Vladislav A. Vashchenko, Andrei Shibkov (Auth.) - ESD Design For Analog Circuits-Springer US (2010) PDF

ESD Design for Analog Circuits
Vladislav A. Vashchenko · Andrei Shibkov
ESD Design for Analog

Circuits
123
Vladislav A. Vashchenko Andrei Shibkov
National Semiconductor Angstrom Design Automation
Semiconductor Drive M/S E-155 1336 Fruitdale Ave. A3
Santa Clara, CA 95052 San Jose, CA 95126
USA USA
[email protected] [email protected]
ISBN 978-1-4419-6564-6 e-ISBN 978-1-4419-6565-3

DOI 10.1007/978-1-4419-6565-3
Springer New York Dordrecht Heidelberg London
Library of Congress Control Number: 2010929991
© Springer Science+Business Media, LLC 2010

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are
not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject
to proprietary rights.
Printed on acid-free paper
Springer is part of Springer Science+Business Media (www.springer.com)

To our families
Preface
This Book and Simulation Software Bundle Project

Dear Reader, this book project brings to you a unique study tool for ESD
protection solutions used in analog-integrated circuit (IC) design. Quick-start
learning is combined with in-depth understanding for the whole spectrum of cross-
disciplinary knowledge required to excel in the ESD field. The chapters cover
technical material from elementary semiconductor structure and device levels up
to complex analog circuit design examples and case studies.
The book project provides two different options for learning the material. The
printed material can be studied as any regular technical textbook. At the same time,
another option adds parallel exercise using the trial version of a complementary
commercial simulation tool with prepared simulation examples.
Combination of the textbook material with numerical simulation experience
presents a unique opportunity to gain a level of expertise that is hard to achieve
otherwise. The book is bundled with simplified trial version of commercial mixed-
mode simulation software from Angstrom Design Automation. The DECIMMTM
(Device Circuit Mixed-Mode) simulator tool and complementary to the book sim-
ulation examples can be downloaded from www.analogesd.com. The simulation
examples prepared by the authors support the specific examples discussed across
the book chapters.
A key idea behind this project is to provide an opportunity to not only study the
book material but also gain a much deeper understanding of the subject by direct
experience through practical simulation examples.
Each section of the book is accompanied by a set of simulation examples directly
related to the main topics addressed in the section.
The examples are not just snapshots of the simulation results. Instead the reader
has an option to use real simulation software tool. This allows the reader a practical
opportunity to understand the ESD simulation examples by interactively studying
the simulation results and changing simulation parameters within the limits of the
trial version.
At the same time, the simulator software does not require any advanced skills
in the technology computer sided design (TCAD) area. The authors of this book
vii
viii Preface
and DECIMMTM simulation tool developers believe that interactive features of the
new tool will allow any electrical engineer or circuit designer to run the simulations
successfully.
Although the free version of the simulation has some limitations compared to the
full version, functionality of the free version is more than sufficient for it to be an
indispensable tool in mastering of the subject of this book.
Nevertheless, the readers who do not want to take an advantage of the simulation
or prefer to postpone the experience can read the textbook as any regular technical
textbook.
The body of the book is prepared absolutely independent from the simulation
examples, which are referred to only at the end of each chapter.
Subject and Purpose of This Book

ESD design for analog circuits is a very diverse and cross-disciplinary field. It
involves an understanding of semiconductor device physics in strong non-linear
operation regime deep knowledge of modern CMOS, BICMOS, and BCD process
technologies, expertise in analog circuit design mixed with understanding of the
product application conditions and specs, and even trends in marketing for analog
integrated components.
Therefore, one of the major challenges accepted by the authors of this book con-
sists in selection of an appropriate depth of material that will provide practical help
in successful ESD design and can still be accessible enough to be used by a broad
audience. This challenge requires both an appropriate simplification to fit the mate-
rial within the limits of a single textbook and a new methodology to present the
material. The methodology is based on combining the phenomenological approach
and simulation results.
The book is organized in a hierarchy from the semiconductor device level to
the product circuit level. In theory, each chapter can be studied independently. The
seven chapters of the book address the following hierarchical levels:
I. Semiconductor Structures
II. Integrated Standard and ESD Devices
III. ESD Clamp Design Principles
IV. ESD Protection Network Design Principles
V. Protection of Signal Path Analog Integrated Circuits
VI. Protection of Power Management Analog Integrated Circuits
VII. System Level and Discrete Component ESD
This book targets all major aspects of ESD protection: device, network, and cir-
cuit design levels, mainly focusing on modern integrated components. System level
and discrete component’s ESD protection is addressed too in the last chapter.
Preface ix
This hierarchy is established in order to enable both sequential and independent

study of the material depending of specific reader expertise and preference.
The authors expect that ESD engineers as well as students would appreciate a
systematic representation of the material in increasing level of complexity. This
approach covers the major background knowledge required for understanding the
material.
Professionals already working in this field with engineering background exper-
tise in device design may find it useful to skip the device chapters and proceed
to the ESD network and analog circuit design material, while experts in circuit
design may benefit from the device physics material that covers physical principles
of conductivity modulation in semiconductor structures.
The authors believe that an option to interactively study the simulation examples
will greatly benefit all readers.
The overall purpose of this book is to help professionals in the field to deal
with the Analog ESD Design issues in their everyday professional work, attack-
ing problems at all hierarchical levels starting from the device ESD level up to
an implementation of integrated self-protecting solutions. The book is supposed to
“arm” readers not only with important practical and technical knowledge, but also
to add a complementary simulation experience that can be further developed with
the light version of new industrial mixed mode simulation software DECIMMTM
from Angstrom Design Automation.
The Book Structure

The book is organized in increasing complexity of the discussed ESD subjects. This
is not complexity in terms of understanding, since that will depend on the level
of expertise – circuit or device design – with which an engineering professional
is approaching this book. Thus, the level of complexity of the material across the
chapters for circuit designers will perhaps be opposite to the level of complexity for
device engineers.
The Chapter 1 to the book below represents a short review of background mate-
rial relevant to the ESD field, describing the authors’ understanding of the field
itself, ESD pulse specification and several other general aspects. At the same time,
the introduction mainly refers the readers to other previously published books in the
field, in order to maximize the space in the current book for material pertaining to
the goal above. As well, the introduction establishes important definitions necessary
for understanding the presented material.
Chapter 2 presents introductory material to the device design field. It provides
a fundamental knowledge of conductivity modulation processes in the elementary
semiconductor structures: p-n, p-i-n, n-p-n, p-n-p, p-n-p-n. The physical processes
in these structures are described on a simplified phenomenological level that is
easy to understand and further supported by rather simple independent simulation
examples.
x Preface
The material presented in Chapter 2 about conductivity modulation in the

elementary structures provides a necessary fundamental background both for under-
standing pulsed safe operating area (SOA) of the standard devices and for the
snapback mode operation of the ESD devices. Chapter 3 covers both these aspects.
It presents pulsed SOA for most typical integrated components in CMOS, BiCMOS,
BCD, SOI, and SiGe process technologies, highlighting the parasitic devices formed
in their structures. Pulsed SOA understanding is critical for the so-called ESD
protection window realized for given circuit pins.
In parallel to each standard device, Chapter 3 deals with most typical basic ESD
protection devices that could be developed as “free” components in the given pro-
cess technology. The important point of low-cost ESD design methodology is the
creation of such “free” ESD devices. In case of “free” devices, it is assumed that the
device can be obtained using only the available mask layers without violation of the
minimum design rules exceeding physical capability of the process tools. Moreover,
the “free” approach challenges self-aligned solutions and solutions that are closely
based upon supported standard devices. In this case, reliability of the ESD devices
can be linked to the corresponding supported device characteristics.
Chapter 4 targets the ESD clamp level using the ESD devices described in
Chapter 3 as building blocks for the variety of ESD protection clamps with desired
characteristics according to pin functionality.
Chapter 5 presents a summary of ESD network design. It explores ways to apply
different ESD clamp solutions to “assemble” an embedded ESD power circuit across
all integrated circuit pins. The ESD network provides an ESD current path between
different pin combinations that, in the case of analog circuits, may involve not only
the ESD pad ring components but the internal functional circuit components as well.
This chapter is a major reference point for practical ESD design of analog circuits
that is further expanded on in Chapters 6 and 7, covering signal path and power
analog circuits.
Chapter 6 is focused on ESD protection for signal path analog applications. It
contains condensed introductory material that highlights design aspects specific to
the ESD on a sub-block level. Signal path circuits and products are mainly rep-
resented in this chapter by high speed, precision and audio amplifiers, interface
application, and digital–analog converters.
The scope of Chapter 7 lies within different power management products.
Analog ESD circuit design is discussed with the examples of dc–dc converters with
integrated power devices, controllers, light management units, and LED drivers.
The final Chapter 8 is focused on the system level of ESD design and ESD for
discrete components. The goal of this chapter is not to describe the system itself,
but instead the integrated components with pins interfacing directly with the sys-
tem terminals. The major aspects are realization of the system level ESD protection
on chip. Due to this being a relatively new topic for ESD design, this chapter con-
tains a much more comprehensive introductory portion in comparison with the well
described across literature sources introduction to ESD in Chapter 2.
Preface xi
Definitions
Conductivity Modulation
Pulsed SOA
ESD Devices
ESD Clamps
ESD Network
Application
Case Studies
Fig. 1 Roadmap for the book composition
Finally, the last section of each Chapter provide a brief description of the sim-
ulation examples directly relevant to the material described in the Chapter. The
examples are available for download from http://www.analogesd.com
Acknowledgments
The authors would like to acknowledge all their colleagues in the field who directly
or indirectly helped with this book.
Vladislav Vashchenko would like to gratefully recognize his closest colleague
and friend Ann Concannon from National Semiconductor for the many years of
close collaboration in the ESD field that directly impacted and contributed to the
creation of this book.
He would also like to acknowledge the contributions of his colleagues Alan
Segerval, Donald Archer, Peter Hopper, Philippe Lindorfer, Alexei Sadovnikov,
Lihui Wang, and David Lafonteese, as well as the many lead circuit designers and
product engineers across the National Semiconductor Corporation with whom he
worked on ESD projects with during the last decade. He is also deeply thankful
for the multiple discussions of ESD subjects during the past 10 years with his col-
leagues in the field: Prof. Gaudenzio Meneghesso from the University of Padova,
Prof. Elyse Rosenbaum from the University of Illinois, Robert Gauthier from IBM,
Gianluca Boselli from Texas Instruments, Vesselin Vassilev from Novorel, Prof.
Juin Lion from the University of Central Florida, Markus Mergens from QPX
GmbH, Prof. Marise Bafleur from LAAS, Theo Smedes from NXP, Kai Goebel
from AMES NASA, and many other colleagues and professional friends from the
ESD Association and Industry. In particular, Vladislav considerably appreciates the
many contributions and support of his current and past direct researcher collabora-
tors and co-authors Dimitri Linten Mirko Scholz, Philippe Jansen, and Steven Thijs
from IMEC, as well as PhD students Nicholas Olson, James Di Sarro, and Blerina
Alija. He is also appreciative of his former mentors Prof. Vladimir Sinkevitch from
Pulsar, Slava Osipov from AMES NASA and Boris Kerner from Daimler AG.
Andrei Shibkov would like to acknowledge many of his former colleagues from
PDF solutions, in particular Carlo Guardiani.
The authors are grateful for analog design expert Vladislav Potanin, National
Semiconductor, who reviewed Chapters 6 and 7 and provided very valuable
comments to the material.
The authors greatly value the support and understanding provided by their
families during the time devoted to this book project.
Finally, both authors would like to emphasize a critical contribution to this book
by Yana Vashchenko, a first-year UCLA student, who spent many days during the
xiii
xiv Acknowledgments
summer of 2009 performing a complete language and style edition for the whole
book. They hope this practice will help her to become a better bioengineer in the
future.
Contents
1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Analog and Digital in Prism of ESD Design . . . . . . . . . . . 1
1.2 Important Definitions . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 ESD Protection Network . . . . . . . . . . . . . . . . . . 4
1.2.2 ESD Clamps . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.3 Absolute Maximum Limits and Pulsed SOA . . . . . . . 7
1.2.4 ESD Pulse Specification . . . . . . . . . . . . . . . . . . 8
1.2.5 Breakdown and Instability . . . . . . . . . . . . . . . . . 9
DECIMMTM Simulation Examples for Introduction . . . . . . . . . . 14
2 Conductivity Modulation in Semiconductor Structures
Under Breakdown and Injection . . . . . . . . . . . . . . . . . . . 15
2.1 Important Definitions and Limitations . . . . . . . . . . . . . . . 15
2.1.1 Basic Semiconductor Structures . . . . . . . . . . . . . . 15
2.1.2 Conductivity Modulation and Negative
Differential Resistance . . . . . . . . . . . . . . . . . . . 17
2.1.3 Spatial Current Instability, Filamentation, and
Suppression . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.4 Snapback Operation . . . . . . . . . . . . . . . . . . . . 20
2.1.5 Notes to the Methodology of Material Presentation
in This Chapter . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Avalanche Breakdown in Reverse-Biased p–n Structure . . . . . 23
2.2.1 Analytical Description of the Avalanche
Breakdown Phenomenon . . . . . . . . . . . . . . . . . 24
2.2.2 Numerical Analysis of the Avalanche Breakdown
in the p+ –p–n+ Structure . . . . . . . . . . . . . . . . . . 26
2.3 Double-Avalanche–Injection in p–i–n Structures . . . . . . . . . 30
2.3.1 An Analytical Description of the Effect . . . . . . . . . . 30
2.3.2 Numerical Analysis for the p–i–n Diode Structure . . . . 31
2.4 Avalanche–Injection in Si n+ –n–n+ Diode Structure . . . . . . . 33
2.4.1 Analytical Approach . . . . . . . . . . . . . . . . . . . . 34
2.4.2 Simulation Analysis . . . . . . . . . . . . . . . . . . . . 36
2.5 Conductivity Modulation Instability in n–p–n Diode Structures . 37
xv
xvi Contents
2.5.1 Conductivity Modulation in a Floating Base

Region: Diode Operation Mode . . . . . . . . . . . . . . 37
2.6 Conductivity Modulation in the Triode n–p–n Structure . . . . . 40
2.6.1 The Case of Grounded Base Breakdown
Operation U EB = 0 (BVCES) . . . . . . . . . . . . . . . 40
2.6.2 The Floating Emitter Case I E = 0 . . . . . . . . . . . . . 41
2.6.3 Avalanche–Injection in a Common Emitter
Circuit: The Case of I B <0 Regime . . . . . . . . . . . . . 41
2.6.4 Avalanche–Injection in the Common Emitter
Circuit with Positive Base Current I B > 0 . . . . . . . . . 47
2.6.5 Avalanche–Injection in the Common Base Circuit . . . . 51
2.7 Avalanche–Injection in PNP Structures . . . . . . . . . . . . . . 52
2.8 Double Injection in Si p–n–p–n Structures . . . . . . . . . . . . 53
2.8.1 Equivalent Circuit . . . . . . . . . . . . . . . . . . . . . 53
2.8.2 Simulation of Conductivity Modulation in
p–n–p–n Structures . . . . . . . . . . . . . . . . . . . . 56
2.9 Spatial Current Instability Phenomena in Semiconductor
Structures with Negative Differential Resistance . . . . . . . . . 59
2.9.1 Current Filamentation at Avalanche–Injection . . . . . . 60
2.9.2 Current Filamentation Effect in Double-
Avalanche–Injection Conductivity Modulation . . . . . . 63
2.9.3 Current Filamentation Effect in the Case of
Double Injection . . . . . . . . . . . . . . . . . . . . . . 66
2.10 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
DECIMMTM Simulation Examples for Chapter 2 . . . . . . . . . . . . 68
3 Standard and ESD Devices in Integrated Process Technologies . . 69
3.1 ESD Specifics in Integrated Process Technology . . . . . . . . . 70
3.1.1 Typical DGO CMOS Process with Extended
Voltage Components . . . . . . . . . . . . . . . . . . . . 70
3.1.2 ESD Specific for BCD and BiCMOS Integrated
Process Flow . . . . . . . . . . . . . . . . . . . . . . . . 83
3.2 Safe Operating Area in ESD Pulse Regime . . . . . . . . . . . . 87
3.2.1 SOA and Current Instability Boundary in Reliability . . . 88
3.2.2 Pulsed SOA for ESD Regimes . . . . . . . . . . . . . . . 90
3.2.3 ESD SOA for Typical Devices in BCD Process . . . . . . 92
3.2.4 Instability Boundary and SOA for ESD devices . . . . . . 96
3.2.5 Physical Limitation of ESD Devices. Spatial
Thermal Runaway . . . . . . . . . . . . . . . . . . . . . 98
3.3 Low-Voltage ESD Devices in CMOS Processes . . . . . . . . . 102
3.3.1 Snapback NMOS . . . . . . . . . . . . . . . . . . . . . 103
3.3.2 FOX (TFO) ESD device . . . . . . . . . . . . . . . . . . 105
3.3.3 LVTSCR and FOXSCR . . . . . . . . . . . . . . . . . . 109
3.3.4 Low-Voltage Avalanche Diodes . . . . . . . . . . . . . . 111
3.4 ESD Devices in BJT Processes . . . . . . . . . . . . . . . . . . 114
Contents xvii
3.4.1 Integrated NPN BJT Devices . . . . . . . . . . . . . . . 114

3.4.2 Bipolar SCR . . . . . . . . . . . . . . . . . . . . . . . . 116
3.5 High-Voltage ESD Devices in BCD and Extended
Voltage CMOS Processes . . . . . . . . . . . . . . . . . . . . . 117
3.5.1 LDMOS-SCR and DeMOS-SCR Devices . . . . . . . . . 118
3.5.2 Lateral PNP BJT Devices . . . . . . . . . . . . . . . . . 121
3.5.3 High-Voltage Avalanche Diodes . . . . . . . . . . . . . . 126
3.6 Dual Direction Devices . . . . . . . . . . . . . . . . . . . . . . 127
3.6.1 Dual-Direction Device Architecture in CMOS Process . . 128
3.6.2 High-Voltage Dual-Direction Devices . . . . . . . . . . . 131
3.6.3 Dual Direction ESD Devices Based upon Si–Ge
NPN BJT Structure . . . . . . . . . . . . . . . . . . . . 134
3.7 ESD Diodes and Passive Components . . . . . . . . . . . . . . . 139
3.7.1 Forward-Biased ESD Diodes . . . . . . . . . . . . . . . 139
3.7.2 Passives . . . . . . . . . . . . . . . . . . . . . . . . . . 142
3.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
4 ESD Clamps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
4.1 Active NMOS Clamp . . . . . . . . . . . . . . . . . . . . . . . 158
4.2 Low-Voltage Clamps with Internal Blocking Junction
Reference or dV/dt Turn-on . . . . . . . . . . . . . . . . . . . . 161
4.2.1 Snapback NMOS Clamps . . . . . . . . . . . . . . . . . 161
4.2.2 Transient-Triggered PMOS Clamp . . . . . . . . . . . . 167
4.2.3 10 V FOX Snapback Device . . . . . . . . . . . . . . . . 169
4.2.4 LVTSCR and FOX-SCR Clamps . . . . . . . . . . . . . 171
4.2.5 High Holding Voltage LVTSCR Clamps . . . . . . . . . 172
4.2.6 Triggering Characteristics Control in SCR Clamps . . . . 176
4.3 Voltage and Current Reference in ESD Clamp . . . . . . . . . . 182
4.3.1 Low-Voltage Clamps in BiCMOS process technology . . 183
4.3.2 NPN Clamps with Voltage Reference . . . . . . . . . . . 185
4.4 High-Voltage ESD Devices . . . . . . . . . . . . . . . . . . . . 188
4.4.1 20 V NPN with Blocking Junction Internal Reference . . 189
4.4.2 NPN Clamp with External Lateral Avalanche
Diode Reference . . . . . . . . . . . . . . . . . . . . . . 190
4.4.3 SCR-Based High-Voltage Clamp . . . . . . . . . . . . . 190
4.4.4 Lateral LPNP Clamp . . . . . . . . . . . . . . . . . . . . 190
4.4.5 Mixed Device-Circuit Dual Mode Solutions . . . . . . . 191
4.5 The Concept of Self-Protection . . . . . . . . . . . . . . . . . . 196
4.5.1 Device-Level Self-Protection . . . . . . . . . . . . . . . 196
4.5.2 Array-Level Protection . . . . . . . . . . . . . . . . . . 198
4.6 ESD Protection of Ultra High Voltage Circuits . . . . . . . . . . 200
4.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
xviii Contents
5 ESD Network Design Principles . . . . . . . . . . . . . . . . . . . . 213

5.1 Rail-Based ESD Protection Network . . . . . . . . . . . . . . . 215
5.1.1 Rail Based and Local ESD Protection . . . . . . . . . . . 215
5.1.2 Rail-Based ESD Protection Using Snapback Clamps . . . 217
5.1.3 Rail-Based ESD Protection Using Active Clamps . . . . 219
5.1.4 Specific of Active Clamp Design in BiCMOS Processes . 223
5.1.5 Bipolar Differential Input Protection . . . . . . . . . . . 232
5.1.6 Bipolar Output Protection . . . . . . . . . . . . . . . . . 234
5.1.7 CMOS Input and Output Protection . . . . . . . . . . . . 235
5.1.8 Array-Level Consideration . . . . . . . . . . . . . . . . 237
5.1.9 Concept of Two-Stage Protection . . . . . . . . . . . . . 240
5.2 Local Clamp-Based ESD Protection Network . . . . . . . . . . . 247
5.2.1 Local ESD Protection . . . . . . . . . . . . . . . . . . . 247
5.2.2 Serial Data Line Pin Case Study . . . . . . . . . . . . . . 248
5.2.3 Erase Pin Protection in EEPROM . . . . . . . . . . . . . 250
5.2.4 Local Protection of the Internal Pins . . . . . . . . . . . 253
5.2.5 Local Protection of the High-Speed I/O pins . . . . . . . 256
5.3 ESD Network for Multiple Voltage Domains . . . . . . . . . . . 258
5.3.1 Multiple Voltage Domains . . . . . . . . . . . . . . . . . 258
5.3.2 Protection of Multiple Voltage Domains with
Single Active Clamp Network . . . . . . . . . . . . . . . 260
5.3.3 Local Bi-directional ESD Protection of
Differential Input . . . . . . . . . . . . . . . . . . . . . 261
5.4 ESD Network Simulation with ESD Compact Models . . . . . . 263
5.4.1 Compact Model for Snapback NMOS and PMOS
Devices . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
5.4.2 Snapback LVTSCR Model . . . . . . . . . . . . . . . . . 265
5.4.3 Extended Voltage Snapback Compact Models . . . . . . 265
5.4.4 High-Voltage Open Drain Circuit Analysis . . . . . . . . 270
5.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
6 ESD Design for Signal Path Analog . . . . . . . . . . . . . . . . . . 281
6.1 Amplifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
6.1.1 Amplifier Product Families and Specifications . . . . . . 282
6.1.2 ESD Solutions for Amplifiers . . . . . . . . . . . . . . . 288
6.1.3 Bipolar Output High-Voltage Audio Amplifiers . . . . . . 290
6.1.4 Bipolar Output Protection in Low-Voltage Amplifiers . . 292
6.1.5 Input Protection . . . . . . . . . . . . . . . . . . . . . . 293
6.1.6 CMOS Output . . . . . . . . . . . . . . . . . . . . . . . 295
6.2 Digital-to-Analog and Analog-to-Digital Converters . . . . . . . 296
6.2.1 Functional Blocks for High-Speed DAC . . . . . . . . . 297
6.3 High-Speed Interface IO pins . . . . . . . . . . . . . . . . . . . 301
6.3.1 Interface Analog Products . . . . . . . . . . . . . . . . . 301
Contents xix
6.3.2
Cable Discharge Event Test Procedure for
Integrated Circuits . . . . . . . . . . . . . . . . . . . . . 302
6.3.3 ESD Protection of Interface Pins with CDE
Requirements . . . . . . . . . . . . . . . . . . . . . . . 305
6.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
7 Power Management Circuits’ ESD Protection . . . . . . . . . . . . 317
7.1 Power Management Products . . . . . . . . . . . . . . . . . . . 318
7.1.1 Power Management Products and ESD Challenges . . . . 318
7.1.2 Integrated DC–DC Converters and Controllers . . . . . . 321
7.1.3 Integrated Power Arrays . . . . . . . . . . . . . . . . . . 323
7.2 Low-Voltage Power Circuit ESD Cases . . . . . . . . . . . . . . 338
7.2.1 LV Power Switching Blocks . . . . . . . . . . . . . . . . 338
7.2.2 Step-Down DC–DC Converters . . . . . . . . . . . . . . 340
7.2.3 Local Snapback Protection of LV Switch Pin . . . . . . . 343
7.3 ESD Protection of Integrated High-Voltage Regulators . . . . . . 347
7.3.1 Asynchronous Integrated Buck Regulator Case . . . . . . 347
7.3.2 Synchronous Regulators . . . . . . . . . . . . . . . . . . 351
7.4 Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 357
7.4.1 Asynchronous Buck-Boost (SEPIC) Controller . . . . . . 359
7.4.2 Synchronous Buck Controller . . . . . . . . . . . . . . . 362
7.5 Light Management Units and LED Drivers . . . . . . . . . . . . 364
7.5.1 Analog LED Technology . . . . . . . . . . . . . . . . . 364
7.5.2 LED Drivers . . . . . . . . . . . . . . . . . . . . . . . . 366
7.5.3 Light Management Units . . . . . . . . . . . . . . . . . 367
7.6 A Few More Case Studies . . . . . . . . . . . . . . . . . . . . . 374
7.6.1 Power Array–ESD Clamp Interaction . . . . . . . . . . . 374
7.6.2 Nepi–Nepi Transient Latch-Up Scenario . . . . . . . . . 377
7.6.3 CDM Case of the High-Voltage Pin Protection . . . . . . 380
7.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
8 System-Level and Discrete Components ESD . . . . . . . . . . . . 395
8.1 System-Level Specifications and Standards . . . . . . . . . . . . 396
8.1.1 Meaning of ESD Robust System . . . . . . . . . . . . . 396
8.1.2 System-Level ESD Pulse and Model . . . . . . . . . . . 400
8.1.3 Transient Latch-up During a System-Level Event . . . . . 405
8.1.4 System-Level Protection Components . . . . . . . . . . . 408
8.2 On-Wafer Human Metal Model Measurements . . . . . . . . . . 409
8.2.1 On-Wafer HMM Tester and Equivalent Circuit of
the Pulse . . . . . . . . . . . . . . . . . . . . . . . . . . 410
8.2.2 HMM-HBM Component Correlation . . . . . . . . . . . 412
8.3 On-Chip Design for System-Level Pins . . . . . . . . . . . . . . 416
8.3.1 Examples of Circuits with System-Level Protection . . . 416
8.4 Hot Swap and Hot Plug-in . . . . . . . . . . . . . . . . . . . . . 422
xx Contents
8.4.1 The Concept of Two-Stage SCR ESD Devices . . . . . . 422

8.5 System-on-Package (SOP) Protection . . . . . . . . . . . . . . . 428
8.6 ESD Robustness of Discrete Components . . . . . . . . . . . . . 429
8.6.1 Discrete Components in High Reliability Systems . . . . 429
8.6.2 ESD Requirement for Discrete Components . . . . . . . 429
8.6.3 Preliminary Numerical Analysis for Devices with
Defects and the Two-Transistor Model . . . . . . . . . . 432
8.6.4 Experimental Evaluation of Discrete Components
Robustness . . . . . . . . . . . . . . . . . . . . . . . . . 436
8.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 442
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 447
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455
Chapter 1
Introduction
Readers proficient in the ESD field may decide to skip this introductory section.
The section briefly summarizes the background directly relevant to the ESD field.
This is done to facilitate the study of the following chapters for those who are
not directly involved in the field or want to refresh the most important aspects of
the knowledge. A broader spectrum of ESD background material is brilliantly cov-
ered in many books and reviews [1–7] written in the field, as well as in EOS/ESD
Symposium Proceedings [8]. The purpose of this section is to summarize the general
ESD approach to integrated components design and provide condensed reference
material related to the ESD pulse specification, and standards. Finally, the most
important definitions used across this book are established.
1.1 Analog and Digital in Prism of ESD Design

Historically, integrated circuits can be roughly subdivided into analog and digital
categories. These classifications are currently used in the industry and are based
upon functionality and design principles of ICs. Even in the case of complex
mixed-signal circuits, different analog and digital domains can be identified and
corresponding ESD methodology specific to either analog or digital circuits can be
applied for each specific domain. The book title emphasizes “analog” as a major
focus of the presented material due to several reasons. However, practically the
material can be treated as universal ESD design guidelines for mixed-signal cir-
cuits, perhaps excluding practical examples of the most scaled down digital CMOS
processes of 90–32 nm. At the same time, the digital domains in analog IC com-
ponents built using 0.13–1 μm process technologies are well covered in Chapters
4 and 5 by active clamp solutions. These circuit blocks represent digital interface,
digital control pins, and digital domains.
From an ESD design point of view, the digital circuits and digital circuit domains
in mixed-mode circuits are different from analog circuits. The most straightforward
way to identify the type of circuit is by the type of signal transferred through the
circuit pins.
V.A. Vashchenko, A. Shibkov, ESD Design for Analog Circuits, 1

DOI 10.1007/978-1-4419-6565-3_1, C Springer Science+Business Media, LLC 2010
2 1 Introduction
From the practical ESD design point of view, the difference between analog and
digital designs is often reflected in the ESD pad ring circuit block. In case of the
fully digital circuit, ESD protection is usually designed in the pad I/O (input/output)
and power domains. In this case, the periphery of the digital circuit is expected to
be practically fully isolated from the internal circuit.
When I/O and ESD library is designed and validated it can support wide vari-
ety of internal digital blocks with different functionalities. In this case, beyond the
ESD/IO library creation itself, the major focus of ESD chip design is to ensure
proper pad ring layout design that accounts for the voltage drop on metallization
busses, number of clamps, and RC timers and make sure that current path for every
pin-to-pin combination is addressed. As well, at appropriate voltage limitation, the
current path is always expected to be confined within the ESD pad ring network. In
the case of high-pin count digital ICs the problem complexity requires automated
tools in form of ESD rule checkers.
The above is not intended to trivialize digital ESD design. Digital ESD design has
its own complexity. For example, one of the most significant challenges is covering
the specifics of high-pin count digital IC products, especially in the case of CMOS
processes scaled down to 90–32 nm gate dimension. In this case, one of the major
challenges is CDM (charged device model) pulse protection of the large form factor
packages. There are many challenges in protection of high-speed and RF I/O pins,
as well as system-level protection.
However, in the case of widely used analog 0.5 μm process technologies the
digital pin protection hardly presents any real challenge. The methodologies for
such protection are very well established.
The major reason for discussion of digital vs. analog circuitry is to emphasize
the specific of ESD protection approach and solutions.
In the case of digital design, one can expect that the ESD network building
blocks in pad ring design will typically provide only a single connection to the
internal circuit. This connection can be relatively easy to analyze in the ESD
current path analysis. At the same time, latch-up isolation is automatically pro-
vided by ESD library solutions with no or minimized interaction of the ESD
and I/O circuit with other circuit nodes due to appropriately designed guard
rings.
In opposite, analog ESD design guarantees no such condition. In general, the ana-
log ESD protection design can be expected to account for the fact that the analog
pad might have multiple connections to the internal circuit node. The connections
can be realized both directly and indirectly by the coupling through the power
components, for example, large drain–gate capacitance of multimillimeter width
NLDMOS power array.
Moreover, most of these internal circuit nodes will generally have an unknown
transient bias and current conditions for different pin-to-pin zap combinations. This
makes it rather difficult to guess at what point of safe operating area the active
device connected to the pin is. During ESD pulse, power components of the analog
circuit may switch to on-state during ESD event and conduct a substantial amount
of current during a part of the ESD pulse. A “sneak” current path can be formed
1.1 Analog and Digital in Prism of ESD Design 3
between the ESD clamp and internal circuit components depending on the layout of
the circuit, especially in case of high-voltage device.
We would like to use this type of differentiation to underscore the essence of sep-
aration between the digital and the analog ESD designs. The fact that a substantial
amount of analog products can contain digital interface pins and domains does not
change this approach to ESD design principles.
Thus, in the case of digital ESD design, the ESD approach can be unified and
formalized to be somewhat independent from the internal circuit blocks as long as
their type is identified and appropriate ESD or ESD/IO library cells selected with
ESD pad ring and created in accordance with the library guidelines. In opposite, in
the case of analog ESD design, a practical ESD design should often be customized
for new circuit changes, taking into account possible alternative scenarios for ESD
current conduction through the internal circuit. These potentially new scenarios
should be then taken into account for both ESD network design and ESD clamp
choice.
Thus, one of the most critical features of the analog vs. digital designs is the
ESD protection network and the internal circuit in general cannot be separated. This
creates a need for the ESD engineer to understand the analog circuit at much greater
depth.
Another important distinct feature from ESD perspective is the product pin
count. Digital products often have hundreds of pins, while some small form fac-
tor analog products might have as few as three to five pins. This fact automatically
brings into consideration the critical issue of the space used on the chip for ESD
protection.
In the case of digital circuit with a high pin count, the chip periphery is relatively
large and a distributed active clamp solution is relatively spatially optimal, usually
requiring space, the size of one pad at each pad. Therefore, perhaps over 90% of the
digital circuits are protected by the distributed active clamp solution.
A different situation can be often found in case of analog circuits. There are many
examples of small-pin-count analog ICs where over 50% of the silicon die space is
taken up by ESD clamps. In this case, space saving and small footprint solutions
can provide a major impact on the product cost.
At the same time, the voltage tolerance for the ESD clamps that corresponds
to the digital signal levels is rather low. In the case of mature process technolo-
gies, this makes even the local protection option relatively easy to address. The
opposite situation is found in case of high-voltage analog products where an opti-
mal high-voltage solution might be rather hard to find without changes to process
technology.
The purpose of this book is to help professionals in the field to deal with the
analog ESD design issues in their everyday professional work, attacking problems
at all hierarchical levels starting from the device ESD level up to an implementation
of integrated self-protecting solutions. The book is supposed to “arm” readers not
only with important practical and technical knowledge but also to add a complemen-
tary simulation experience that can be further developed with the new mixed-mode
simulation software DECIMMTM from Angstrom Design Automation.
4 1 Introduction
1.2 Important Definitions
1.2.1 ESD Protection Network

Practically every book written in the ESD field well describes the nature of phenom-
ena behind ESD charge accumulation and the discharge events [1–4]. Over time the
specs created for industrial application move further away from real events that may
cause charging or the objects that can provide discharge through IC pins.
Therefore, to avoid redundancy and save space in this book, the starting point for
the ESD subject is presented from a slightly different angle.
Without loss of generality, we will assume the following approach to the subject.
An integrated circuit represents an object that contains both internal circuitry com-
ponents and external pads. A discrete component can be treated as a particular case
of IC where internal circuitry is represented by a single or few devices packaged
together. Similarly a system can be considered as a combination of circuit blocks
that contain mounted, integrated, and discrete components with external terminals.
The pads are bonded to package and pins. The pins could be package pins, system
socket connectors, or device leads.
The ESD protection task will be further treated simply as an added ability to the
integrated components or systems, or discrete components to withstand certain ESD
pulse specification.
In this case, the ESD protection capability is treated as a part ICs or system
specification similar to other normal specification parameters for the internal circuit
performance and reliability.
In most practical cases of analog integrated components design, this added ability
to withstand certain level of ESD pulse spec is achieved through a co-design of the
internal circuitry and the ESD protection network. Therefore, an idea of separation
between the ESD network and internal circuitry as well as corresponding require-
ments for non-conflicting existence of the ESD protection solution with the circuit
performance is perhaps irrelevant.
Nevertheless, the ESD co-design goal is to minimize the impact of the ESD
network components on the ideal circuit performance that could have been accom-
plished if ESD performance was not specified.
Similar reasoning is relevant for system-level ESD where normal product
operation during ESD and electrical overstress (EOS) events may be required.
System-level ESD protection is presented in Chapter 8.
Thus, a problem of ESD development can be formulated as the implementation
of an embedded capability of integrated products toward withstanding standard ESD
tests according to defined specification.
From a circuit design point of view, this essentially means an extra functionality
of the integrated product circuit at some specific high-current pulse conditions in
addition to normal operation specification targets.
There are several ways to realize such capability. One of the major principles is
to implement additional peripheral ESD protection networks connected in parallel
with the original functional circuit blocks. However, analog ESD design often relies
1.2 Important Definitions 5
on the self-protection capability of the internal components. The self-protection of

the internal components can be exploited both up to the level of the full ESD current
conduction and using much smaller current levels through the internal circuit to
enable a two-stage protection. In principle, a pin-specific combination of the above
measures is usually used.
Thus, in general, the product can be considered as a superposition of inter-
nal functional circuit blocks and additional pulsed power circuit dealing with
ESD current. This circuit can partly use the components of the functional circuit
blocks. This pulsed power circuit can further be understood as the ESD protection
network.
The major functionality of the ESD protection network is to provide a high-
current path with appropriate voltage limitation in case of ESD discharge applied
for every pin-to-pin combination. Under the appropriate voltage limitation require-
ments, a limitation of the voltage in the ESD time domain below the pulsed
safe-operating area (SOA) is understood for the device. Thus, the ESD protection
network should not only provide the ESD current path but also limit the volt-
age below absolute maximum rating conditions realized at each pin. The pulsed
absolute maximum rating provided by the circuit at each pin is rather complex
figure of merit that depends on time domain, rise time, maximum ratings of
the devices connected to the pin, coupling of their control electrodes, and other
factors that could be related to a parasitic current path realized in the actual
layout.
Thus, the task can be converged into co-design and embedding of the ESD
network. ESD network design principles are discussed in Chapter 5.
As has been already mentioned, this network can partly be realized using the
self-protection capability of the active circuit components. However, major building
blocks are the ESD protection clamps, connected by an appropriate metallization
routing for given pulsed current level.
An ESD protection clamp usually presents itself as self-triggered in the high-
current state. The self-turn-off event of clamps usually occurs due to discharge of
the ESD pulse. There are two major categories of ESD protection clamps. The first
uses the RC network to control on and off state conditions for the active devices. In
this case, the high-current device is fully controlled in high-current mode and can
be turned off by the RC network.
Another major class of ESD clamps is based upon avalanche breakdown and
snapback ESD devices that involve different conductivity modulation mechanisms
in order to achieve high-current conditions. Unlike active clamps, the turn-off of
ESD devices with conductivity modulation usually cannot be practically controlled
by a driver circuitry. The ESD device operates in the conductivity modulation mode
up until the voltage conditions are changed simply due to the end of the ESD dis-
charge, or the voltage level will be naturally reduced below the holding voltage of
the snapback components or below the breakdown voltage.
The major principles for ESD clamps design are discussed in Chapter 4. The
ESD devices are presented in Chapter 3. The principles of conductivity modulation
are presented in Chapter 2.
6 1 Introduction
1.2.2 ESD Clamps

The building blocks of the ESD protection network are the ESD clamps. Under ESD
clamp we will further understand a simple circuit that provides pulsed ESD current
path under certain conditions. These conditions are generally defined as achieving
the critical voltage level or a fast rise time. In most sophisticated cases, the clamp
operation can be controlled or enabled by the additional control electrode. ESD
clamps are discussed in Chapter 4.
The clamp can be designed using integrated components operating in normal
operation mode of monopolar or bipolar current conduction. This principle is the
basis of active clamps discussed in Chapter 4.
One of the most critical requirements of the ESD protection network toward com-
petitive advantage of the product is to occupy the feasibly smallest space on the chip.
Therefore, a proper way to achieve such capability is the implementation of small
footprint device-level solutions, taking an advantage of several isothermal conduc-
tivity modulation mechanisms that can be realized in such devices. Thus, for analog
design, the most useful clamps are the ones reversibly operating in high injection
and breakdown modes.
Typical device-level ESD solutions are essentially pulsed power devices with
some biasing components that are specifically designed to work in the high
injection, breakdown, and conductivity modulation modes.
ESD clamp development involves a rather broad set of cross-disciplinary tasks in
the fields of circuit design, physics of semiconductor devices, physical ESD clamp
design with non-linear physics aspects of the operation, effects in interconnects,
materials, and topological array problems to balance the current density distribution.
Finally, the most important necessary expertise involves a deep understanding of
process technology including the options and the limits.
Meantime, as has been shown in [9] and will be particularly illustrated in detail
in Chapter 2, the physical phenomena responsible for conductivity modulation used
in ESD devices are very limited to several physical mechanisms that can be ana-
lyzed using elementary diode, triode, and thyristor semiconductor structures. These
practically useful mechanisms include avalanche breakdown, avalanche–injection,
double-avalanche–injection, and double injection. At this point of ESD field evolu-
tion, there were no reports of using other conductivity modulation mechanisms for
practical ESD design. Similarly, due to short ESD pulse time domain, the conductiv-
ity modulation mechanism based upon thermal carrier generation plays a secondary
role, mainly limiting the high current capabilities of the ESD device itself [9]. This
limitation is discussed in Chapter 3.
The operation in these breakdown conditions generates high current densities
presented in Table 1.1. In case of avalanche breakdown, the carriers are generated in
the region with a rather high electric field and are separated proving practically no
mutual space charge neutralization. In this case, relatively low current density can
be achieved. In case of avalanche–injection, the mutual space charge compensation
of the carriers generated in the avalanche region and injected creates more favor-
able conditions for high current density. The highest current levels are provided in
Table 1.1 Conductivity modulation mechanisms realized in ESD devices for high current density
Conductivity modulation Typical lateral current

mechanism Typical ESD devices density (mA/μm)
Avalanche breakdown Avalanche diodes; blocking 0.01–0.1

junctions, PMOS, PNP
Avalanche–injection Snapback NMOS; NPN, 0.1–3
field oxide devices
Double-avalanche– P–i–n, M–i–n diodes 0.1
injection
Double injection LVTSCR, SCR, bipolar 10–100
SCR, LDMOS-SCR
the case of double injection, when the avalanche multiplication is supressed in the
structure (see Chapter 2).
One of the challenges in creating an appropriate ESD device is balancing the
conductivity modulation current density inside the device to achieve an appropriate
current level. In this case, a linear width scaling of the ESD current can be expected
in the device. Since some of the mechanisms provide a positive feedback that in
general can result in uncontrollable current density increase, a negative feedback
loop should be implemented in the device to limit the current density below critical
limits.
ESD devices are discussed in Chapter 3, while basic principles of the conductiv-
ity modulation in semiconductor structures are discussed in Chapter 2.
1.2.3 Absolute Maximum Limits and Pulsed SOA

As has been stated above, on a formal circuit design level upon given spec, the ESD
functionality is achieved by implementation of a pulsed power circuit embedded
into the normal circuit using shared and dedicated components. This partly virtual
secondary pulsed power circuit should not only provide a pulsed power operation
for the pin-to-pin current path scenario but also limit voltage below the damage level
of the internal circuit blocks.
Thus, realization of the voltage waveform parameters in the ESD time domain is
a major challenge, considering the absence of the limitations in normal circuit oper-
ation regime. Thus, the ESD protection network should provide voltage waveforms
that will limit the voltage below the absolute maximum limits on the pin, but at
the same time will not cause false turn-on during normal operation. Practically, this
means that the turn-on voltage should be realized above absolute maximum limits
specified for the circuit operation modes in all data sheet conditions including the
temperature range.
These two limits define the so-called ESD protection window. While the lower
limit of ESD protection window can be sourced from the circuit application spec,
the upper limit is a more complex matter. In principle, a preliminary idea about
8 1 Introduction
absolute maximum limits for the particular pin can be “extracted” from the pulsed
SOA of the devices directly connected to the pin. However, the problem is that most
of electrical design rules for process technology usually do not provide pulsed SOA
in the ESD time domain.
Another major issue is unknown coupling effect of the control electrodes of the
devices connected to the pin in the ESD test modes. In this case, the uncertainty
is related to the selection of the SOA regime for identifying the ESD protection
window.
There are no universal recipes for identifying the ESD protection window
for analog circuit pins before the actual ESD tests. One of the most productive
approaches is based upon transmission line measurements (TLP) for pulsed SOA.
These SOA aspects are discussed in Chapter 3 for most typical devices realized in
BCD (bipolar CMOS DMOS) process technology.
1.2.4 ESD Pulse Specification

During the era of integrated product development, various specifications for ESD
pulse have been created and became industry or custom standards. The detailed
description of the ESD pulse characteristics and the equivalent circuits can be found
in the introductory chapters of numerous popular books in the field, for example
[1–4], original publications [8], and standard documents, for example [10–13].
The standards for the human body model (HBM) pulse are as follows: ESDA
(ANSI) STM5.1-2001; JEDEC JESD22-A114-E; IEC 613240-3-1; AEC Q100-002
REV-D; and EIAJ ED-4701/304. Similar standards can be found for the machine
model (MM), the charged device model (CDM), and the ESD system level [13].
The consolidated summary of most commonly used ESD pulses is presented in
Table 1.2. Standard non-system-level corporate requirements for the HBM, MM,
and CDM ESD passing levels are 2 kV, 200 V, and 1 kV, respectively.
Table 1.2 The most common examples of non-system-level ESD pulse parameters
Peak current to pulse Rise time/pulse

ESD pulse voltage ratio (A/kV) width (ns)
Human body model 0.67 2–10/150

Machine model 17.5 2–10/66–90
Charged device model 9 0.25/2
EN (IEC) 61000-4-2 3.75 0.2/50–150
From an ESD device design challenges perspective, there are two major spec
types: the non-system level and the system level. The non-system-level spec usu-
ally targets some minimal requirements to protect the integrated circuit component
while incorporating it into the system, packaging, handling, and electrical tests. For
example, the HBM ESD pulse spec for the non-system level requires withstanding
some pulsed current level of ∼1.33 A with the rise time of ∼2–10 ns (Table 1.2).
Respectively, the test is conducted in the conditions of the unpowered circuit.
In opposite to the non-system packaged specs, system-level specs are targeting
protection of some circuit pins under normal operation conditions. In addition to
this, usual system-level requirements target much higher current levels that can be
practically realized in a non-ESD protected environment. The complexity of the
system-level protection problem is related to a possibility of transient latch-up.
Transient latch-up can be realized in case if ESD clamp provides a holding volt-
age lower than the power supply voltage under the minimum holding current below
the current that can be provided by the power supply.
Due to fast rise time, in most cases ESD pulse automatically provides the
conditions for pure electrical turn-on. Electrical current instability is initiated in
quasi-isothermal conditions and provides further triggering of a high-current con-
ductivity modulation state. In most practical cases, the lattice temperature change
can be neglected before the triggering due to uniform current distribution and short
time before the triggering.
After the switching, heat dissipation becomes significant. However, the heat dis-
sipation scenario is significantly different from dc operation. Due to rather short
pulse duration, the heat dissipation is realized in a rather small area of few microns
in the vicinity of the device’s active region. Thermal heat dissipation (as well
as electro-mechanical stress and dielectric breakdown) and backend limits pro-
vide physical limitations for ESD device operation. These effects are illustrated in
Chapter 3.
1.2.5 Breakdown and Instability

In the majority of cases, ESD events in semiconductor structure are close to adi-
abatic conditions. In spite of high current density, due to very fast switching time
and fast ESD pulse (∼100 ns), the heat generation is confined in rather local area
(Fig. 1.1).
This phenomenon eliminates a significant amount of physical effects related to
the carrier generation in the drift regions, epi-layers, substrate, and corresponding
current instability phenomena.
The thermal effects further play an important role in the final irreversible
catastrophic phenomena when ESD device fails.
The phenomena of irreversible breakdown, burnout, thermal and isothermal
instability, and current filamentation are often mentioned in respect to semiconduc-
tor device failure in the case of both ESD and EOS events.
In this book, the notion “breakdown” preserves its elementary primary sense.
The breakdown is treated similar to the effect in p–n semiconductor junctions as a
process of sharp current increase that is caused by the carrier generation current.
Practically for the ESD field, the only important cases are isothermal avalanche
and tunneling breakdown.
10 1 Introduction
Fig. 1.1 Simulation example: local heating in collector region of NPN transistor caused by 100 ns
stress pulse
In avalanche breakdown of the reverse-biased p–n junction, current I increases

with voltage U increase according to the known empirical formula [14]:
1
I∼ n ,
1 − U UBR
where n = 4–6 for Si material and UBR is the avalanche breakdown voltage.
The dependence I(U) in this case of “classical” avalanche breakdown is strictly
monotonic.
A different kind of breakdown could be observed in a case where injection
and conductivity modulation are involved. In this case, current increase can create
complex S-shaped I–V characteristics of the devices.
The current instability can be defined as a process of uncontrollable current
increase due to voltage decrease.
This phenomenon in general may or may not result in irreversible device failure
due to the appearance of some limiting factors. In the case of the ESD device, the
current instability is used to create low dissipated power conditions in conductivity
modulation mode. To avoid high-amplitude filaments, the local current density is
limited internally on the device. Implementation of such limitation is an essential
part of ESD device architectural design.
Thus, the physical definition of the breakdown is just a sharp current increase
under positive differential conductivity. In other words, the breakdown itself does
not include any positive feedback. The major breakdown mechanisms in semicon-
ductors are the avalanche and the thermal, although several additional mechanisms
can be found, for example, the breakdown related to the change in trap charge state
or dielectric breakdown.
In opposite to the breakdown, different electrical and thermal instabilities [9]
include a positive feedback. The thermoelectrical instability phenomena in semi-
conductor devices are rather complex. To provide a “quick start” in understanding
the physical sense of these phenomena, we present below the example of thermal
instability in semiconductor structure.
The best example is the avalanche–injection conductivity modulation in NPN
structures discussed in Chapter 2. An accurate analytical description for conduc-
tivity modulation mechanisms is rather complex. An example of the analytical
description for the case of thermal breakdown is used below to explain the current
instability phenomenon itself. Often this thermal instability case can be responsible
for the physical limitation of the current level provided by ESD device [9].
In the example of bulk semiconductor structure, if at a constant voltage U << UBR
the devices are heated by some external current source, then from some temperature
level the current through
the
sample
will grow sharply according to an exponential
dependence I ∼ exp −EG kT , where EG , T, and k are the energy of band gap, lat-
tice temperature of semiconductor material, and Boltzmann’s constant, respectively.
This thermogeneration process for carriers is considered as a thermal breakdown.
It is assumed that on a uniform sample of bulk semiconductor with length l and
area S the voltage
U is supplied. The current density j through the sample is equal to
j = σ E = σ U l, where E is the electric field in the sample and σ is the conductivity
of semiconductor material. The conductivity
of semiconductor material in this case
can be expressed by σ ∼ exp −EG kT .
Then, the generated heat per volume unit is equal to σ E2 . It is also further
assumed that heat dissipation is provided by exchange with ambient space of fixed
temperature TS . Then, heat dissipation from the surface will be proportional to
(T − TS )β , where T is the temperature of semiconductor region and β = 1–2. The
heat balance equation is given by

EG U 2
exp − = K (T − TS )β , (1.1)
kT l2
where the proportionality factor K depends on physical and geometrical parameters

of the device structure. In a certain range of values of U this equation can have two
solutions: T1 and T2 (T1 , T2 > TS ). The solutions correspond to two different current
values I1 and I2 .
The I–V characteristic of such a structure is no longer as simple as avalanche
breakdown and has an S-shaped view (Fig. 1.2b).
The physical meaning of the I–V characteristic can be explained as follows. On
the initial region “OC,” the current increase obeys dependence close to Ohm’s law.
12 1 Introduction
Fig. 1.2 I–V characteristic at avalanche breakdown of p–n junction (a) and at the thermal
instabilities in semiconductor resistor (b)
An appreciable deviation from linear dependence begins when the heating reaches
a higher level. The conductivity increase is connected with intensive thermogenera-
tion of electrons and holes in the sample. Insignificant increase of the voltage on the
thermal breakdown region “CA” results in a sharp increase of heat generation.
Up to a certain limit, the heat generation is balanced by an increase in temperature
since the heat dissipation is proportional to (T − TS )β . However, at some U > UCR ,
the exponential increase of the heat generation can no longer be compensated by
heat dissipation. In this state, (1.1) has no solutions and the semiconductor sample
has no stationary states, respectively. This means that in the voltage source regime
at U > UCR , a sample will be uncontrollably self-heated up to its destruction.
A similar loss of thermal stability or an uncontrollable process of transition into
a new state usually means instability. In this case, the instability is of a thermal
nature. From Fig. 1.2b, the stable states of a sample in the case of thermal instability
can be achieved only at corresponding voltage decrease U. If the circuit provides a
sufficient load resistance, then the thermal instability may finally evolve into a stable
state that corresponds to an I–V characteristic with negative differential conductivity
(NDC) (Fig. 1.2b, state R).
The load characteristic CR in the case of such a device in circuit operation could
be called snapback.
Current instability in real device structures is not always easy to interpret due to
an additional spatial current instability phenomenon.
Since in the case of current instability the local current density dependence
upon voltage has negative differential resistance, a typical consequence is current
stratification into filaments and hot spots.
These phenomena may or may not damage the device depending on the local
damping implemented in the ESD device.
This damping can often act up to a certain electrical power limit followed by the
hot spot formation in the silicon surface. Usually, the hot spot is a narrow region with
a dimension of a few micrometers with concentrated current and elevated tempera-
ture. In bipolar transistors, similar current localization results in thermal breakdown.
In this case, a sharp temperature increase may result in metallization melting.
COLLECTOR-EMITTER VOLTAGE (V)
Fig. 1.3 Simulation example: snapback I–V in NPN
The example of thermal instability is provided above to deliver a primary

understanding of processes that are extremely widespread in real device structures.
Alternatively to the analytical methods the thermal instability for given NPN
structure can be obtained by numerical simulation results (Fig. 1.3) thus enabling
additional way of learning.
A few summary points emphasized at the end of the introduction.
In opposite to digital, the analog circuits or analog domains in mixed-mode
circuits are rather diverse in terms of the voltage tolerance and signal spec
and often require small pin count. In this condition, practically every pad may
require a separate ESD protection clamp limited by rather critical small footprint
requirements.
This task could be solved in most cases only on the device level and thus requires
understanding in three major key areas:
(i) The principles of operation of ESD devices operating in the breakdown and
conductivity modulation conditions (secondary breakdown)
(ii) ESD clamp and ESD network design based upon these devices
(iii) Application of the ESD network to analog circuits and network–ESD circuit
interaction
These key areas are addressed step by step in this book to enable a proficient,
practical ESD design for those who are involved in the field.
Of course, once this understanding is established, a more detailed experience in
the field is required based upon real product case studies. Nevertheless, we believe
that in-depth knowledge about ESD components, networks, and practical exam-
ples for the analog circuit ESD protection (Chapters 6, 7, and 8) should bring ESD
expertise to a new useful level and will be a relevant contribution to the field of ESD.
14 1 Introduction
DECIMMTM Simulation Examples for Introduction

To download a trial version of the numerical simulation software and request
an electronic license key please visit http://www.analogesd.com
To download libraries with simulation examples for this chapter please visit
http://www.analogesd.com/Chapter1.html
List of examples is subject to change.
Example 1.1 Simulation of the avalanche breakdown I–V in elementary quasi-1D

structures
Example 1.2 Simulation of the S-shaped current instability in 2D NPN device
Example 1.3 Simulation of the transient thermal-coupled mixed-mode solution for
2D NPN, demonstrating hot spot area during ESD pulse confined
within 1 μm
Chapter 2
Conductivity Modulation in Semiconductor
Structures Under Breakdown and Injection
2.1 Important Definitions and Limitations

Understanding semiconductor structure operation under ESD pulse conditions at the
physical level is critical for successful protection circuit design. In spite of the use of
a variety of ESD protection devices and clamps for analog circuit protection, there
are several fundamental physical effects taking place during a high-current ESD
event. These effects are discussed in this chapter.
At high-current operation in short-pulse conditions and under ESD pulse in par-
ticular, semiconductor devices obey certain principles. These principles are confined
within a rather limited set of basic isothermal conductivity modulation mechanisms
[9]. Physical limitation of both the electrical safe operating area (SOA) in standard
(provided by the process technology) devices and the ESD snapback devices is the
result of the same fundamental principles.
This introductory chapter provides a quick start summary required for under-
standing the most important conductivity modulation effects utilized on the device
level.
2.1.1 Basic Semiconductor Structures

Implementation of ESD devices in the given semiconductor process technology usu-
ally involves significant complexity. This complexity is the result of combination of
different factors, for example, multiple complex implant profiles, multiple regions
and interfaces, topology, and 3D effects. In some cases, several alternative ESD cur-
rent paths can be found in real ESD cells depending on the current level. These
paths can change, depending on the transient conditions specific to given ESD pulse
type and cell topology. One of the most typical examples of this occurrence is in
the NPN BJT device in BCD process technology, where competition between the
vertical and the lateral current transport can be observed at a typical combination of
structure parameters.
Nevertheless, based upon present understanding [9], it can be concluded that all
these processes can be identified within in a very limited number of avalanche–

16 2 Conductivity Modulation in Semiconductor Structures
injection effects. Moreover, most of these effects can be treated as adiabatic due to
a relatively short ESD pulse duration with the generated heat dissipated within very
small region (∼1 μm). This condition significantly simplifies analysis of the ESD
device operation.
In spite of a superposition or a “domino” effect observed in a real life scenario,
the avalanche–injection effects can be understood based upon a type of conductivity
modulation. The type of conductivity modulation can be linked to a conductivity
modulation in one of the elementary semiconductor structures. These structures are
analyzed in this chapter.
Thus, the major goal of this chapter is to provide an understanding of cur-
rent conduction in elementary structures under avalanche–injection conditions. This
understanding is directly useful for analysis of the processes in real ESD devices and
clamps in pulsed operation mode.
Since this book is written in the field of analog IC design, the major focus is
confined to Si and Si–Ge semiconductor materials used in today’s microelectron-
ics for analog circuits. However, similar principles can be applied to compound
semiconductor materials GaAs and GaN [15–22].
Due to short-pulse conditions, certain assumptions are made about the electrical
regime in ESD time domain. In particular, most of the phenomena can be treated as
an isothermal case. Another important assumption is silicon semiconductor material
parameters. Under these assumptions there are only five elementary semiconductor
structures that can provide a practically relevant current density level under condi-
tions of breakdown and injection. These elementary or primitive structures are as
follows:
(i) p–n structure operating at reverse bias in avalanche breakdown mode

(ii) n–p–n structure operating in avalanche–injection conditions
(iii) p–n–p structure operating in avalanche–injection conditions
(iv) p–n–p–n structure operating in double-injection conditions
(v) p–i–n structure operating in double-avalanche–injection conditions
To stay within practicality, these five elementary structures are presented and ana-
lyzed with heavily doped contact regions of type p+ or n+ . Respectively, the internal
p, i, or n-base or drift regions are assumed to have significantly lower doping, in
comparison with the contact regions. Consequently, the conductivity modulation
analysis is accomplished within the area outside of the contact regions which remain
quasi-neutral, i.e., the dominant part of the space charge region is confined within
the bulk of the lower doped drift regions.
Thus, in reference to p–n structures, the structures p–n–n+ , p+ –p–n+ , or p+ –p–n–
+
n are assumed to be practically relevant cases, where the acceptor and donor levels
in the corresponding regions are NAp+ ∼ NDn+ >> NAp , NDn .
There are several other structures that are potentially useful for ESD. For exam-
ple, Schottky diodes operating in avalanche–injection conditions, tunneling (Zener
diode) structure, discharge gaps, polymer suppressors. However, these devices are
2.1 Important Definitions and Limitations 17
not discussed in this book due to a particular focus on the most practical analog ESD
design applicable to the integrated circuits.
2.1.2 Conductivity Modulation and Negative Differential

Resistance
Under conductivity modulation in the semiconductor device structure we will sim-
ply understand a non-linear behavior of the structure that results in a significant
change of conductivity due to voltage or current applied to the structure contact
regions. This is a definition that is based upon the external structure parameters.
Similarly, conductivity modulation can be defined through internal structure param-
eters as a change in the conduction properties of particular structure regions as
a result of change in carrier balance due to avalanche and injection processes.
The conductivity modulation in multiterminal structures is the result of an internal
avalanche–injection process initiated by change in the applied current or volt-
age. This is emphasized in opposite to the change of the structure conductivity
due to injection/extraction of carriers from the base contact region or creation
accumulation/depletion regions by the field control electrode.
From a practical ESD device design point of view, the most interesting cases are
related to a non-linear change of the conductivity that results in negative differen-
tial resistance effect. The negative differential resistance can be simply defined as a
decrease in the voltage drop on the structure contacts with a current increase through
the contacts. This effect is usually observed as a formation of S-shaped I–V char-
acteristics, although in some cases the saturation resistance of the drift and contact
regions compensates the internal negative differential resistance.
A different known type of non-linearity is the negative differential conductance
that corresponds to formation of N-type I–V characteristics. In this case, the voltage
increase results in current decrease. This phenomenon is practically observed under
formation of the traveling Gunn domain in GaAs [9] or in case of trap charge change.
However, it appears that at this point this effect is barely useful for ESD protection
solution design.
In opposite, ESD applications of the negative differential resistance effect are
extremely useful. This physical effect provides a power-efficient way to dissipate
ESD current under a lower voltage drop on the protected pin. As a result, the lower
dissipated power on the ESD device both provides an opportunity for a small foot-
print solution design and simultaneously enables realization of the desired voltage
waveform during ESD pulse. The last is achieved by the low clamping voltage that
guarantees survival of the internal circuit components while keeping the electrical
pulsed regime within the absolute maximum ratings.
In the case of a real circuit, full S-shaped I–V characteristics are hard to observe
during the negative differential resistance (NDR) effects due to a finite load resis-
tance. Therefore, a part of NDR might be represented by the load resistance. For
example, in the case of a 50 transmission line pulse (TLP) system [2], a part of
the S-shaped I–V characteristic will be created with a corresponding 50 load resis-
tance. A 1.5 k HBM waveform tester may enable observation of more details on
the S-shaped I–V characteristics, even demonstrating partly real negative differential
resistance of the structure itself.
Thus, in the case of an operation with finite small load resistance, the device
achieves negative differential resistance conditions at some critical voltage, caus-
ing a self-turn-on or self-triggering into the corresponding high-current state. This
phenomenon is usually referred to as a snapback among ESD application engineers
and circuit designers. This term will further be used across this book to identify the
associated phenomena described above.
At the same time, according to Vashchenko and Sinkevitch [9], as it will be
demonstrated below, the real physical processes responsible for such behavior repre-
sent a very limited set of mechanisms based upon a positive feedback loop between
avalanche and injection phenomena.
Beyond the practical use of negative differential resistance in ESD devices to pro-
vide ESD current discharge and desired voltage waveforms, this phenomenon has a
much more general nature. Practically, as shown in [9], together with thermal con-
ductivity modulation effect, this phenomenon is responsible for physical limitation
of electrical regimes of all real structures. In particular, it is essentially responsi-
ble for physical limitation of the ESD device operation in pulsed conditions. These
aspects will be addressed in Chapter 3.
2.1.3 Spatial Current Instability, Filamentation, and Suppression

Negative differential resistance is often linked in literature with the current insta-
bility and spatial current instability phenomena. Indeed, on the circuit level, the
current instability in the structure can be intuitively understood as an uncontrol-
lable current increase when some critical voltage level is applied to the structure.
Specifically, such an effect can be observed during snapback depending on the load
resistance.
Most of the ESD pulses produce a current level of ∼1–10 A, while most of the
structures are capable of providing current densities of ∼0.1–10 mA/μm. Therefore,
the discussion of ESD devices that can be used in the clamps automatically involves
largely distributed objects designed as multifinger devices and arrays.
At the same time, one of the important non-linear physical effects observed in
distributed structures with structure-level positive feedback is directly related to a
different type of current instability [9, 23–27]. This type of instability is the insta-
bility of the uniform current density distribution across the structure width. Thus,
in the case of negative differential resistance, the current distribution in the suffi-
ciently wide semiconductor structure is in general spatially unstable and results in
the formation of current crowding, current redistribution, and current filamentation
effects.
The spatial current instability evolution scenario depends upon the electri-
cal regime and load characteristics. However, the typical spatial dimensions for
the structure’s positive and negative feedback processes determine the possible
solutions for non-uniform current states. Spatial instability of the current density
distribution may result in the formation of rather complex non-uniform states with
a current density that may significantly exceed the initial uniform current density
[26]. A simplistic example of such a state is the isothermal current filament.
Formation of spatially non-uniform states in distributed semiconductor structures
is one of the interesting phenomena of non-linear physics [26, 28, 29]. In this chap-
ter, numerical simulation results will be presented for several elementary structures
in Section 2.8 in order to demonstrate the general physical principle behind the
phenomenon.
One of the critical targets for successful design of the ESD device with certain
positive feedback is to implement structure-level negative feedback. This negative
feedback needs to be engaged at a certain current level and limit the current den-
sity to below the safe conditions. For example, in grounded gate snapback NMOS,
discussed below, the positive feedback realized due to avalanche–injection conduc-
tivity modulation in the parasitic n–p–n structure is compensated on the structure
level by the negative feedback provided by current saturation in the drain ballasting
region. This device will be discussed in detail in Chapter 3.
In general, the negative differential resistance effect leads to the spatial current
instability in the device. The spatial current instability results in current filamen-
tation. The filamentation itself is not necessarily an effect that immediately results
in irreversible damage of the device. Moreover, the snapback NMOS at high ESD
current represents an operation mode in which the device supports the current fila-
ment regime visible at a small current level. In this case the filament amplitude is
limited by the drain ballasting region below the safe limits in pulsed regime. At the
maximum level of current, the operation mode practically corresponds to a filament
mode with a width equal to the structure width. At the same time, burnout of such
a device is the result of conductivity modulation of the ballasting region that in its
own turn results in much greater amplitude filament that usually almost immediately
locally melts the device.
Thus, while in an appropriate ESD device the peak current density is limited,
the pulsed operation of the ESD device is expected to be fully reversible. An oppo-
site situation is realized in a standard device. With some exceptions, the standard
devices for the given integrated process will not provide a reversible operation in
negative differential resistance conditions and following snapback mode of oper-
ation. Therefore, specifically such an effect limits the pulsed safe operation area
(see Chapter 3) due to local burnout of the device structure [9] in the formed high-
amplitude current filament. Implementation of the device-level negative feedback is
one of the practical measures used to enable self-protection capability or improve
pulsed SOA.
Since the pulsed current density of the semiconductor device in conductivity
modulation mode can be up to 10 mA/μm, the physical limit of the operation
regime is not necessarily always related to the semiconductor part of the device.
The backend metallization and contacts should also be taken into account, especially
in high-voltage devices. This mainly occurs due to electromigration limits because
of high dissipated power and thermal heating of the backend region. For the given
process technology, the electromigration limits are usually a part of the process tech-
nology electrical design rules. The correlation between the electromigration rules
and electromigration in the ESD pulse domain can be experimentally measured. To
estimate backend limits in ESD pulse conditions the conventional electromigration
limits can be multiplied by some process-specific factor. Usually, in 0.13–1 μm
process technologies, with a good safety margin, this factor is above 40. A typical
ESD device of standard 2 kV with peak current of 1.33 A usually requires W =
50–1000 μm total width depending on the current density provided by the
implemented elementary semiconductor structure. In this case, adequate backend
metallization routing can be easily achieved following certain rules for clamp design
(see Chapter 4).
A very important consideration should be emphasized at the end of this sec-
tion. It is the fact that local structure-level current instability involves a necessity
of implementation of device-level local current density suppression below a certain
safe level. An intuitive attempt to use backend ballasting for this purpose is usually
not successful if the current distribution in the finger of the device results in uncon-
trollable current density increase. Backend ballasting is used only to balance turn-on
or current across multiple fingers [30].
Thus, the local positive feedback in the structure caused by conductivity mod-
ulation should be suppressed by some local negative feedback on the device level,
although the spatial parameter for negative feedback can often be bigger. This mea-
sure is achieved by implementation of current ballasting regions inside the device
that provide local current saturation effects. Total current limitation by the external
circuit components is usually unsuccessful due to local current filamentation inside
the device structure.
Practically, current density suppression in the ESD device is usually done so
that all possible current filament states have limited amplitude, which ensures that
heating during ESD pulsed domain (∼100 ns) is sufficiently below the critical level
for structure damage.
However, this reliable operation does not absolutely guarantee that such a result
is achieved in the event of much longer electrical overstress pulses or in dc condi-
tions. Thus, ESD operation in the negative conductivity regime can be considered
as an operation in the filament state with limited amplitude that in normal con-
ditions would melt the device locally or result in other irreversible accelerated
degradation effects. Physical limitations of the ESD device usually occur at a higher
current level, when conductivity modulation of either the ballasting or contact
regions results in formation of corresponding high-amplitude filaments followed by
unlimited injection from the contacts.
2.1.4 Snapback Operation

As mentioned above, ESD device operation with negative differential resistance
on a certain load resistance is usually called snapback in practical electrical
engineering. In the case of snapback, the ESD device turns on or triggers on into
a high-conductivity or high-current state depending on the load characteristics in

the circuit.
One of the most popular ways to measure snapback characteristics is the
transmission line pulse (TLP) setup that usually provides a 50 load [2].
In principle, there are no published standards that would require any TLP charac-
teristics to be achieved by the integrated product. Historically, the capability of the
HBM, MM, and other ESD testers provided mainly the results of pass–fail measure-
ments, while the voltage waveforms of the ESD pulse operation were impossible to
measure. Therefore, for device-level ESD design and debugging, a less expensive
and more informative pulse method was required to compare different ESD devices
and provide fast debugging of the IC’s ESD failures.
The TLP is widely used currently, although a new state-of-the-art method that
allows complete waveform measurements for an HBM pulse with a ∼1.5 k load
has been recently developed [32].
One of the major purposes of TLP evaluation is not only to confirm the maximum
current level provided by the clamp but also to deliver insight on the most probable
voltage waveform realized by the ESD device. This is achieved by extraction of
several major figures of merit from the TLP characteristics.
From a physical point of view, the ESD device in snapback mode operates similar
to a voltage-controlled switch with resistive load. The first pair of parameters, crit-
ical triggering voltage VT1 and triggering current IT1 , is used as figures of merit for
turn-on of the device into snapback. These parameters are critical in providing both
device turn-on within the so-called ESD protection window and guaranteeing that
normal circuit pin operation will not result in the transient latch-up phenomenon.
The next important figure of merit parameter is the holding voltage VH . This
parameter depends on the load resistance and snapback voltage. Thus, for example,
when using a 50 TLP system, the measured holding voltage generally can be
higher than the real minimum holding voltage for the example when measured using
a waveform HBM system with a load resistance of 1.5 k.
After turn-on into the high-current state, the device provides a certain on-state
resistance due to the internal positive feedback and a saturation region that deter-
mines the voltage waveform in the ESD pulse domain at a high current level.
Usually, this parameter is an important practical criterion. For standard package
level specifications it can be defined, for example, as 1.33 A peak current, at HBM
conditions of V2 kV .
Finally, at a certain point, physical limitation of the device operation results in
irreversible changes to the device structure. This regime is usually named IT2 and
VT2 (Fig. 2.1)
The parameters IT2 and VT2 in general cases cannot be seen from the pulsed
I–V characteristic itself, because it may keep the same trend even after irreversible
changes. A separate functional test is usually required to establish reversible opera-
tion in the device. In most TLP systems, this functional test is usually implemented
by a simple leakage current measurement at a given voltage level with assumed para-
metric failure criteria, for example, a leakage deviation of one order of magnitude
from the original level.
PULSED CURRENT (A)

PULSED CURRENT (A) 4 4
IT2
2 2
IT1
0 0
0 VH VT2 25 VT1 10 –1 10 0 10 1 10 2
VOLTAGE (V) LEAK.CURR. (@20V) (nA)
Fig. 2.1 Typical TLP snapback characteristics with major figures of merit indicated on the plot
In most TLP systems, the leakage voltage is measured between each pulse
(Fig. 2.1). For example, in the case of data for the 20 V snapback device presented
in Fig. 2.1, the left plot for pulsed I–V demonstrates no peculiarity associated with
the irreversible failure that is already observed in the leakage current functional test
plot at an IT2 current level of ∼2.5 A.
TLP characteristics are very convenient for comparative analysis. They are
widely used across this book both to represent the device parameters, pulsed SOA,
and for debagging of the analog circuit product pin characteristics.
A new methodology combines the pass–fail test with waveform capture [33]. It
is presented below as well, as an important future ESD design tool that enables a
much faster time-to-market.
2.1.5 Notes to the Methodology of Material Presentation

in This Chapter
The elementary semiconductor structure presented below can be used to control the
pulsed characteristics of the developed device and clamp in a wide range to achieve
a desirable voltage waveform at ESD pulse conditions.
The purpose of the following section is to provide a condensed, thorough
background for understanding of conductivity modulation in the semiconductor
structure. This knowledge, combined with a certain level of physical understanding,
is directly relevant to both ESD devices and clamp design.
This goal can be achieved in several different ways. In most books published
in this field, authors prefer presenting analytical descriptions of the phenomena or
extensive experimental data.
We believe that in most practical cases, the analytical approach requires a signif-
icant trade-off between the complexity of real objects and simplifications required
for the analytical description itself. Although such an approach is fit for teaching on
2.2 Avalanche Breakdown in Reverse-Biased p–n Structure 23
a physical level, there is little practical use of such material for a wide audience of
engineers and designers, due to the diminished practical relevance of the simplified
structure parameters and the physical models of semiconductor material and spe-
cific devices. Thus, this approach is usually applicable within a rather narrow range
of parameters and provides limited accuracy of conclusions that greatly depend on
the expertise and experience of the theoretician.
Without questioning the usefulness of analytical approach in general, we believe
that a much broader pool of readers with more practical needs will appreciate an
alternative approach specially “engineered” for this book, which takes advantage of
modern physical processes and device simulation tools.
This alternative approach combines empirical description with rather accurate
numerical simulation results.
The advantage of such an approach lies in the ability to avoid excessive simplifi-
cation on the semiconductor level in simulation, while still presenting the data using
accurate representations of the drift-diffusion model for carrier transport equations
and semiconductor material parameters for silicon. At the same time, a complete
visualization of the intrinsic structure state is fully possible.
Namely, this simulation approach is followed consistently across the book
chapters in each of the suggested methodology levels.
However, there are some important limitations involved. In this chapter, when
describing the elementary conductivity modulation structures, we will limit our-
selves within the following constraints:
(i) Silicon semiconductor material parameters

(ii) Isothermal cases specific to ESD pulse domain
(iii) Selecting the doping profiles and structure dimensions close to parameters of
the integrated structures in 0.5 μm process, as this process technology is most
widely used today for analog design
(iv) Staying within practical limits of the current densities in conditions where
conductivity modulation will be limited below 10 mA/μm2 level
In spite of these assumptions, both the description and the simulation results
could be easily expanded for broader cases.
How successful this approach will prove to be will be known through the future
feedback of those who read this book. Meanwhile, possible gaps in understanding
can be covered by the recommended books in the field where the analytical approach
to learning is presented in greater detail [1–9, 14, 26].
2.2 Avalanche Breakdown in Reverse-Biased p–n Structure

In ESD devices, avalanche breakdown of the blocking junction is a physical phe-
nomenon that is typically used to initiate the self-triggering of a voltage referenced
in snapback ESD devices. Avalanche diodes can be used directly to enable voltage
limitation by the avalanche current generation. The alternative way to initiate snap-
back is to implement a triggering circuit that is often based on avalanche breakdown
voltage reference as well.
In the case of snapback devices, conductivity modulation and current instability
follow the avalanche breakdown mode. The same phenomenon is responsible for
the physical limitation of pulsed SOA of standard semiconductor devices supported
in the given process technology in the ESD time domain.
In this section, avalanche breakdown is first discussed using simplified analyt-
ical models. Then, a simple quasi-1D numerical simulation analysis using a 2D
simulation tool is presented in order to demonstrate the internal effects of the
structures.
This is the only place in the entire book where the analytical approach is
compared to the methodology of combining the phenomenological approach with
numerical simulation, as implemented in this text.
For ESD devices, the reverse-biased p–n junction presents a direct interest in
several aspects. First of all, this structure forms a blocking junction that provides
a corresponding high-voltage tolerance with low leakage of the self-triggered ESD
device. The positive feedback and current instability inside the ESD device can usu-
ally be initiated at a relatively low level of the avalanche current. Therefore, the
contribution of an additional voltage drop in avalanche mode is rather low. This
automatically provides a correlation between the avalanche breakdown voltage and
the snapback triggering voltage in a more complex device.
Typically, the high-current operation of the avalanche diode device can produce
current levels of 0.1 mA/μm. This makes the avalanche diode directly useful not
only as a source of reference voltage and current, but as a clamp. In principle, the
total width for a packaged-level spec for lateral avalanche diodes of the clamp is
rather large: ∼10,000 μm. Therefore, a more practically efficient way to imple-
ment the ESD protection circuit is the use of avalanche diodes in the second stage
of the two-stage ESD protection network. In this case, the avalanche diode is typi-
cally designed to provide current of 1–50 mA that can be achieved at a relatively
small size of laterally or vertically self-aligned or non-self-aligned devices (see
Chapter 4).
2.2.1 Analytical Description of the Avalanche Breakdown

Phenomenon
This section provides an analytical description of avalanche breakdown.
Generation of the electron–hole pairs due to impact ionization in the semicon-
ductor structure becomes evident at an electrical field of E > 104 V/cm. At some
critical level of electric field, this process has an avalanche nature. The impact ion-
ization coefficients α n for electrons and α p for holes are used for analysis of the
avalanche breakdown process. The coefficient denotes the number of ions generated
by the charged particle on the pathway unit. The coefficient is a strong function of
the electric field. It is given by exponential dependencies [14]:

E
αn , αp ≈ exp −
E0
In the case of 1D asymmetric reverse-biased p+ –n junction, the coefficient depen-
dency may be written in the following way, without taking into account the diffusion
and recombination in the balanced equation:
djn djp
− = = αn jn + αp jp , (1.1)
dx dx
where jn = qnvn , jp = qnvp are the electron and hole current densities, vn , vp are the
electron and hole drift saturation velocities that are independent from E in a strong
field: vn ≈ vp = 107 cm/s. The boundary conditions for the p+ –n junction are given
by
jn (x = 0) = 0; jp (x = W) = js , (1.1a)
where W is the width of the space charge region (SCR) of the p+ –n junction and jS is
the saturation current density. The current discontinuity equation is added to (1.1):
j = jn + jp = const. (1.2)
From the solution of (1.1) and (1.2) at αn = αp = g, the current density is
js
j= W . (1.3)
1− 0 g dx
The ratio of the total current density to the saturation current density j/jS is
denoted as the multiplication coefficient M:
1
M= W . (1.4)
1− 0 g dx
W
For the conditions of high avalanche breakdown j → ∞, M → ∞, 0 g dx → 1.
Usually, as a criterion of the avalanche breakdown, a certain given integral value
δ << 1 is accepted:
W
g dx = δ (1.5)
0
Since the total current density j(E) is a sharp function of E, most of the avalanche
current in the p+ –n junction is generated in rather narrow layer <<W. Therefore,
the breakdown condition (1.5) can be approximated by
g (EBR ) = δ. (1.6)
Thus, the defined electric field EBR can be adopted as the avalanche breakdown
value. This value is changed slightly at and δ variation.
For analytical estimation of the I–V characteristic in the avalanche breakdown
mode, the following assumptions are used in addition to the assumptions above:
(i) Uniform distribution of the breakdown across the uniform p–n junction area
with the lateral dimension of the p–n junction significantly higher than its
thickness and thus the 1D problem is automatically justified.
(ii) The drift velocity v does not depend upon the electric field starting from the
levels of electric field above E > 104 V/cm.
(iii) E(x) is a smooth function of coordinate, with the simple function of ionization
coefficient g ≈ Em .
Under these assumptions, the basic equation for I–V dependence is given by the
Poisson equation for the 1D case:
dE ρ
=− , (1.7)
dx ε
where ρ is the charge density and ε is the dielectric constant of semiconductor
material.
In case of developed avalanche breakdown (jS > j), the space charge density in
the n-region is given by
j
ρ = qND − , (1.8)
v
where ND is the donor concentration, j is the electron current density in the n-region,
and the avalanche generation is concentrated in the multiplication layer <<W.
From [14], the voltage on the p–n structure U is presented by the expression

j −1
U = UBR 1 − , (1.9)
jN

where UBR = εEBR 2 2qND and jN = qND v is the critical current density at total
space charge density equal to zero.
At j → jN , the I–V dependence
I(U) of the p–n junction aspires to satura-
tion. At this condition, dI dU > 0 , i.e., the p–n junction has positive differential
conductivity.
2.2.2 Numerical Analysis of the Avalanche Breakdown

in the p+ –p–n+ Structure
In the above section, an example of the analytical approach is demonstrated to
illustrate the most basic phenomenon. It provides an in-depth understanding of
the physical nature of the phenomenon that is very useful in practice. However,
the application of the analytical approach to integrated ESD device engineering
is not straightforward. For example, the application of the analytical approach to
practical design of the ESD avalanche diode in the given integrated process tech-
nology with complex profiles, multiple interface regions, and reduced surface field
(RESURF) effects would require numerous simplifications and assumptions. Thus,
this approach would hardly guarantee appropriate accuracy.
The practical alternative to this method is a numerical simulation of devices
using finite element models (FEM). Numerical simulation is actually based on the
analytical model for carrier transport of semiconductor devices.
Numerical FEM methods can be used to simulate devices with arbitrary geometry
and doping profiles in 2D or 3D and can be calibrated and solved with an accuracy
that can hardly be achieved with the analytical approach, while also providing fast
simulation speed.
This approach does not imply that numerical simulation requires no expertise.
Previously, exclusively TCAD engineers ran the industrial TCAD software, which
involved writing substantial amount of code and extensive knowledge of the tool-
specific software limitations and peculiarities. Today, with a new interactive tool
DECIMMTM from Angstrom Design Automation [31], the learning overhead is sig-
nificantly reduced. The DECIMMTM tool is designed to be user-friendly and usable
by any electrical engineer, both device and circuit designers, and to provide results
that can be directly used in engineering work. This in particular is exploited across
this book by the complementary examples referred to at the end of each chapter.
In the case of real device, the p–n structure in integrated process technology
usually has a more complex architecture. At least two heavily doped contact regions
are required: p+ and n+ regions. Then, at least one lightly doped region is required
to tolerate practical levels of operation voltage not limited by the tunneling effect.
An example of simulation structure used to analyze avalanche breakdown phe-
nomena in the silicon p–n junction is presented in Fig. 2.2. The solution for different
doping levels in the p-drift region (or p-base) (Fig. 2.3) can be obtained within sec-
onds providing an accurate solution for initial avalanche breakdown voltage at low
current level, and the saturation I–V dependence. Moreover, the solutions for all
internal characteristics, for example, electric field and carrier densities (Fig. 2.4a–c),
can be easily plotted from the simulator output for different current levels, helping
to reveal the internal processes in the device.
The particular device in this example has an L = 3 μm space between the anode
(positively biased) and cathode (negatively biased) contacts with a W = 1 μm width.
Further in this chapter, we will deviate from the commonly used approach of
calling the electrodes of the diode structure that are subjected to the positive and
negative bias anode and cathode, respectively. This is done to avoid misunderstand-
ing and confusion of avalanche diodes and other diode structures, for example, with
forward-biased diodes.
Current density is normalized for a 1 μm structure depth. Three uniform doping
profiles are defined in the silicon region of the structure: p+ region with acceptor
concentration NAp+ =1020 cm–3 defined from y = 0 to position y = 0.5 μm, n+ -
region with donor concentration NDn+ =1020 cm–3 from y = 1.5 μm up to the end of
a)
b)
Fig. 2.2 Diode p+ –p–n+ structure with additional drift regions (a) and 1D cross section through
the center of the structure showing detailed doping profile (b)
the device. Uniform doping across the whole structure p– with NAp is variable and
defines the drift or a base region of the structure with a length of LP = 1.
In agreement with the analytical approach in the simulation results (Fig. 2.3),
a decrease of the breakdown voltage is observed with an increase of the drift
region acceptor doping. The avalanche breakdown current is generated upon the
predicted exponential dependence up to a certain current level, when current sat-
uration effect in the drift region becomes dominant. With further current increase,
Fig. 2.3 Calculated I–V characteristics of the structure in Fig. 2.2 of different NAp doping values
a current instability provides for the appearance of the S-shaped I–V portion with
negative differential resistance (Fig. 2.3). According to the analysis of the internal
parameters distribution, it can be easily concluded that the observed current insta-
bility in this avalanche diode example is the result of p-base region conductivity
a) b)
c)
Fig. 2.4 Calculated distributions of the electric field (a), electrons (b), and holes (c) for the
structure with NAp =1017 cm–3 at various bias and current levels
modulation after the level of injected carriers exceeded the level of acceptor
doping.
Thus, the process in this structure can be easily analyzed in greater detail by
comparing the magnitude of the electric field and the carrier density (Fig. 2.4a–c).
Cutline profiles are taken for the device with NAp =1017 cm–3 in the middle of the
structure in the Y-direction for different bias and current levels. At a low current
level of ∼10–10 A (bias ∼14 V), the electric field profile forms a triangular shape at
the p–n+ junction peaking at Y = 1.5 μm.
As the current increases up to ∼10–5 A, it is supported by the exponential depen-
dence of the avalanche multiplication peaking at Y = 1.5 μm. After the current
saturation at voltage level of ∼24 V, the electric field increase begins in the base
region, thus providing a higher integral voltage drop. The carriers injected into the
base region provide partial space charge neutralization, which results in the increase
of the electric field magnitude in the p-base. This occurs up to the formation of
the trapezoidal electric field distribution in the p-base at a voltage level of ∼38 V
(Fig. 2.4a).
The high electric field provides carrier multiplication across the whole p-base
region, while the injected carrier density is above the NAp =1017 cm–3 level.
This effect produces a flow of avalanche-generated electrons and holes drifting in
the strong electric field in opposite directions across the p-base. At high current
level, the space charge of the carriers is compensated, providing the electric field
distribution with two maxima at 32 V (Fig. 2.4a).
Thus, at the high current density typical for ESD operation, the conductivity of
the p-region is modulated and the structure provides negative differential resistance.
As it will be demonstrated in Section 2.8, the current filamentation effect further
results in high-amplitude current filament formation. Unless additional contact sat-
uration region is provided in the structure, the density of the carriers injected into the
filament will reach the doping level in the contact regions, resulting in tremendous
locally generated power. These phenomena, in most practical cases, are responsible
for local burnout of the reverse-biased p–n diodes.
2.3 Double-Avalanche–Injection in p–i–n Structures
2.3.1 An Analytical Description of the Effect

At first, operation of this structure in reverse bias is somewhat similar to the opera-
tion of the reverse-biased p–n junction in avalanche breakdown mode. However, the
presence of the i-region creates an interesting positive feedback effect. This effect is
essentially similar to the p-base modulation in the avalanche diode discussed in the
previous section, but observed at a much lower current level.
An analytical description of the structure with simplified parameters is presented
below, followed by more detailed numerical simulation results with exact model
parameters.
2.3 Double-Avalanche–Injection in p–i–n Structures 31
On the empirical level, the conductivity modulation of the i-region in the p–i–n
structure is rather easy to understand.
At reverse bias, the main part of the potential difference U is supplied to the
i-region of the p–i–n structure. At U ≈ UBR , the electric field in the i-region is
constant and impact ionization begins uniformly through the i-region.
The generated electrons in the i-region are accumulated near the n-region and the
generated holes near p-region. The result of this distribution is the increase of the
electric field near the junctions and decrease in the middle of the structure due to
space charge mutual neutralization.
The negative differential resistance of the p–i–n structure is easy to prove using
the analytical approach; the p–i–n diode I–V characteristic can be determined by the
following equations. The electric field in the i-region at I = δI = 0 is given by
E = EBR + δE (x, δI) . (1.10)
For the initial part of the characteristic at δE/EBR 1, the changing electric
field δE is limited by the breakdown condition:
W
g (E) dx = const. (1.11)
0
By substitution of (1.10) into (1.11), expanding g(E) and limiting by the first
three terms
W W
1 g BR
δU = δEdx = − (δE)2 dx. (1.12)
2 g BR
0 0

Since g BR ≡ dg dE EBR and g BR ≡ d2 g dE2 EBR are positive, the increment
of the voltage drop δU is negative. This simple fact means that at least the internal
differential resistance of the p–i–n diode at avalanche breakdown is negative.
2.3.2 Numerical Analysis for the p–i–n Diode Structure
As it was mentioned before, the numerical solution for the p–i–n structure is similar
to the high-current operation of the reverse-biased p+ –n–n+ or p+ –p–n+ structure
at the current level that corresponds to the density of injected carriers above the
base (drift) region doping level. However, at low current levels, the electric field
distribution is different and the saturation region is not present. Essentially, from the
conductivity modulation point of view, the p–i–n structure is the particular case of
p+ –n–n+ or p+ –p–n+ structure where the base doping level approaches zero.
A simulation study of the p+ –i–n+ is presented in Figs. 2.5 and 2.6. This
structure is geometrically identical to that described in Section 2.2.2 with NAp
Fig. 2.5 Numerical analysis of the conductivity modulation in the p–i–n structure. I–V character-
istics
a) b)
c)
Fig. 2.6 Numerical analysis of the conductivity modulation in the p–i–n structure; distributions
for the electric field magnitude (a), electron (b) and hole density (c) at selected bias and current
levels
2.4 Avalanche–Injection in Si n+ –n–n+ Diode Structure 33
set to 1014 which is equivalent to intrinsic Si from the practical point of view.
Simulated I–V curve is shown in Fig. 2.5. The distribution of electric field mag-
nitude, electrons, and holes in the middle of the structure in the Y-direction is
summarized in Fig. 2.6a–c. With the voltage increase, the original electric field
distribution in the structure’s i-region is uniform. At the breakdown voltage level,
the energy of the carriers is sufficient to cause the impact ionization effect that is
initiated across the whole i-region. This forms a flow of electrons and holes in
the opposite direction. At some critical level, the quasi-neutral area in the mid-
dle of the i-region provides for a decrease of the electric field due to partial
space charge compensation of the injected carriers. At the same time, the total
avalanche current increase is supported by a slight increase of the electric field
magnitude at the n+ –i and p+ –i junctions (Fig. 2.6, plot for current 10–4 A/μm).
With current increase due to exponential dependence of the multiplication coeffi-
cients upon the electric field, the total voltage drop on the multiplication regions
is less than the total voltage drop in the partially quasi-neutral middle of the i-
region (Fig. 2.6, plot for current 10–3 A/μm). As a result, these internal processes
provide for negative differential resistance of the entire device (Fig. 2.6, plot for
current 10–2 A/μm).
2.4 Avalanche–Injection in Si n+ –n–n+ Diode Structure
The structure of the n+ –n–n+ diode is the best representation of the saturation resis-
tor structure that is often used in ESD circuits for two stage ESD protection networks
(Chapter 4).
The structure of the n+ –n–n+ diode (Fig. 2.7a) is helpful for modeling avalanche–
injection in an n+ –p–n–n+ bipolar transistor for active regime operation (the
positive n+ -contact is a collector equivalent). The physical processes in this Si
structure are very common for other materials too. For example, the n+ –i–n+
parasitic structure is formed by source, i-buffer, and drain regions and is respon-
sible for current instability in GaAs MESFETs [18], similar to how the n+ –p–n+
parasitic structure is responsible for the same effect in discrete Si NMOSFET
devices.
As for integrated electronic components, the structure represents the drift
region in the high-voltage devices, drain extended MOS in CMOS pro-
cesses, N-channel lateral double-diffusion MOS (NLDMOS) devices fabricated
in conventional BCD processes or a lateral insulated gate bipolar transistor
(LIGBT), and many other high-voltage devices including avalanche diodes.
Thus, understanding the conductivity modulation in such devices is rather
important.
Similar to the previous sections, the n+ –n–n+ structure is first discussed within
an analytical approach. Then, the numerical simulation example is provided to
demonstrate operation in high injection conditions for a case close to typical analog
BiCMOS and BCD processes.
X
n+ n n+
a)
+ U –
I
I SCL
CURRENT
b)
I0
0 VOLTAGE
2
CURRENT
1
c)
I0
0 EBRL/2 EBRL
0 VOLTAGE
Fig. 2.7 Structure of the n+ –n–n+ diode (a) and typical I–V characteristic (b) for the boundary
cases of SCN type “1” and SCL type “2” (c)
2.4.1 Analytical Approach

For an n+ –n–n+ diode at a fixed current I, Ohm’s law is true only at rather small
current levels. In this case, the resistance of n-region R0 is equal to
L
R0 = , (1.13)
qμn ND A
√
where ND is the donor concentration and A is the total diode area A L . With

further current increase, the electric field E = U L increases up to the level E >

103 V cm, where the drift velocity is saturated at the level v ≈ 107 cm/s.
The density of free electrons in the n-region is equal to n = j v. At the dielectric
relaxation time in the n-region that is less than electron drift time, n = ND is true
and the current density is saturated at level j0 = qvND .
In theopposite case when n > ND and the total space charge is negative ρ =
qND − j v, the I–V dependence of the n+ –n–n+ diode at j > j0 is not saturated and

has a slope equal to dI dU = 2εvA L2 .
A typical experimental I–V characteristic of the n+ –n–n+ diode is presented in
Fig. 2.7b. For adequately high U, the diode current I is equal to the sum I = j0 A +
2.4 Avalanche–Injection in Si n+ –n–n+ Diode Structure 35
ISCL , where ISCL is the space charge limited current component (SCL). The ratio
between I0 and ISCL depends upon U, ND , and L.
For theoretical analysis, an understanding of two boundary cases is rather helpful:
space charge neutralized (SCN) diode and SCL diode (Fig. 2.7).
In the SCN diode ρ = 0, the electric field is distributed uniformly: E = U L ≈
const, I–V characteristic has a saturation region at I = I0 .
In the SCL diode ρ < 0, the electric
field is increased linearly in the anode contact
direction: EMAX = E (0) = 2U L. Near the cathode contact, E (L) ≈ 0 and the
electron injection in the n-region is rather limited.
Since both the doping concentration and the length of the diode play a major role,
a relative boundary between the SCN and SCL diodes for Si devices is ND L = 6 ×
1011 cm–2 [35]. At ND L 6 × 1011 cm–2 the diode is SCL; at ND L 6 × 1011
cm–2 the diode is SCN. However, this statement is true with the assumption that the
diffusion length of major carriers is bigger than the device length L, LD ∼ L.
At a small diffusion length or a long n-region, for example, due to a high recom-
bination rate, the SCL case is dominant. This situation is also realized in GaAs
n+ –i–n+ diodes.
Using the avalanche breakdown criterion (1.6), the avalanche multiplication in
the SCN diode begins at U ≈ EBR L. Electrons are accumulated near the anode
electrode; the field is increased in this region and shifted in the multiplication
region near the n+ –n boundary. The holes drift to the cathode, thus being injected
into the n-region, namely this phenomenon resembles the notion of “avalanche–
injection.”
This uncompensated hole space charge causes injection of electrons from the n–
n+ junction. At I >> I0 , a multiplication region is formed at the anode region and the
maximum field slightly exceeds EBR due to a sharp dependence g(E). The negative
space charge ρ = ND − n + p in the multiplication region provides electric field
decrease to E(L) ≈ 0. This SCL diode regime is realized at U = (EBR L) 2.
With the current increase above I > I0 , the SCN diode voltage is decreased
twofold (Fig. 2.7c) and an S-shaped I–V characteristic is observed in the critical
regime I = qND vA; U ≈ EBRL. Obviously, avalanche multiplication for the SCL-
diode begins once U = EBR L 2 and a negative differential conductivity region is
not presented in the I–V characteristic (Fig. 2.7c).
In the n+ –n–n+ structure in the avalanche–injection condition M = j jn [9];
therefore, in high injection j → ∞ , jn → jp , and M → 2. This fact namely pro-
vides the difference between the avalanche breakdown process in reverse-biased p–n
junctions, where M → ∞, and the avalanche–injection conductivity modulation that
occur at rather low M → 2.
From the experimental study [35], for samples with various ND and L, it follows
that the I–V characteristic of real n+ –n–n+ diodes has an intermediate shape between
the boundary cases of the SCN and SCL diodes.
It should be emphasized again that the above conclusions are true for cases
where diffusion length LD > L. For a diode structure with LD < L, the S-shaped
I–V characteristic is realized independently of the type and doping level. For exam-
ple, in n+ –i–n+ structures, the deep S-shaped I–V characteristic is observed both in
experiments and in numerical simulation studies [9].
2.4.2 Simulation Analysis

A significant expansion of understanding can be achieved at the simulation level.
A summary of the isothermal quasi-static numerical simulation analysis for the
n+ –n–n+ structure case is presented using the corresponding diode structure
(Fig. 2.8). This structure is geometrically identical to that described in Section 2.2.
Fig. 2.8 Isothermal I–V characteristics of n+ –n–n+ diode structure calculated for different n-region
doping levels
The device has a similar L = 3 μm space between the anode (positively biased)
and cathode (negatively biased) contacts with a W = 1 μm width. Current density is
normalized for the 1 μm structure depth. Three uniform doping profiles are defined
in the silicon region of the structure: the cathode n+ region with acceptor concentra-
tion NDn+ =1020 cm–3 defined from y = 0 up to position y = 0.5 μm and the anode
n+ -region with donor concentration NDn+ =1020 cm–3 from y = 1.5 μm to the end of
the device. Uniform doping is defined across the whole structure n– , with NDn as a
variable parameter that defines the doping level in the n-drift (or the n-base) region
of the structure with fixed length LP = 1 μm (Fig. 2.8).
The simulated quasi-static isothermal I–V characteristics of the device with dif-
ferent uniform doping levels are presented in Fig. 2.8, practically demonstrating that
the SCN structure is formed for the given structure parameters.
According to the analytical solution in the previous section, the device indeed
provides a linear resistance, followed by the formation of the saturation region as a
function of the n-region doping level.
After reaching the high injection current level, the conductivity modulation in
this structure is observed. This phenomenon is similar to the corresponding phys-
ical effect in n–p–n diodes discussed in the following section. It is related to the
positive feedback between avalanche multiplication at the anode junction and injec-
tion from the cathode junction. This will be demonstrated by comparative analysis
of the electric field and carrier density distributions at different current densities.
2.5 Conductivity Modulation Instability in n–p–n Diode Structures 37
2.5 Conductivity Modulation Instability in n–p–n Diode

Structures
Isothermal current instability (or the current mode secondary breakdown) was his-
torically first observed in power epitaxial n+ –p–n–n+ structures. The first studies
demonstrated that the thickness L and the doping level ND in the n-region pro-
vide a considerable impact on the condition of isothermal instability. Particularly,
it was determined in [34] that in structures with a thick epitaxial layer and in tran-
sistor structures without the n-layer, S-shaped I–V characteristics are not formed in
a rather wide current range. At the same time, in other BJT structures, isothermal
current instability has been observed at a relatively small collector current, followed
by catastrophic failure.
Different types of breakdown S-shaped I–V characteristics were measured using
a high-speed experimental technique. The particular type of technique used depends
on the structure, operation regime, and transistor type of the device.
From ESD design point of view, n–p–n devices are usually considered both as
parasitic structures formed in the integrated design and as components that pro-
vide the conductivity modulation effect for the ESD device itself. However, at a
submicron dimension, the diode n–p–n structure provides a rather high-temperature-
dependent leakage. Therefore, instead of a diode with a floating base, a triode n–p–n
structure is used, with the p-base held at the same initial potential as the n-emitter.
The triode structure or simply the bipolar junction transistor structure is discussed
in Section 2.6.
2.5.1 Conductivity Modulation in a Floating Base Region: Diode

Operation Mode
The n–p–n structure essentially represents the case of a bipolar junction transistor
in common emitter operation with an open base circuit.
For simplification of the following discussion, the 1D n+ –p–n–n+ diode structure
will be described under avalanche breakdown conditions.
2.5.1.1 The Case of Floating Base Breakdown (BVCEO) IB = 0

The operation regime of transistors with a floating base is roughly similar to the
n+ –n–n+ diode described in the previous chapter. However, this transistor operation
has several significant peculiarities.
Two versions of the transistor structure with different doping levels in the n-
collector region present a particular interest: A structure with ND L ≈ 6 × 1011
cm–2 (L = 10 μm) and C structure with ND L = 6 × 1012 cm–2 >> ND L = 6 × 1011
cm–2 (L = 9 μm).
The C-type n+ –n–n+ diode corresponds to the SCN type. The A-type diode is the
intermediate case between the SCL type and the SCN type.
© 1970 IEEE
I CR ICR
CURRENT
FIELD
diode a) b)
I > Io I=0
Io
EBR
Io X
+
UCR U [p] [n] [n ]
I1
I=0
CURRENT
FIELD
diode d)
Io
c)
I1
X
EBR L /2 ISCL
VOLTAGE
Fig. 2.9 Calculated I–V characteristics and electric field distributions for the structures with ND L
= 6 × 1011 cm–2 (a, b) and ND L = 6 × 1012 cm–2 (c, d) at different current levels [35]
The shape of the I–V characteristics of the transistor structures and the electric
field distributions in the n-region are presented in Fig. 2.9, in comparison with the
I–V characteristics of the corresponding n+ –n–n+ diode structures.
In operation regime with base current IB = 0, the collector current is equal to
the emitter current and proportional to (1 − αM)−1 . The increase in current begins
under conditions UCE → Uα , (αM → 1), i.e., when the electric field is approaching
a certain maximum value EMAX ≈ EBR on the p–n junction (Fig. 2.9b). Since the
value EMAX is limited by the condition αM = 1, below EBR the initial current
increase is observed at an increase of the collector–emitter voltage starting at level
L
UCE ≈ 0 E dx.
In the “A” structure, at j = j0 = qND v, the electric field levels off in the n-
collector region as a result of the electron space charge. Due to a lower n-collector
doping level at j > j0 , the avalanche multiplication region is formed on the n–
n+ junction, followed by a voltage decrease. The I–V characteristics of the A-type
transistor structure have a negative slope and aspire to the I–V characteristics of the
corresponding n+ –n–n+ diode structure. At significant current density I > ICR , both
the transistor and the diode have similar characteristics.
For the more heavily doped structure “C,” the I0 value is so high that breakdown
on the p–n junction is initiated when the space charge region (SCR) is confined
within the n-collector region (Fig. 2.9d). In this case, the I–V characteristic has a
positive slope that increases at U → EBR L 2, when SCR attains the n–n+ junction.
Therefore, for the C-type structure with a higher n-collector doping level, the dif-
ferential resistance remains positive dI dU > 0 up to the current level I0 , which
2.5 Conductivity Modulation Instability in n–p–n Diode Structures 39
exceeds the corresponding level in the A-type structure approximately by one order
of magnitude.
2.5.1.2 Numerical Solution for IB = 0 Case

A summary of the simulation analysis is presented in Fig. 2.10. It reveals simi-
lar regularities to the analytical approach above and provides a better accuracy for
the 1 μm p-base region. The quasi-neutral region formation at the emitter junction
with a corresponding reduction of the electric field can be observed through com-
parison of the electric field and carrier density profiles at different current levels
(Fig. 2.10c). However, a more accurate model demonstrates that the SCL regime is
0.0 0.5 1.0 1.5 2.0

ELECTRIC FIELD (kV/cm)
400
VHB
300
VHA
200 VTA
100 VBR
0
10 20
VHB
ELECTRONS (cm–3)
10 19
10 18 VHA
10 17 VTA
10 16
10 15
10 14
10 13
10 12
a) 10 11
10 10
10 –2 VHB VHA 10 9 VBR
Base Region Doping NPA(cm–3)
10 –3 10 8
10 –4 10
14 10 20
VTA 10 19
10 –5 10 18 VHB
CURRENT (A)
15
10
HOLES (cm– 3)
10 –6
10 17 VHA
10 –7 10 16
10 –8 16
10 15 VTA
10 10 14
10 –9 10 13
10 –10 1018 10 12 VBR
10 –11 10 11
17 10 10
10 –12 VBR 10 10 98
10 –13 10
10 –14 10 7
0 2 4 6 8 0.0 0.5 1.0 1.5 2.0
CATHODE VOLTAGE (V) b) Y (um) c)
Fig. 2.10 n–p–n diode structure (a), simulated I–V characteristics for different base doping levels
(b), and distributions of the electric field magnitude, electron and hole density at representative
current levels (c) for the structure with p-base doping level 1016 cm–3
actually observed above the level NA L = 1017 × 10−4 = 1013 , thus bringing the
accuracy of the analytical estimation within one order of magnitude.
2.6 Conductivity Modulation in the Triode n–p–n Structure

The most important applications of the n–p–n structures in submicron semicon-
ductor processes involve the triode operation regime. To analyze such an operation
mode in numerical simulation, a third contact region should be provided, thus essen-
tially making the device structure equivalent to the bipolar junction transistor (BJT)
structure.
2.6.1 The Case of Grounded Base Breakdown Operation UEB = 0

(BVCES)
In case of shorted emitter and base terminals, the initial location of avalanche break-
down is observed at the collector p–n junction. In this case, avalanche breakdown
is not accompanied by injection from the closed emitter junction, due to the contact
potential difference. Therefore, the collector current is provided by multiplication
of the saturation current IC0 IC ∼ MIC0 . The breakdown voltage is UCEBR > Uα
and is practically equal to the collector–base breakdown voltage UCBBR .
The avalanche-generated holes flow out of the base in negative base current
regime. This hole current results in an additional voltage drop across the p-base,
which results in an increase of the base potential at the emitter junction on the cor-
responding value rB IC , where rB is the corresponding internal base resistance. The
rB can be extracted for the simulation or measurement data. Thus, when the voltage
level rB IC is equal to or above the potential of the emitter junction opening (approx-
imately 0.7 V at room temperature conditions), injection of the electrons begins
from the emitter, followed by a positive feedback mechanism. Thus, this voltage
drop controls both the critical current level and the critical voltage of instability.
In experimental conditions, the beginning of the instability practically coincides
with the opening of the emitter junction. Therefore, for estimation of the critical
regime, the following equation may be used [34]:
ICR (RB + rB ) = 0.7, (1.14)
where RB is the external resistance of the base circuit. With RB increase, the criti-
cal current decreases [34], while the breakdown voltage value UCEBR is practically
unchanged. This results in the reduction of the probability of alternative thermal
instability mechanisms.
Condition (1.14) covers the case of the base grounded with the external resistor
of value RB as well. In device terminology, this case corresponds to the so-called
BVCER breakdown voltage.
2.6 Conductivity Modulation in the Triode n–p–n Structure 41
2.6.2 The Floating Emitter Case IE =0

In the case of floating emitter circuit, the injection of electrons becomes impossible.
However, from experiment [34], it follows that in a number of power transistors with
open emitters, the instability is observed in the IE = 0 regime as well. However, in
this case, the critical current level is approximately 10 times higher. The reason
for this effect is the increase of the emitter potential with an increase of the residual
current in the base up to some critical value. This critical value corresponds to a level
of surface breakdown voltage in the region between the emitter and the grounded
base.
Due to this specific breakdown scenario, the electrical contact between the
emitter and the base is restored and provides a possibility of injection from the
emitter junction. In this case, the critical voltage remains the same, but the crit-
ical current in reality corresponds to an increase of the emitter potential up to
the value UEBBR (which, for the case discussed in [34], was approximately 6 V).
Therefore, the criticalcurrent for positive feedback and corresponding instability is
ICR = (0.6 + UEBBR ) (RB + rB ). Apparently, this condition can be satisfied by the
implementation of proper device architecture.
2.6.3 Avalanche–Injection in a Common Emitter Circuit: The Case

of IB < 0 Regime
From numerous experimental studies of isothermal instability, it follows that for-
mation of negative differential conductivity in the pulsed regime results in current
filamentation in power transistors and ESD devices. The pulse duration tP is
adequate for the filamentation process (for example, tP ≥ 1 μs). Therefore, for
estimation
of the critical conditions, it is sufficient to determine the regimes with
dI dU ≤ 0.
In the case of active device structures, these conditions can be realized at
overloads in the collector circuit in the pinch-off condition. A similar effect can
be observed in a large signal operation at load mismatching. However, the most
important instances for this book are ESD events.
The application reliability aspect at IB <0 regime is critical for power transistors.
In this regime, the critical drain current level for instability sharply decreases down
to less than 1 mA.
The main feature of the negative base current operation is described below. Hole
current flow from the base results in the instability of emitter current: the higher
the jB , the higher the value of base potential near the emitter junction. Therefore,
the injection level of electrons from the emitter is respectively higher as well.
Flowing along the emitter junction (Fig. 2.11) in the base contact direction, the
IB (y) increases emitter current density jE (y) in the center of emitter strip and dumps
it in the edge of the strip. This effect mirrors the Fletcher effect. At fixed emitter
current IE (in common base configuration), the negative base currents might result
Ly
X
Lx
n+ p
Y IB IB(y)
IE n
Z
IC Wb n+
Fig. 2.11 Simplified structure of the power bipolar junction transistor with implanted emitter in
case of negative base current operation in common emitter circuit
in filamentation of the emitter current in the “OX” direction to the base current flow
IB in the “OY” direction.
Under negative base current conditions in the common emitter circuit, the
following expression can be used for the base current IB :
IC0 (1 − αM)
IB = − + IC . (1.15)
α αM
For sufficiently high potential on the emitter junction, the current is determined
by avalanche multiplication in the collector region: M >> 1 and IB ≈ –IC . This is
similar to the diode circuit for the case IE = 0 discussed above.
For the pinch-off regime, the injection efficiency of the emitter junction is very
low, α << 1. Therefore, for essential current generation, the conditions M >> 1 and
UCE > Uα are necessary. The appearance of holes in the base enhances injection
and results in emitter current increase. This results in immediate increase of gain α
to its maximum value. In this case, high multiplication coefficient values M are no
longer necessary to support the already increased current level. At the given current
source in the collector circuit, the multiplication coefficient M is decreasing due to
UCE reduction. This corresponds to an S-shaped region formation.
The negative differential resistance region was well studied to address reliabil-
ity problems in NPN BJT. It can be realized at rather small current levels under
dα dIE > 0. At the same time, in structures with high values of the saturation
current ISE and ISC , the S-shape may not form at all.
Obviously, from the reliability point of view, this S-shape is not “dangerous,”
because current redistribution is realized only at rather small current levels and
automatically terminated by saturation of the gain α. While filamentation at small
current levels does not result in failure, it should be considered that any NDC region
on the I–V characteristics might be accompanied by spontaneous switching of the
transistor into a new state, according to the load characteristic. This may be critical
for operation, as in a typical example of the latch-up scenario.
The Ic increase at UCE > Uα is determined by ratio (1.15). According to the
avalanche–injection concept, the critical current level occurs when jc increases up
to the j0 value (at least locally). At negative base currents (αM ≥ 1), emitter current
redistribution is possible. Therefore, the critical level of current density may be real-
ized at lower Ic levels, in comparison with IB = 0. The typical instability boundary
for transistors with this effect is presented in Fig. 2.12a.
COLLECTOR CURRENT (mA)

LND = 6 × 10–12cm–2
COLLECTOR CURRENT
Ib > 0
Ib > 0 100
a) b)
R=∞ 50 R=∞
R=0
R=0
0
U UCB BR 40 80 120 160
COLLECTOR-EMITTER VOLTAGE COLLECTOR-EMITTER VOLTAGE (V)
© 1970 IEEE © 1970 IEEE
Fig. 2.12 Stability boundary in the negative base current regime (IB < 0) for transistors with
collector region parameters ND L << 6 × 1011 cm–2 (a) and ND L >> 6 × 1011 cm–2 (b) [35]
To demonstrate transistor behavior at negative base current IB < 0, it is useful

to analyze the common emitter circuit with resistor RBE between the base and the
emitter at a given collector current IC . This circuit can be used both to measure the
pulsed stability boundary UCECR (IC ) at different values of R (from RBE = 0 to ∞)
and to simulate various base current levels.
A high, but finite, resistance results in a decrease of the gain coefficient, in com-
parison with the open base circuit case RBE = ∞ (IB = 0). Some portion of holes
flows out of the base. This corresponds to the condition αM ≥ 1 and UCE ≥ Uα
(1.8).
A small negative base current IB < 0 results in the appearance of a double effect.
On the one hand, negative base current IB < 0 reduces the emitter injection in com-
parison with IB = 0 and, on the other hand, it enhances the local current density in
the central part of emitter. In the region UCE > Uα (Fig. 2.12a), the redistribution
jC is dominant: while the peak current density jMAX is increasing, the total critical
current is decreasing. With RBE decrease, the total injection level is reduced. For
the conditions of emitter junction opening, the current increase of IC (IC ≈ –IB ) is
necessary:
ICCR (RBE + rB ) ≈ 0.7. (1.16)

For a transistor with ND L 6 × 1011 cm–2 , the falling region of IC at UCE > Uα
may not even be presented [35]. Indeed, in such structures, the main part of the
collector current IC is provided by avalanche generation. In this condition, the
injection efficiency of the emitter is rather low, so the emitter impact on field
redistribution in the n-collector region is negligible. Apparently, in this case, the
negative base current flow does not cause any significant redistribution of emitter
current. Therefore, the stability boundary has a shape similar to the one presented
in Fig. 2.12b. Practically, at RBE = 0, the I–V characteristic is equivalent to one of
the base–collector diode structure.
Evaluation of the critical regime for the avalanche–injection isothermal insta-
bility at IB < 0 condition is not as simple as the positive base current case IB >
0. However, analytical estimation for this effect is in rather good agreement with
experimental data.
2.6.3.1 Numerical Analysis for IB < 0 Case

In the case of high-gain standard bipolar devices (Chapter 3), the base is much thin-
ner and the collector has a complex profile. Nevertheless, the structure (Fig. 2.13)
represents rather practical case of a parasitic device that could both be used for ESD
protection and be present in an NMOS device with a submicron gate length. In this
chapter, the operation regimes of this important structure are further analyzed at a
comparative level of simplification. Integrated NPN BJT devices in analog BiCMOS
process technologies with complex profiles and regions are analyzed in Chapter 3,
for the cases of both vertical and lateral current transport. This section focuses on
essentially the quasi-1D model of the NPN BJT device.
VIN
NPN
NPN
RBE
GND
a) b)
Fig. 2.13 Model of the NPN BJT structure (a) and circuit equivalent (b) used for numerical
analysis and the common emitter device operation
A model of the NPN BJT structure is presented in Fig. 2.13. The structure has
three electrodes: base, emitter, and collector. The heavily doped n+ -emitter and
n+ -collector regions with ND =1020 cm–3 and the p+ -region for base contact with
NA =1019 cm–3 form contact regions of the structure.
In spite of significant simplification, the internal p-base and n-collector regions
are important to preserve in the device structure. The base–collector junction is
formed by the n-region with NDn =1016 cm–3 and the p-region with NAp =1017 cm–3 ,
respectively.
Conductivity modulation in the BJT structure in avalanche–injection conditions
is a strong function of the collector and base profiles. However, the major physical
process in the device can be properly illustrated.
The collector–emitter I–V characteristics for different base resistances of the
NPN BJT structure in common emitter mode are presented in Fig. 2.14 for two
base–collector doping compositions:
NAp = 1017 cm−3 /NDn = 1016 cm−3 and NAp = 1018 cm−3 /NDn = 1017 cm−3 .
10–2 BASE CONTACT

10–2 10–3 RESISTANCE 10 1
COLLECTOR CURRENT (A)
3
10–3 1 10–4 (Ohm/um) 4 10

10 10
10–4
104 10–5 105
10–5 105 10–6
10–6
10–7 10–7 10 7
107
10–8 10–8 108
108 10–9
10–9 109
10–10 10–10
BASE CONTACT 10–11
10–11
RESISTANCE (Ohm/um)
10–12 10–12
10–13 10–13
10–14 10–14
0 5 10 15 20 25 30 0 5 10 15 20
COLLECTOR–EMITTER VOLTAGE (V) COLLECTOR–EMITTER VOLTAGE (V)
a) b)
Fig. 2.14 Simulated collector–emitter I–V characteristics for two base–collector doping compo-
sitions: NAp = 1017 cm–3 /NDn = 1016 cm–3 (a) and NAp = 1018 cm–3 /NDn = 1017 cm–3 (b)
As can be seen from the simulation data, there are two S-shaped regions
(Fig. 2.14) that can be identified in the output collector–emitter I–V characteristics.
After avalanche breakdown and small current saturation, the structure exhibits
negative differential resistance with two negative differential conductivity regions.
At the first critical triggering point of instability VTA (Fig. 2.14a), the device
reaches the condition with a minimum holding voltage VHA , followed by a new
current saturation region that brings the voltage drop to a new triggering point of
instability VTB .
For the first current instability region, the dependence upon base contact resis-
tance remains in good agreement with the analytical expression (1.16) for silicon
material at room temperature conditions ICR (RB + rB ) = 0.8 =const.
However, this correlation is mainly true only at rather high base resistance and
respectively relatively low critical current below 0.1 mA/μm (Fig. 2.14b).
In case of strong injection in the base–collector region, additional effects become
involved. These effects are current saturation and conductivity modulation of the
collector region itself.
To explain these phenomena, the I–V characteristics of the NPN structure with
NAp =1018 cm–3 /NDn =1017 cm–3 and base contact resistance RBE =104 /μm
(Fig. 2.15a) are compared to the dependence of base contact bias and negative base
current (Fig. 2.15b).
0.0 0.5 1.0 1.5 2.0

ELECTRONS (cm–3) ELECTRIC FIELD (kV/cm)
10–2 VHB 700

VBR VTA
600
VTB VHB
500 VHA
10–3 400
VHA 300
VTB
200
10–4 100
VTA
0
1020
1019
VBR 1018 VHB
10–5 1017
0 5 10 15 20 1016 VHA
VTB
1015
COLLECTOR-EMITTER VOLTAGE (V) 1014
a) 1013 VTA
1012
1011
1010
NEGATIVE BASE CURRENT (mA)
0.10 VHB 1.6

BASE-EMITTER VOLTAGE (V)
VTB 1098 VBR

VHA 1.4 10
0.08 10 20
VTA 1.2 10 19
10 18 VHB
HOLES (cm–3)
1.0 10 17
0.06 10 16 VHA
0.8 10 15
10 14 VTB
0.6 10 13
0.04 10 12 VTA
0.4 10 11
10 10 VBR
0.02 0.2 10 9
10 8
0.0 10 7
0 5 10 15 20 0.0 0.5 1.0 1.5 2.0
COLLECTOR-EMITTER VOLTAGE (V) Y (um)
b) c)
Fig. 2.15 Conductivity modulation of the base and collector regions in the NPN BJT structure
model. Dependence of the collector current (a), negative base current (b), and distribution of the
electric field magnitude and electron and hole density (c) in the selected conductivity modulation
regimes VBR , VTA , VHA , VTB , VHB
In comparison of these data in the range before the instability point VTA , the
negative base current is practically equal to the collector current. In the first current
instability point VTA , the base contact voltage is reaching ∼0.8 V, thus eliminating
the potential barrier from the n+ -emitter region injection into the p-base region, for
the case of room temperature silicon material.
In the VTA regime, the distribution of the electric field magnitude corresponds to
the peak at the collector–base junction with the space charge (Fig. 2.15c, plot for
the electric field distribution).
With the following current increase, the density of the injected carriers in the
n-collector region exceeds the donor concentration in this region (Fig. 2.15c, distri-
bution for electrons) and results in the current saturation effect in this region. Due to
a much higher current in the base–collector region, current conduction is supported
at a much lower multiplication coefficient under lower electric fields (Fig. 2.15c,
distribution of electric field in regime VTB ).
At further current increase, modulation of the n-collector region conductivity is
observed. The density of injected carriers in the n-collector exceeds the NDn level.
This effect results in electric field redistribution with the maximum of the depen-
dence formed at the n–n+ collector interface. With the higher current density, the
quasi-neutral region is expanded toward the collector n+ –n junction, thus provid-
ing a lower voltage drop on the structure, which is associated with the second-stage
negative differential resistance.
In these conditions of strong injection, the injected carriers density in the struc-
ture is higher than NAp =1018 cm–3 /NDn =1017 cm–3 and the device operation is
similar to the high-current operation of the n+ –p–n+ structure with a floating base,
where conditions on the base provide practically no effect (Fig. 2.15b).
2.6.4 Avalanche–Injection in the Common Emitter Circuit with

Positive Base Current IB > 0
The common emitter NPN-based ESD clamp circuit is perhaps the best known ESD
protection configuration. In the case of ESD clamp, the positive base current can
be provided by an additional reference clamp component. This component can be
represented by an avalanche diode with an appropriate breakdown voltage, an open
base NPN BJT with an appropriate BVCEO voltage, or another device with break-
down type I–V dependence. An alternative solution can be realized with a more
complex control circuit that will provide a positive base current according to a more
sophisticated algorithm that accounts for conditions on other electrodes and dif-
ferent time delays (Chapter 3). In appropriate design of the reference circuit, the
snapback clamp can eliminate the transient time-dependent turn-on effect (dV/dt
effect).
This mode of operation is discussed using both the analytical and numerical
simulation approaches for the structure (Fig. 2.13).
2.6.4.1 Analytical Description of the IB > 0 Case

For simplification, the base current in the common emitter biasing circuit is assumed
constant and positive: IB = const, IB > 0. Neglecting the saturation current con-
tribution, in comparison with the high base current, the collector current is equal to
αMIB
IC = . (1.17)
1 − αM
Since the multiplication coefficient is the function of the collector–base voltage

and the collector current M = M(UCE , IC ), (1.17) determines the device I–V char-
By differentiating the right part of (1.17) on UCE at α = const, T = const,
acteristic.
and ∂M ∂UCE > 0, the simple criterion of instability is provided by
IC ∂M
> 1. (1.18)
M (1 − αM) ∂IC
Unfortunately both criteria (1.18) are obtained involving both the current gain
α(IC ) and the multiplication coefficient M(T) dependencies. Thus, the criterion is
hard to use in practical cases since it requires the exact dependencies α(IC ) and
M(IC ,UCE ).
For some particular cases, it is possible to complete the calculation of the insta-
bility boundary. Meanwhile, the result from this calculation is no more informative
than the qualitative estimation presented below.
At low base currents when jC < j0 = qND v, the transistor behavior is simply
similar to the case of IB = 0 presented above. In the vicinity of Uα , the collector
current begins increasing according to (1.17) up to the critical value I0 , when the
negative differential conductivity is formed, followed by the current instability.
At a corresponding base current level, when the collector current jC > j0 , the
maximum electric field is formed on the subcollector n–n+ junction. Resulting con-
ditions are formed for avalanche–injection instability immediately with avalanche
generation, i.e., at the electric field EMAX = EBR . This effect is observed at
UCEBR < Uα , where Uα is the breakdown voltage in the CE bias configuration
derived from the condition α 1 M(Uα ) = 1.
It is quite clear that with the increase of IC , a consequent value of UCEBR will
be decreased. For IC > 2I0 , the thickness of the space charge region W will still be
lower than the thickness of the n-epilayer L. With the assumption of triangular elec-
tric field distribution, the integral
value of the collector–emitter
voltage
in the critical
point is UCECR ≈ EBR WCR 2, where WCR ≈ (EBR ε) j v − qND . Therefore, for
adequately high current IC I0 , the isothermal stability boundary must be close to
the hyperbolic function:
εEBR
2 v
UCR ICR ≈ . (1.19)
2
2.6.4.2 Numerical Analysis of the IB > 0 Case

In order to compare with the analytical approach, the same NPN BJT device struc-
tures with NAp = 1017 cm–3 /NDn =1016 cm–3 and NAp =1018 cm–3 /NDn =1017 cm–3
(Fig. 2.13a) are solved with current boundary conditions for the base contact. The
same silicon parameters, room temperature, and quasi-static solution have been
calculated.
At different positive base currents, it can be observed that the instability point
depends on the device design that provides different current gain. The structure
with parameters NAp =1017 cm–3 /NDn =1016 cm–3 in Fig. 2.16a provides current
gain of over 1, with dependence close to hyperbolic (1.19), while the device with
parameters NAp =1018 cm–3 /NDn =1017 cm–3 has gain lower than 1 and provides a
different type of collector–emitter characteristic (Fig. 2.16b).
10 –2
BASE CURRENT 10 –2
–2 BASE CURRENT
10 (A/um)
–2 (A/um)
–3 10
10
10 –3 10 –3 –3
10
–4 10
–5
10
–4
0 10
10 –4 10 –4
–5
10
0
10 –5 10 –5
0 5 10 15 0 5 10 15
COLLECTOR-EMITTER VOLTAGE (V) COLLECTOR-EMITTER VOLTAGE (V)
a) b)
Fig. 2.16 Simulated collector–emitter I–V characteristics for two base–collector doping com-
positions: NAp =1017 cm–3 /NDn =1016 cm–3 and NAp =1018 cm–3 /NDn =1017 cm–3 at positive base
current
This example demonstrates that numerical simulation analysis provides a more

accurate and simple insight into device characteristics than a simple analytical
estimation.
In the device with low current gain, completion of the condition αM > 1 requires
much higher multiplication coefficients, and thus a much weaker dependence upon
base current is observed.
While high-current gain is one of the major figures of merit for NPN devices
supported in integrated process technology, low current gain structures are not rare
cases. Most practical examples of such occurrences are provided by parasitic struc-
tures formed in NMOS devices, as well as by parasitic structures that might be
formed between different devices in the drawn layout, for example, between the
n+ -source and n+ -region of the antenna diode in Pwell of the NMOS device (see
Chapter 6).
Using numerical simulation, the physics of conductivity modulation in a structure
with positive base current can be understood by comparing the electric field and
carrier densities in the conditions VTA and VHA (Fig. 2.17a).
0.0 0.5 1.0 1.5 2.0

700
600 X = 0.3 μm
500
400
300 VTA
200
VHA
100
0
20
10
ELECTRONS (cm–3)
10 19
VHA
18
10
10 –2 VTA
VHA
10 17
IB = 10–3 V TA 10 16
20
10 –3 10
X = 0.3 μm
HOLES (cm–3)
10 19
VHA
10 –4
18
10
VTA
10 –5 10 17
0 5 10 15 0.0 0.5 1.0 1.5 2.0
COLLECTOR-EMITTER VOLTAGE (V) Y (um)
Fig. 2.17 Collector–emitter I–V characteristics of the NPN structure (Fig. 2.13a) with the base–
collector doping level NAp =1017 cm–3 /NDn =1016 cm–3 under positive base current IB =10–3 A
and distribution of the electric field magnitude, electron and hole density for cutline X = 0.3 μm
in the triggering and holding regimes VTA and VHA
Similar to n+ –p–n+ structures, discussed above, negative differential resistance

is formed due to the reduction of the electric field in the n-collector region between
Y = 1 and 1.5 μm (Figs. 2.13a and 2.17b). Thus, namely conductivity modulation
of the n-collector region is responsible for the formation of S-shaped characteristics
in this structure.
2.6.5 Avalanche–Injection in the Common Base Circuit

The common base circuit operation has relatively rare ESD applications. However,
for completeness of the presented material, this case is discussed in this section
based upon the results of early experimental studies.
The conditions of isothermal instability in the common base bias circuit (CB)
and common emitter bias circuit (CE) are slightly different from each other at
UCB < Uα . For a current density above the critical level qND v, the instability begins
immediately with avalanche–injection at UCB of avalanche ionization:

jC
UCB ≈ εEBR
2
2 − qND .
v
At jC > 2j0 , the device turns on into the negative base current regime. For CE cir-
cuits, this effect is accompanied by the loss in collector control by the base current.
For CB circuits, the voltage range Uα < UCB < UCBCR is still operational.
The negative base currents may result in a redistribution of the emitter current.
In CE circuits, this redistribution involves a part of injected electrons from the emit-
ter followed by a redistribution of the total collector current that has an avalanche
component. In CB circuits, at UCB < UCBCR the collector current is practically
equal to the emitter current (jC = αMjE ). Therefore, the degree of current redistri-
bution might be significantly higher. At IE = const, it can result in a scenario where
the isothermal filamentation of the emitter current is not translated to the external
circuit as negative differential resistance, but essentially takes place under positive
differential conductivity I (UCB ).
At UCB > Uα , the hole space charge is increasing the forward bias level of
the emitter junction. The non-uniformity of this process results in redistribution
of emitter current as follows: the higher the hole density, the higher both emitter
forward bias and the emitter current density. When flowing through the avalanche
multiplication region, the density of the electron current results in an increase of
the avalanche hole density. If in these conditions the current density jCMAX in the
middle of filament becomes equal to the critical value j0 , then the triggering-on is
observed according to the classical avalanche–injection mechanism.
From a theoretical point of view, the isothermal instability in CB might have a
two-stage evolution. The evolution of avalanche–injection in the collector n-layer
is not a necessary condition for transistor failure. Emitter current filamentation
at αM > 1 cannot cause the catastrophic consequences by itself, because the
total dissipated power PC = UCB IC is practically unchanged, namely, the current
redistribution results in a sharp increase in maximum specific power jCMAX UCB .
A distinct feature of the isothermal instability in CB at real operation is the low
level of excessive currents before voltage redistribution (at least one order of mag-
nitude lower than in the CE bias circuit). Critical excessive current ICR means that
there is a difference between the values of ICR and IC before the NDC formation
(Fig. 2.18a). In CE, the value of ICR may be achieved at ∼0.1 mA, but in CB
ICR in the Uα < UCB < UCBCR range it is higher by an order of magnitude.
600
CRITICAL CURRENT (mA)

400
I0
a) b)
200
ICR
0
U UCB CR 0 25 50 75 100 125 150
COLLECTOR-BASE VOLTAGE CRITICAL VOLTAGE (V)
Fig. 2.18 Isothermal instability in the common base circuit: I–V characteristic (a) and experimen-
tal stability boundary for two samples (b)
The experimental shape of the stability boundary (Fig. 2.18b) corresponds to a

concept of primary emitter current filamentation. The decrease of ICR at UCB > Uα
is apparently the same as the decrease of a similar value in CE circuits. It should
be considered that the value varies for various devices (of a given transistor type)
and is the result of device parameter scattering and presence of local structural
defects (Fig. 2.18b). This problem will be discussed in the following section for
filamentation criterion.
2.7 Avalanche–Injection in PNP Structures

ESD clamps based upon the p–n–p structure are key components for ESD protection
solutions for fast transient pins, hot-plug-in, and system-level applications.
In principle, the processes in p–n–p BJT structures can be described using the
same analytical approach that is presented above for the other NPN structures.
However, the major difference in the case of silicon-based process technologies is
implied by the difference in carrier mobilities and multiplication coefficients.
This difference in parameters for electrons and holes in the case of real device
parameters provides a much higher holding voltage and much smaller negative dif-
ferential resistance. In addition, the positive feedback due to avalanche–injection
processes has a much higher compensation by the negative feedback of lower
conductivity in the case of holes as majority carriers.
At the same time, a mild negative differential resistance is a rather useful device
feature that provides an ideal vertical I–V dependence of the clamping voltage upon
current by compensation of the higher resistance of the p-regions (Fig. 2.19).
Practical realization of both effects requires substantial accuracy. Therefore,
numerical simulation of such p–n–p structures presents a valuable advantage. To
demonstrate the difference between silicon n–p–n and p–n–p structures, a simula-
tion was completed for a similar PNP BJT structure and is presented in Fig. 2.20.
2.8 Double Injection in Si p–n–p–n Structures 53
BASE CURRENT (A/um)

10–2 –2
10

–3 –3
10 10
–4
10
10–4 –6
10
10–5
10–6
10–7 10
–9
10–8 –12
10
–9
10
0 5 10 15 20
a) b)
Fig. 2.19 Simulated collector–emitter I–V characteristics for the PNP structure with reversed dop-
ing type in comparison with the NPN structure (Fig. 2.13a) for different base current levels in the
common emitter circuit
The 0.5 μm long p-collector and 0.5 μm long n-base with a uniform doping pro-
file were defined similar to the corresponding levels in the NPN device (Fig. 2.16a)
NDn =1017 cm–3 and NAn =1016 cm–3 , respectively. The base current direction and
collector voltage have been, respectively, reversed.
According to the cutline plots, namely modulation of the n-base provides for the
S-shaped dependence. The distribution of the electric field and carrier density across
the cutlines at X = 0.3 μm is presented in Fig. 2.20.
Thus, this simple numerical experiment provides several practical guidelines for
PNP-based ESD clamp design (Chapter 4).
2.8 Double Injection in Si p–n–p–n Structures
2.8.1 Equivalent Circuit

This section deals with structures that are key components of power electronics and
parts of many power management devices: insulated gate bipolar transistors (IGBT),
silicon-controlled rectifiers (SCR), DIACs, TRIACs, and many others [14]. In addi-
tion, these elementary structures represent the parasitic devices in many different
latch-up scenarios.
The basic theory of thyristor can be found in [14]. This section is focused
mainly on p–n–p–n structures specifically related to ESD application and latch-up.
0.0 0.5 1.0 1.5 2.0 2.5 3.0

500
V BR V H X = 0.3 μm
400 VT
300
200
100
0
1018
10 17 VH
ELECTRONS (cm )
–3
1016
1015 VT
1014 V BR
1013
1012
1011
1010
10 –2
NEGATIVE COLL. CURRENT (A)
109
BASE CONTACT VH 108
10 –3 1 VT
RESISTANCE 107
–4 (Ohm/um) 3 106
10 10
4 1020
10 1019
10 –5 1018 VH X = 0.3 μm
HOLES (cm )
7 1017
–3
10 –6 10 1016
1015
1014 VT
10 –7 V BR 1013
1012
10 –8 1011
1010 VBR
10 –9 10
9 1098
107
106
10 –10 10
0 5 10 15 20 25 0.0 0.5 1.0 1.5 2.0 2.5 3.0
NEGATIVE COLL.-EMIT. VOLTAGE (V) Y (um)
a) b)
Fig. 2.20 Calculated collector–emitter I–V characteristic for the PNP structure with different
lumped base resistances (a) and comparison of the electric field and carrier density distribution
for the cutline at X = 0.3 μm in representative regimes (b)
As well, the region parameters are chosen typical of 0.5 μm process technolo-
gies. Conductivity modulation effects in such devices are demonstrated mainly by
numerical simulation.
The final stage of the physical mechanism of positive feedback realized in p–n–
p–n structures is called double-injection conductivity modulation. A distinct feature
of such a modulation mechanism is that it can be realized at rather low voltages of
∼1.5 V. In these low electric field conditions, avalanche breakdown in the structure
is suppressed and not required to support conduction of high current, unlike in the
case of conductivity modulation mechanisms discussed above.
Due to low holding voltage in high-current regime, the dissipated pulsed power is
low as well. This provides a significant advantage for ESD design that can be taken
from both low dissipated power in the active region and the low clamping voltage
itself. In a properly designed device, the double-injection conductivity modulation
provides internal current density over a 10 mA per micron width. This provides a
very important device-level solution for high-speed applications, due to such a low
parasitic capacitance required for a standard package-level spec.
The low (below 2 V) clamping voltage has a critical advantage over protection
by a leaky diode stack for nanoscaled CMOS processes.
However, the low clamping voltage has a downside as well. In case of high-
voltage and transient pins, where transient latch-up conditions can be created, it is
hard to realize SCR devices with holding voltage above the power supply level.
Several solutions for this will be discussed in the next chapter.
In some cases, the p–n–p–n structure can be analyzed as made of four layers with
a corresponding sequence of the p+ -emitter, n-base, p-base, and n+ -emitter.
Most of the physical effects in the device can be understood using a so-called
Ebers circuit [14] for device representation (Fig. 2.21). In this structure, the n–p–n
and p–n–p devices are sharing corresponding collector and base regions. In sub-
micron dimensions, connection of the bases is necessary to suppress gain of the
parasitic n–p–n and p–n–p in the structure (Fig. 2.21a).
ANODE
ANODE
rnB
pnp pnp
npn npn
rnB
CATHODE CATHODE
a) b)
Fig. 2.21 Ebers circuit used to represent the p–n–p–n structure with open external bases (a) and
shorted bases (b)
Therefore, the correct way to study conductivity modulation in the p–n–p–n

structure requires the use of a version where the n-base and p-base are connected.
The most typical circuit configuration for this structure in submicron ESD designs
is presented in Fig. 2.21b. This parasitic structure can be formed by CMOS inverter
circuit (PMOS and NMOS) in the following sequence: p+ -source and n-well of
PMOS and p-well and n+ -source of NMOS.
A two-terminal diode structure is realized in the case of floating base regions
(Fig. 2.21a). In general, the blocking junctions can be connected to a reference volt-
age and current subcircuit in order to realize a desired level of the triggering current
and voltage (Chapter 3).
One of the particular cases realized in submicron process technologies for ESD
protection is device implementation with a p–n–p–n structure where the base regions
are connected to the corresponding anode and cathode regions (Fig. 2.21b). The
most useful operation regime for ESD protection is one where the p+ -contact region
is positively biased. In this case, the blocking junction is created by the high-voltage-
tolerant n-base to p-base junction.
The phenomenological principle of device operation can be understood using the
theory of bipolar devices and the equivalent circuits (Fig. 2.21).
Criteria for positive feedback that results in conductivity modulation in p–n–p–n
structure can be obtained by
αNPN MN + αPNP Mp > 1, (1.20)
where α NPN , α PNP , MN , MP, are current gain and multiplication coefficients in the
internal n–p–n and p–n–p structures.
In the case of floating base regions (blocking junction contacts), the p–n–p–n
structure operation depends upon the formed SCN or SCL structures.
The collector current of each structure provides for the base current of the other
structure. Thus, even if one of the devices is achieving the condition of αM >1,
instability is unavoidable.
In submicron process, the base regions are usually rather small and comparable
with diffusion length. In this case, a more practical ESD device is realized using a
connection of the blocking junction regions (Fig. 2.21b).
2.8.2 Simulation of Conductivity Modulation in p–n–p–n

Structures
A corresponding numerical solution can be obtained using a 2D structure with the
base layers made of a sequence of the anode p+ -region, n-base, p-base, and cathode
n+ -contact region.
The conductivity modulation processes in p–n–p–n structures with floating and
connected base regions are demonstrated by means of numerical simulation in
Figs. 2.22 and 2.23, respectively.
2.8.2.1 Floating Base Case

On the phenomenological level, the principle of conductivity modulation in p–n–
p–n structure operation is rather easy to understand using simulation results for
different doping levels in the base regions.
In the four-layer device with floating blocking junction regions (Fig. 2.22a),
when the p+ -anode is positively biased, the current in the device is blocked by the
reverse-biased junction that is formed by the n–p bases.
With voltage increase, the critical voltage depends on the space charge region
length as compared to the base region length. At a few volts drop on the structure, in
the case of low base doping of <1016 cm–3 , the space charge region covers the whole
1 μm base region. This provides a scenario for current conduction mode under high-
current gain conditions with α NPN , α PNP >>1 under small MN , MP ∼1. Significant
10–2
10–3
Base Doping NA&ND (cm–3)

16
10–4
10 &10
10–5
16
CURRENT (A)
10–6
19
18
17
10 &10
10 &10
10 &10
10–7
19
18
17
10–8
10–9
10–10
10–11
10–12
0 1 2 3 4 5 6 7
CATHODE VOLTAGE (V)
a) b)
Fig. 2.22 Two-dimensional model for simulation of the conductivity modulation effect in p–n–
p–n structures (a) and simulated isothermal quasi-static I–V characteristics with different doping
levels in the base regions (b)
amplification of the saturation current at low voltages results in conductivity modu-

lation (1.20) at very small multiplication coefficients. With further current increase,
the high-conductivity region is formed (Fig. 2.22 for NAp =NDn =1017 cm–3 ).
However, at higher doping levels of the base regions, the critical voltage for con-
ductivity modulation rises. In this case, the space charge region length is less than
the base length, and thus the punch-through operation is not realized. This enables
an increase of the electric field at the blocking junction up to a level suitable for
avalanche breakdown. This provides high multiplication coefficients MN or MP >>
1 and corresponding multiplication of the saturation current and low current gains
with completion of condition (1.20).
At higher current, the saturation of the n and p regions is observed and results
in expansion of the space charge region up to the contact n+ and p+ regions. After
creation of the bias of ∼0.7 V, the potential barrier is eliminated and injection from
the corresponding p+ –n and n–p+ forward-biased junctions begins. This forms a
positive feedback that creates the double-injection conductivity modulation mode.
In high injection mode, the injected high-density electron–hole plasma over-
modulates the conductivity of both base regions. Due to positive feedback, the
structure can support high current conduction without high avalanche multiplication
MN , MP ∼1.
That is why this type of conductivity modulation is called double injection, in
opposite to avalanche–injection as in n–p–n structures, where the positive feedback
0.0 0.5 1.0 1.5 2.0 2.5 3.0
ELECTRONS (cm–3) ELECTRIC FIELD (kV/cm)

VT X = 0.65 μm
100
VBL
VBR
10 VHA
VHB
1
1020
1019 VHB
1018 VHA
1017
1016
1015 VT
1014
1013
1012
1011
a) 1010
1098
107 VBR
V HB 106
10 –2 105
104
103
VBL
10 –3 V HA 102
10 –4 10
VT
Base Doping NA &ND (cm )
1020
–3
10 –5 1019 VHB
CURRENT (A)
10 –6 1018
0 16
1017 VHA
10 6
17
1016
1 0 1 7& 1
10 –7
HOLES (cm–3)
& 10 17
&1
18
10 &10
1015
10 &10
10 16
10 –8 1014 VT
1013
17
10 16
18
10 –9 1012
1011
10 –10 1010
V BR 1098
10 –11 107 VBR
10 –12 106
105
10 –13 V BL 104 VBL
103
10 –14 102
10
0 5 10 15 20 25 30 35 40 0.0 0.5 1.0 1.5 2.0 2.5 3.0
ANODE VOLTAGE (V) b) Y (um)
c)
Fig. 2.23 Two-dimensional model for simulation of the conductivity modulation effect in p–n–p–n
structures with shorted blocking junctions (a); simulated isothermal quasi-static I–V characteristics
at different base doping levels (b) and electron and carrier density distributions for different current
with different doping levels in the base regions (c)
is realized due to low-field injection from the emitter n+ –p junction and avalanche
generation at the collector p–n+ junction. Respectively, this type of conductivity
modulation in reverse-biased p–i–n is called double-avalanche–injection due to the
positive feedback from two multiplication regions formed in the structure.
2.8.2.2 Connected Base Case

The simulation example for a two-terminal device with connected base regions is
presented in Fig. 2.23. To simplify the example, corresponding interdigitated n+
and p+ base contact regions are defined in the structure.
The upper and lower contact regions are connected to the corresponding base and
emitter regions, thus forming a diode structure.
2.9 Spatial Current Instability Phenomena in Semiconductor 59
The final two-terminal device is a good physical representation of a real ESD

device implemented in integrated process technologies. The scenario is discussed
above, with corresponding correction to the operation of the internal n–p–n and
p–n–p structures with negative and positive base currents, respectively.
This mode of operation provides corresponding high blocking voltages at low
base doping levels, by reducing the current in the multiplication region due to base
connection to the anode and cathode.
Since the bases are externally connected, the saturation current has an escape
path. In this case, even at low base doping, the structure is capable of achieving high
breakdown voltage. Only when avalanche breakdown followed by the avalanche–
injection in either n–p–n or p–n–p provides a sufficient current level, the double-
injection conductivity modulation becomes dominant. This conductivity modulation
provides positive feedback that terminates avalanche breakdown. In the final state,
the device is in on-state and conducts current similar to the forward-biased p–i–n
diode, since the injected carrier density significantly exceeds the doping level in the
base regions. In Chapter 3, various SCR devices based upon the p–n–p–n structure
with double-injection conductivity modulation will be discussed in greater detail.
2.9 Spatial Current Instability Phenomena in Semiconductor

Structures with Negative Differential Resistance
The formation of spatially non-uniform states in distributed systems can be observed

at certain conditions of different spatial parameters for positive and negative feed-
back processes. The parameters for these conditions are internal to the system. There
are many examples of such systems and phenomena in traffic, biological, chemical,
social, and semiconductors [26].
Usually, the current density instability processes in the active device can be
described physically using two parameters-activator and inhibitor. Spatially non-
uniform states can be formed such that the inhibitor and activator have different
characteristic spatial or temporal parameters [26].
In chemical systems, these internal processes can be represented by reaction and
diffusion processes [23–27]. In the case of semiconductor systems, current instabil-
ity with a small spatial parameter acts as an activator, while a diffusion or current
spreading in the quasi-neutral areas acts as an inhibitor.
As for distributed systems, the case of ESD devices practically always presents a
distributed object where the required total current is achieved by appropriate width
scaling.
Practically all elementary semiconductor structures, as described above, provide
at certain conditions an example of the system where interplay between positive
and negative feedback mechanisms provides for formation of spatially non-uniform
current conduction.
The activation process in the structure is the avalanche–injection process itself
that provides positive feedback on the local level. Usually, the spatial dimensions of
such a parameter are approximately equal to the space charge region of the structure
where the conductivity modulation is realized.
The inhibition effects are provided by the ballasting and contact regions in
the structure, due to current spreading to such regions. Thus, the typical spatial
dimension of the inhibitor is roughly equal to the saturation region’s length.
In the ESD specific case, the primary current filamentation effect, due to conduc-
tivity modulation of the discharge regions during the snapback, is controllable and
reversible. The secondary filamentation effect is usually responsible for damage of
the ESD devices in the pulsed mode.
This phenomenon is similar to the physical limitation of any device [9].
Practically, spatially uniform damage is never observed in distributed semiconductor
structures. On the contrary, the most typical scenario is a solitary or multiple-
structure melting region (Fig. 2.24). However, often the original damage which
occurred in pulsed regime might be significantly altered by the subsequent pulse
zaps or functional tests. Often, the following EOS event which occurs under
power-on conditions creates a more significant damage due to a domino effect.
a) b)
c) d)
Fig. 2.24 Examples of failure analysis photos for catastrophic burnout in the ESD snapback
NMOS device (a), ESD diode (b) NLDMOS as result of HBM (c) and MM (d) pulses
2.9.1 Current Filamentation at Avalanche–Injection
The first 2D numerical simulation of the avalanche–injection current filamentation

effect has been presented in [15] in order to explain the physical nature of the drain-
source burnout effect in MESFETs [16–19]. The phenomenon has been identified
in the parasitic n-i-n bipolar device.
Similarly, the avalanche–injection current filamentation effect has been simulated
in [20] for silicon n–p–n structures.
To model the base region connection with 2D numerical simulation capabilities,
a multicell mathematical approximation structure has been suggested (Fig. 2.25a).
Thus, an essentially 3D problem has been solved within 2D numerical simulation
capability by the composition of a 2D triode n+ –p–n+ [p+ ] structure. The structure
X(µm)
DRAIN
n+
2
p -Si a)
1
–n+ –p+ –n+ –p+ –
Z(µm) 24 20 16 12 8 4 0
+
SOURCE/p –CONTACT
DRAIN AND P+–CURRENT (A/cm2)
105
F IL A
Copyright 1997, with permission from Elsevier

M EN
104 T
VP + = 0
103
ORM
102 b)
UNIF
101
100
0 5 10 15 20
VOLTAGE (V)
20
delay time
filamentation time
UNIFORM
DRAIN VOLTAGE (V)
15
VP + = 0 c)
IP + = 0
10 UNIFORM
FILAMENT
5
0 1 2 3 4
TIME (ns)
Fig. 2.25 Multicellular n+ –p–n+ [p+ ] structure (a), calculated ID –VDS characteristics for the
grounded and floating p+ -cell configuration (b), and the drain voltage vs. time after a stepwise
current increase from ID = 0 up to ID = 20 kA/cm2 (c) [20]
presents itself as a periodic structure with multiple n+ - and p+ -cells of the source
and substrate contact regions.
In order to reduce the impact of the spatial structure of the device regions on
the spatially non-uniform solution, the dimensions of n+ - and p+ -cell contacts have
been chosen smaller than the expected filament dimension. An iterative method can
be used in the simulation to demonstrate that at certain dimension of the contact
regions, there is no observed influence on the final filament state. The structure can
be understood as a distributed cross section of low gain multifinger NPN BJT, where
emitter and base regions are reduced below reasonable process limits and butted,
while the p-base region is rather big.
A numerical solution has been obtained using transient isothermal simulation for
the conditions of applied constant current.
From general understanding of the non-linear theory [23–27], in order to initiate
the stratification of the solution into a spatially non-uniform state, small fluctua-
tions of the uniform distribution are needed. While in real devices the fluctuations
are always present due to statistical effects and non-uniformities, a physical equiv-
alent for fluctuation is required in the numerical simulation of the ideal structure to
enable the possibility of current stratification. According to Vashchenko et al. [18]
numerical error of solution provides an adequate source of fluctuations. As a result,
this spatial current instability effect can be simulated [18]. This numerical “noise,”
due to finite accuracy of the numerical solution, provides conditions for filament
excitation with no additional fluctuation.
As has been demonstrated [20] (Fig. 2.25b), the transient solution for the dis-
tributed structure (Fig. 2.25a) was spatially non-uniform. The initial current density
applied at the beginning of the transient solution was chosen at the level corre-
sponding to the negative differential resistance in a minimum contact width device.
After a fast transient process, the first unstable uniform state is formed during a
very short period of time (<1 ns) (Fig. 2.25c), and then the final quasi-static solution
corresponds to the current filament. The distribution of the electric field and carrier
density for the current filament state is presented in Fig. 2.26.
Within the filament region, the peak of the electric field peak at the drain (col-
lector) junction (Fig. 2.26a) corresponds to avalanche–injection conditions at high
current, with corresponding quasi-neutral electron–hole plasma in the p-region of
the structure (Fig. 2.26b, c).
The amplitude of the filament is roughly limited by the current saturation effect
in the contact n+ -regions. Therefore, further current increase results in broadening
of the filaments up to the structure width.
The important practical conclusion from this numerical experiment is the effect
of the contact regions in ESD devices. If the n+ -region is inadequate, excessive injec-
tion from the contacts and power dissipation in the metal–semiconductor interface
results in very fast local damage of the structure. This effect is significantly ampli-
fied due to a much lower melting temperature of the contact eutectic of ∼600◦ C in
comparison with melting effects in the semiconductor material of ∼1100◦ C.
The technique of the current limitation in conductivity modulation mode is prac-
tically realized in many ESD protection structures, including snapback NMOS
devices (Chapter 3).
400
V/cm)
300
FIELD (k
200
100 2.5
2.0
25 1.5
20 15 1.0 m)
10 5 0.5 (μ
0 X
Z (μm)
(a)
Copyright 1997, with permission from Elsevier

–3 )
Y (cm
18
10
DENSIT
17
10
16
10
15 2.5
ON
10 2.0
14
ELECTR
10 1.5 )
20 1.0 m
15 10 5 0.5 (μ
0 X
Z (μm)
(b)
m )
–3
18
10
NSITY (c
17
10
16
10
15 2.5
E
10
HOLE D
14
2.0
10 1.5 )
1.0 m
20 15 10 0.5 (μ
5 0 X
(c)
Fig. 2.26 Calculated distribution of the electric field (a), electron (b), and hole (c) density in the
2D n+ – p–n+ [p+ ] structure for current density I = 20 kA/cm2 [20]
Similar results can be reproduced using a commercial simulator, for the dis-
tributed device (Fig. 2.10a) with the following modifications. Two 0.25 μm wide
oxide regions have been added to the structure with the silicon width extended to
10 μm (Fig. 2.27a). The current filament state with two filaments is formed at 7 ns
(Fig. 2.27b).
2.9.2 Current Filamentation Effect in Double-Avalanche–Injection

Conductivity Modulation
The first numerical solution for current filaments due to double-avalanche–injection

conductivity modulation has been obtained in [37] to solve the problem of elec-
trical Schottky gate burnout in GaAs MESFET structures. For the current filament
solution, a model of the M–i–n+ structure with a Schottky contact has been used to
12
15
ANODE VOLTAGE (V)

10
8
10 6
4
5 2
0
0
1e-11 1e-10 1e-9 1e-8 1e-7 1e-6
TIME (ns) a)
b)
Fig. 2.27 Transient characteristics during the current filamentation effect in the n–p–n structure
with a floating p-region at current density 10 mA/μm (a) and current density depth profile for the
formed state with two filaments (b)
obtain the solution for a physical equivalent of the 1D case (quasi-1D case) with a
width of w = 0.2 μm and the case of filamentation with width W = 10 μm. The
i-region length of the M–i–n+ structure is Li = 1 μm, the n+ -region length is Ln =
0.3 μm, and the donor concentration in the n+ -region is 1018 cm–3 .
Conductivity modulation at Schottky diode breakdown is similar to the phe-
nomenon in the p–i–n structure [9].
Two-dimensional distribution of field and carrier density for the filament state is
presented in Fig. 2.28. For current filament solutions, typically the electric field and
carrier density distribution inside the filament region is close to that of the narrow
(1D) structure in high-conductivity state. At the same time, distributions further
away from the filament correspond to the state of 1D structure before avalanche
multiplication.
SCHOTTKY CONTACT LI Ln
X (μm) 600
0
V/cm)
500
DRAIN
FIELD (k
i-GaAs n+ W 400
1.2
300
Z (μm)
0.8
0.4
)
8
m
6 4
(μ
2 0.0
Z(μm) 0
X
(a) (b)
–3
)
ITY (cm
)
–3
1018 10
18
(cm
1017 10
17
1016 16
ENSITY
ON DENS
10
1015 15
10 14
1014 10 13
1013 1.2 10 12 1.2
HOLE D
1012
ELECTR
0.8 10 11 0.8
1011 10 )
0.4 0.4 μm
8 8 X(
)
6 6
m
4 2 0.0 4 2
μ
0 0
X(
Z(μm)
(c) (d) © 1996 IEEE
Fig. 2.28 Current filament in10 μm-wide M–i–n+

structure (a) and electric field (b), electron (c)
and hole (d) density distribution at 1×105 A/cm2 current density [19]
In case of avalanche–injection, the maximum current density inside the filament

is not simply limited below the current saturation in the n+ -region, but exceeds it.
This is the result of the current spreading to the n+ -region, which essentially sup-
plies more total current to the filament region than the corresponding saturation
current density in the n+ -region. This explains why the very narrow high ampli-
tude filaments, realized in Schottky diode breakdown, are very hard to observe
experimentally.
A summary of the simulation results for the silicon p–i–n structure with doping
profiles (Fig. 2.5a), expanded total width, and SiO2 regions at the device boundary
is presented in Fig. 2.29 for a current density of 10 mA/μm.
12
ANODE VOLTAGE (V)
10
CURRENT (mA/um)
60
8
40 6
4
20 2
0
0
1e-11 1e-10 1e-9 1e-8 1e-7 1e-6
TIME (ns)
a) b)
Fig. 2.29 Transient characteristics during the current filamentation effect in the p–i–n structure
with a floating p-region at current density of 10 mA/μm (a) and obtained final state with two
filaments (b)
As can be seen, filamentation occurs within a very fast transient process of tF <
1 ns, providing a rather narrow filament with a width of ∼1 μm (Fig. 2.29b).
2.9.3 Current Filamentation Effect in the Case of Double Injection

A similar solution of current filament can be obtained and studied in detail for a
p–n–p–n structure with shorted base regions, originally presented in Fig. 2.23a.
To obtain the distributed structure, similar SiO2 regions have been added and
the silicon region width of the structure has been increased up to 10 μm, iterating
interdigitated n+ and p+ regions for anode, cathode, and blocking junction connec-
tions (Fig. 2.30a). The quasi-stationary solution corresponds to the current filament
at the right side of the device (Fig. 2.30a), formed during a two-stage process during
tF ∼0.5 ns [19].
12
ANODE VOLTAGE (V)
CURRENT (mA/um)
60 10
8
40 6
4
20 2
0
0
1e-11 1e-10 1e-9 1e-8 1e-7 1e-6
TIME (ns)
a) b)
Fig. 2.30 Transient characteristics during the current filamentation effect in the p–n–p–n structure
with a floating p-region at a current density of 10 mA/μm (a) and obtained final state with two
filaments (b)
2.10 Summary
In this chapter, major characteristics of the conductivity modulation process have
been presented. Timescale of the ESD events allows us to make certain practically
important assumptions. Most of the processes in ESD devices can be described in
assumption of the adiabatic conditions. Due to the short time of the ESD event,
the generated heat is localized within the active area of the devices and typically
propagates in the structure on the distance of ∼1 μm only. In this case, a rather
accurate numerical solution can be obtained using thermal-coupled mixed-mode
simulation of ESD device cross sections with relatively small area not exceeding
that required for accurate simulation without heating effects.
2.10 Summary 67
Due to the same ESD timescale, only electrical mechanisms rather than thermal
determine the initial conductivity modulation phenomena and resulting ESD pulse
operation transient waveforms.
Thermal and electro-thermal mechanisms [9] become critical only in regimes
too close to irreversible failure, thus providing a physical limitation on the max-
imum pulsed power for ESD devices. In most cases, these phenomena provide
constraints for the ESD design that mainly focuses on the provision of appro-
priate waveforms, dc voltage tolerance, and parasitics reduction. Typically, when
the combination of the design targets for ESD device is achieved, the physi-
cal limitation of the ESD device operation regime itself occurs beyond the ESD
protection window. Therefore, isothermal conductivity modulation mechanisms
are directly applicable to the ESD device design, discussed in the following
chapter.
In summary, in isothermal pulsed conditions, only four major basic conductivity
modulation phenomena are realized in semiconductor structures in addition to the
injection in forward-biased diodes:
(i) avalanche breakdown;

(ii) double-avalanche–injection;
(iii) avalanche–injection;
(iv) double injection.
Practically any complex ESD device, as well as standard devices provided by

the manufacturing processes technology, or even parasitic devices formed inside
integrated circuit, can only provide a condition when either one or a combination of
these basic mechanisms is activated.
Often, a more complex domino effect or alternative current path is realized.
For example, snapback operation of an SCR device for ESD protection can be
initiated first by avalanche breakdown of the blocking junction, then by avalanche–
injection in the parasitic NPN structure, and only then double-injection conductivity
modulation will take place.
Another example is avalanche–injection in vertical NPN, realized in the
BiCMOS process when, depending on the current level, the lateral current con-
duction can take over the vertical current path. Double-avalanche–injection can be
initiated due to n-base conductivity modulation under avalanche breakdown of the
p+ –n–n+ structure.
Nevertheless, in spite of a variety of ESD protection devices and clamps, the
physical effects taking place during ESD events are similar to the phenomena in ele-
mentary p–n, n–p–n, p–i–n, p–n–p, and p–n–p–n structures discussed in this chapter.
Chapter 3 will demonstrate how understanding of these phenomena can be used
to study more sophisticated ESD devices and provide physical limitations of elec-
trical SOA in standard devices. Then, in Chapter 4, clamp-level ESD protection
will be discussed using both the equivalent circuits of ESD clamps and the layout
topology.
DECIMMTM Simulation Examples for Chapter 2

Example 2.1 Numerical analysis of the conductivity modulation in the diode p+ –p–
n+ structure.
Example 2.2 Numerical analysis of the conductivity modulation in the p–i–n
structure.
Example 2.3 Numerical analysis of the conductivity modulation in the n+ –n–n+
diode structure.
Example 2.4 Numerical analysis of the conductivity modulation in the quasi-1D
n–p–n structure.
Example 2.5 Numerical analysis of the conductivity modulation in the 2D NPN
BJT structure.
Chapter 3
Standard and ESD Devices in Integrated Process
Technologies
In Chapter 2, conductivity modulation mechanisms were discussed based upon

examples of these mechanisms in corresponding elementary semiconductor struc-
tures. The purpose of Chapter 3 is to describe how these conductivity modulation
phenomena are realized in both standard devices supported by electrical design
rules of given integrated process technology and free ESD devices developed in
the process.
The pulsed safe-operating area (SOA) and physical limitations of the operation
regime are discussed first for standard devices in typical integrated process tech-
nologies. Then, “free” ESD devices are described in greater detail specific to the
device type.
The definition of a “free ESD device” assumes no additional process steps in
device formation. A free ESD device can be built using only the available mask
layers within physical limits of the process. The device can be formed using a
self-aligned approach, where variation of the masks will not change the charac-
teristics, or as a non-self-aligned device. In the last case, the device is sensitive to
misalignment of the mask layers, which requires the use of additional measures to
guarantee repeatable device characteristics (or acceptable parametric yields) within
the specified parameter range.
ESD device engineering requires a strong understanding of the integrated
process technologies, architecture and process step specifics. Therefore, to sup-
port the device-level discussion, Sections 3.1 and 3.2 establish basic definitions.
These definitions cover the most typical for integrated analog circuits manufac-
turing process technology steps and supported devices regions. This material is
required for understanding of the subsequent sections. A deeper understanding of
the process technology can be found in many published books in the field, for
example [39].
The discussion of the pulsed safe-operating area (SOA) in ESD pulse time
domain is presented in Section 3.3. Successive sections are focused primarily on
the most typical ESD device types.

70 3 Standard and ESD Devices in Integrated Process Technologies
3.1 ESD Specifics in Integrated Process Technology
3.1.1 Typical DGO CMOS Process with Extended Voltage

Components
Complementary MOS (CMOS) processes are the most popular low-cost processes
of choice for low-voltage analog design. In the case of the CMOS process,
previously developed intellectual property (IP) for digital blocks can be easily
integrated with analog blocks.
Taking advantage of accurate mask alignment provided by modern tools, over the
recent years these processes were upgraded with extended voltage devices. These
extended voltage devices (drain-extended MOS [DeMOS]) enabled the release of
analog power products with an operating voltage that significantly exceeds the low-
voltage gate oxide limits. For example, creation of non-self-aligned 12 and 20 V
NDeMOS and PDeMOS can be accomplished in the 5 V process within the same
mask set, while 40 V DeMOS can only be obtained using two additional mask layers
for an extended drift region, in order to satisfy hot carrier degradation reliability
requirements.
Thus, low-cost CMOS processes enable a combination of previously developed
IP for complex digital blocks and lateral high-voltage power devices to support dif-
ferent blocks of analog products: power trains, level shifters, low-dropout (LDO)
regulators, and other subcircuits.
Integrated process technology is usually subdivided into front-end and back-end
phases.
The front-end phase includes all the process steps (etch, diffusions, depositions)
from the initial substrate up to contact formation.
The back-end phase starts from the first metal layer. It includes formation of
via’s, metal layers and different back-end modules. These modules can include inte-
grated thin film resistors, copper top metals, high-density MIM capacitors, on-chip
inductances, laser trimmed fuses, MEMS, and others.
This section presents a simplified description of the front-end CMOS dual-gate
oxide (DGO) process flow required to support just six major standard active devices,
in combination with ESD specific discussion. The following device formation is
discussed:
i) low-voltage (LV) NMOS and PMOS devices for the digital core circuit;
ii) high-voltage (HV) NMOS and PMOS device for interface circuit blocks;
iii) extended voltage (NDeMOS and PDeMOS).
In addition to this, a set of major passive devices (inductors, diffusion resistors,

and polycapacitors) is usually formed during the front-end process phase to support
the analog product specification.
3.1 ESD Specifics in Integrated Process Technology 71
3.1.1.1 Initial Wafer Material

The process development usually starts with the selection of the initial substrate
material. To meet low-cost requirements, the substrate can be non-epitaxial, but
in most cases epitaxial substrate material is preferred to guarantee lower latch-up
spacing requirements. At the same time, in the case of high-Q inductors imple-
mented on the chip, a very high-resistant substrate is carefully selected to reduce
electromagnetic induction current losses.
Typically, p-substrate material is used in extended CMOS processes. Non-epi
substrate material is highly resistive. On the contrary, epi substrate material is usu-
ally formed on heavily doped p+ substrate with a 2–6 μm lightly doped p-epitaxial
layer.
In spite of being more expensive, epi material provides an active silicon layer of a
much better quality. From the analog design point of view, epi-material is preferred
because of the resulting relaxed latch-up spacing and guard ring design rules, as
well as much better sub-block isolation due to the usually grounded p+ substrate.
Respectively, the substrate material has a significant impact on the parameters of
the parasitic NPN devices in standard process devices. For example, characteristics
of NMOS devices depend on substrate material.
More expensive process technology option is bonded silicon-on-oxide (SOI)
wafers. In this case, an active silicon region is separated from the substrate by an
oxide layer. The former is isolated by a deep trench region (DTI), thus providing
completely isolated devices and circuit blocks. This type of substrate completely
eliminates the possibility of latch-up current path between isolated devices.
In high-voltage power devices, the substrate effect can impact the breakdown
voltage of the device due to induced charge in the substrate. Usually, it is considered
that buried oxide (BOX) provides isolation of ∼100 V per 1 μm thickness.
In the case of the SOI substrate in dc operation regime, the power dissipation
limitation should be taken into account, as it provides one of the major limitations
of power circuits. However, at least for a relatively thick silicon layer (∼4 μm), the
heat generation has a minor impact on operation under ESD pulse conditions, due
to a relatively small heat propagation range of ∼1–2 μm during ESD pulse. In this
case, elevated lattice temperature is present only in the subsurface area of the device.
Meanwhile, the effects of substrate-generated carriers and the spreading of
the injected current into the substrate can significantly change the characteristics
of the parasitic devices at ESD conditions or pulsed SOA limits, depending on
substrate material. For example, a snapback NMOS device can change character-
istics with different substrate resistance, due to a corresponding change of base
resistance in the parasitic n–p–n structure responsible for the avalanche–injection
process.
Typically, processes steps usually start from precleaning operations for the sur-
face with a high-temperature anneal of excessive oxygen, followed by the formation
of an alignment pattern using etch or laser marking. The alignment pattern is used
for mask alignment in the following process steps.
3.1.1.2 Device Isolation

The next steps in the CMOS process are intended to form a surface isolation between
the devices and device regions on the chip. The regions formed during isolation
include a shallow trench isolation (STI) and, if available, deep trench isolation
(DTI). Some older or low-cost processes can use junction isolation. Surface iso-
lation is useful because it provides process stability by eliminating hard-to-control
surface states and, at the same time, masks shallow implants, creating self-aligned
architecture.
An alternative surface isolation in low-cost high-voltage power process is thick
field oxide isolation (TFO) or LOCOS. This isolation approach avoids chemical
mechanical polishing (CMP) required for STI formation, but provides much bigger
minimum isolation dimensions and stress areas with in the so-called “bird’s beak”
region. In this region, broken bonds of silicon create high-density charge traps. The
last significantly impacts opportunities for extended voltage devices, unless special
measures are applied to the drift region profile.
STI and TFO can be used in very efficient ways for ESD purposes in order to
form free lateral NPN, SCR, and other devices with high-voltage tolerance (Section
3.3). By changing the isolation region dimensions within the physical minimum
process dimensions, appropriate voltage tolerance and triggering characteristics can
be controlled under appropriate yield.
There are many process variations that form shallow and deep trenches. One of
these versions is described below.
An STI isolation formation (Fig. 3.1) starts by Nitride and SiON deposition.
These layers are used to mask the silicon surface during etch and high temperature
cycles. Then, shallow trench etch removes nitride oxide and silicon at the desired
isolation depth (usually at 0.35–0.45 μm), followed by trench liner oxidation. This
important step is required to improve the quality of the interface and eliminate
interface trap regions that might otherwise significantly impact both the hot carrier
degradation (HCD) effects at normal operation conditions and breakdown voltage
walkout in extended voltage devices [40].
The following steps include HDP oxide deposition followed by a reverse
composite mask and an HDP oxide etch. The next step of isolation for-
mation is chemical mechanical polishing (CMP), which is required to pla-
narise the trench oxide filling regions. Finally, the remaining nitride layer is
etched.
Deep trench isolation involves more complex but similar process steps. These
steps are partly shared with STI formation. DTI may involve polysilicon trench
filing. The DTI in the SOI process is formed to merge with the buried oxide (BOX)
region. In high-voltage BCD processes, DTI depth can reach 15–30 μm or even
greater.
The remaining open silicon surface outside of the STI and DTI regions we will
call composite, referring to the corresponding mask layer as Composite Mask.
Since ESD devices at high current usually support the lateral operation mode, the
depth of the trench that defines the current path cannot be neglected.
N+ Buried Layer
STI STI STI
NMOS PMOS NMOSH NMOS PMOS NMOSH
p-Epi Substrate p-Epi Substrate
STI STI STI
PMOSH Isolated NMOSH

PMOSH p-Epi Substrate Isolated NMOSH
p-Epi Substrate
STI STI STI STI STI
PDeMOS p-Epi Substrate

NDeMOS PDeMOS p-Epi Substrate
NDeMOS
p+ n+ p+ n+
P-epi Pwell pldd p+ DnwellNwell nlld n+ p-poly n-Poly SiO2 Si3N4 P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4
Fig. 3.1 Process flow cross-sections for STI formation
3.1.1.3 Deep Nwell isolation

The following process steps provide the deep N-isolation layer. In this book, this
layer will be further called the Deep Nwell (DNWELL) in CMOS processes, or
n-isolation (NISO) in BiCMOS and BCD processes.
The initial process steps include the pad oxide region etch and growth of the
thin, ∼10 nm sacrificial oxide layer, followed by a corresponding masking step and
a high-energy implant of deep n-type species (Fig. 3.2).
In CMOS processes, DNWELL is used to provide vertical voltage isolation from
the p-substrate in the Pwell regions of NMOS, thus enabling the isolated version
of these devices. The same layer might be used to implement isolated PLDMOS
devices. In BCD processes, the NISO region is a deep region that provides isolation
of vertical PNP devices with a P-buried layer (PBL). Both layers are very useful
in suppressing noise and coupling between circuit blocks, and in presenting latch-
up isolation. Similarly, they are used to provide isolated ESD devices. The isolated
ESD devices are used for protection of the voltage nodes relatively to other biased
nodes, rather than the grounded substrate node.
In a snapback NMOS ESD device, the DNWELL can be efficiently used to
increase the intrinsic base resistance of the parasitic NPN devices in NMOS
snapback cells. Availability of the DNWELL implant region enables creation of
dual-direction ESD devices (Section 3.3). Without this region, the double blocking
STI STI STI
NMOS PMOS NMOSH
p-Epi Substrate
STI STI STI
deep Nwell
Isolated NMOSH
PMOSH
p-Epi Substrate
STI STI STI STI STI
deep N-well

NDeMOS
p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4
Fig. 3.2 Process flow cross-sections for DNWELL formation
junction regions cannot be realized in the structure. It is assumed that the formation
of the Nwell regions at the edge of the DNWELL implant is necessary to complete
lateral isolation of the enclosed Pwell regions.
For ESD and latch-up regimes, it should be taken into account that with an added
DNWELL region, an additional parasitic vertical PNP device is formed. In this par-
asitic structure, the isolated Pwell forms an emitter, the DNWELL an n-base, and
the grounded p+-substrate a collector. Depending on the thickness of the DNWELL
doping level, the gain of this parasitic PNP can be high, and thus cannot be simply
neglected. For example, in a dual-direction DIAC ESD device, the additional pos-
itive and negative feedback provides very asymmetrical characteristics at positive
and negative gate bias (Section 3.3).
A similar p–n–p structure is realized in BCD processes, where the PBL, NISO,
and p-substrate act as an emitter, base, and collector, respectively.
3.1.1.4 Well Implants

The most popular CMOS and BiCMOS processes usually include low-voltage and
high-voltage CMOS devices. The low voltage of the CMOS provides the advantage
of implementation of high-density logic circuit blocks. The high-voltage CMOS
devices are used for interface and power pins. Typically, the control and I/O pins
support one of the standard voltage levels, for example, 3.3 or 5 V. The voltage level
is used to drive the gate of the high-voltage lateral DeMOS or LDMOS devices.
Since low- and high-voltage MOS devices require different gate oxide thickness the
process is usually supports dual gate oxide (DGO) or triple gate oxide (TGO) in
order to allow appropriate scaling of the MOS device gate length.
In low-cost processes, in order to reduce the mask count, Nwell and Pwell regions
could be partially shared between LV and HV CMOS devices.
Often, a single drawn mask layer defines both well implants. Usually, for Nwell,
this layer is further used to produce an automatically generated Pwell mask. One of
the important parameters of high-voltage ESD device design is the spacing between
the Nwell and the generated Pwell. Variation of this spacing can provide formation
of the graded junction with the desired breakdown voltage level.
For practical ESD device design, it might be rather useful to interfere with the
generated layers and perform manual editing of the final generated layers. This
approach enables the formation of a variety of new free ESD devices optimized
for appropriate ESD protection. The only remaining limitations of such an effort
are the physical design rules of the process. Therefore, in-depth knowledge of the
particular integrated process technology is necessary.
The most powerful help during design comes from physical process simulations
using TCAD tools. In this case, the opportunities of the free ESD devices with
altered well implants can be efficiently explored, followed by the experimental test
chip verification.
In practical implementation, both the Nwell and Pwell are formed by a combina-
tion of several implant steps with different doses and energies (Fig. 3.3). Different
species might be used in order to form a complex optimal well profile. The profile
may include an anti-punch-through in-depth implant to improve latch-up prevention,
a buried channel, or the formation of a so-called empty well. Often, a retrograded
well profile is achieved by introducing the LV Nwell implant and the anti-punch-
trough (APT) n-type (Phosphorous) implant. During the next masking step, using a
Pwell blocking mask, both the LV Pwell and the anti-punch-through p-APT implant
set are applied using different dose and energy p-type (Boron) species.
Depending on the process guidelines, one or two additional masking steps are
done to create a threshold voltage implant for PMOS VTP n-type (Arsenic) and a
threshold voltage implant for NMOS VTN p-type (Boron). These shallow implants
form the upper portion of the wells. Thus, wells are formed in three regions: VT,
well, and APT.
When low-voltage wells are formed, similar steps are completed for high-voltage
wells HNWELL and HPwell.
p+ p+
STI STI STI STI STI STI
Pwell Nwell Pwell
hpwell hpwell
hpwell hnwell Hpwell
deep Nwell deep Nwell
PMOSH Isolated NMOSH PMOSH Isolated NMOSH

STI STI STI STI STI STI STI STI
hpwell hpwell hpwell hnwell hpwell hnwell

NDeMOS PDeMOS NDeMOS
p-Epi Substrate
p+ n+ p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4 P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4
Fig. 3.3 Process flow cross-sections for Pwell and Nwell implant
3.1.1.5 Gate Oxide

The next typical steps form the thin gate oxide layer for the LV CMOS device and
thick gate oxide for HV CMOS devices.
At first, the sacrificial oxide is removed, followed by growth and etch of the
recovery oxide. These steps are required to eliminate defects formed on the silicon
surface during well implant steps. After subsequent cleaning, HV gate oxide thermal
oxidation is first used to form the thick gate oxide.
In 3.3 V high-voltage CMOS devices, the oxide thickness is ∼6.8 nm. Then, a
gate oxide mask is developed, followed by gate oxide cleaning in the places of future
LV CMOS devices.
Thin LV Gate oxide formation is achieved by corresponding thermal oxidation.
A typical thickness of the gate oxide for 1.2 V devices in 0.13 μm processes is
∼2 nm. During LV gate oxide growth, the HV gate oxide continues to grow up to
the final thickness of 7 nm (Fig. 3.4).
3.1.1.6 Polygate
As soon as thermal oxidation is completed, polysilicon (poly) deposition is also
fulfilled, followed by Poly mask and etch. The etch stops on gate oxides for both
Nwell Pwell Nwell Pwell hpwell

hpwell
hnwell hnwell hpwell

hpwell

STI STI STI STI STI STI STI STI STI STI
hpwell hpwell hnwell hpwell hnwell hpwell hnwell

hnwell
PDeMOS NDeMOS PDeMOS NDeMOS

p+ n+ p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-pol n-Pol SiO2 Si2N4 P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-pol n-Pol SiO2 Si2N4
Fig. 3.4 Process flow cross-sections for gate thin oxide formation and poly deposition
LV and HV devices. The minimum length of the LV gate is the most challenging
optimization parameter to meet.
For ESD purposes, the devices can be produced using minimum LV gate length
in conjunction with HV poly regions. This enables creation of split gate structures,
gated diodes, and a control electrode in other ESD devices. However, the voltage tol-
erance of such devices both to dc voltage levels and to long-term reliability operation
requires additional assessment.
In free ESD devices with MOS regions, the gate oxide absolute maximum rating
is the major limiting factor for voltage tolerance. For example, a 1.2 V CMOS device
might provide a perfect snapback characteristic for 2.5 or 3 V pins. However, long-
term reliability concerns will not allow application of such a clamp in the product
due to gate oxide overstress.
Another ESD aspect related to gate oxide quality is robustness in short-pulse
electrical overstress. Depending on the quality of the dielectric and the process
itself, the maximum pulsed gate voltage before burnout may vary significantly.
Usually, in short pulse conditions, the critical voltage for gate oxide obeys the power
law [41].
Recently, more complex gate materials with a high-K dielectric and sophisticated
metal gate systems specific to NMOS and PMOS devices have been commercial-
ized. However, application of these processes so far is confined mostly to digital
products.
3.1.1.7 Lightly Doped Drain Implants

The scaling of the CMOS devices below 0.5 μm processes requires lightly doped
regions. The regions provide robust hot carrier degradation performance, allow a
high breakdown voltage, and eliminate gate induced drain leakage (GIDL) and drain
induced drain leakage (DIDL).
Before n- and p-lightly doped drain (NLDD and PLDD) implants are performed,
an oxide etch and clean from all silicon and poly areas is necessary, followed by
the application of the rapid thermal process (RTP) to seal poly and clean, creating
a few-nanometers-thick oxide on silicon in the place of all subsequent implants
(Fig. 3.5).
Nwell Pwell Nwell Pwell

hpwell hpwell

hpwell
deep nwell deep nwell

STI STI STI STI STI STI STI STI STI STI
hpwell hnwell hpwell hnwell hpwell hnwell hpwell hnwell
deep nwell deep nwell
PDeMOS NDeMOS PDeMOS p-Epi Substrate

NDeMOS
p-Epi Substrate
p+ n+ p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4 P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4
Fig. 3.5 Process flow cross-sections for pldd and nldd implant
For all LDD implants, the poly gate region acts as a masking layer that creates
self-aligned conditions.
In submicron CMOS process, an LV NLDD n-type is formed with arsenic, fol-
lowed by a Halo NLDD implant in the same opening. The halo implant in the LV
NMOS device is a p-type and made completely of BF2 . The HV device may receive
a separate HNLDD n-type implant (phosphorus).
Then, an LV PMOS PLDD mask p (BF2 ) and halo implant n-type (As) are
performed, followed by an HV HPLDD mask BF2 implant for the HV PMOS
device.
Both PLDD and NLDD lightly doped drain implants might be very useful in
creating ESD devices with a triggering voltage that corresponds to either n+-PLDD
or p+-NLDD.
Implementation of the free self-aligned avalanche diodes achieved by overlap-
ping the corresponding LDD mask layer with an n+ or p+ layer is very advantageous
because it forms a lateral surface structure. In 0.5 μm CMOS processes, the break-
down voltage of the devices is ∼6–8 V and therefore often suitable for the creation
of the 5 V avalanche diode and avalanche diode referenced snapback clamps. These
devices provide appropriate targeting of the ESD protection window by providing
a much better alternative to the n+ -Pwell and p+-Nwell blocking junctions with a
breakdown voltage of ∼11–14 V.
Creative manipulation of different LDD combinations may result in the creation
of other surface blocking junctions and surface BJT ESD devices with low-voltage
triggering characteristics.
The major limiting factor for a surface junction with an LDD region is the forma-
tion of surface states that may create unstable conditions for breakdown voltage and
enable walkout and abnormal leakage variation. Therefore, such free device design
requires further qualification that involves reliability testing and understanding of
the yield impact, especially if non-self-aligned layers are involved.
Another ESD application of the NLDD regions in ESD design is the formation
of surface saturation resistors.
3.1.1.8 Spacer Formation and NPLUS and PPLUS Implants

The following steps create the spacer region. Together with the POLY region, the
spacer provides an additional self-aligned extension of the polygate to mask NPLUS
and PPLUS low-energy implants at a certain distance from polygate edge. This
allows implants to be shifted some distance apart from the gate edge, thus form-
ing a desirable implant profile for electric field distribution, which reduces the hot
carrier degradation effect and increases breakdown voltage of the device.
There are many different recipes for spacer formation that depend on the tools
involved and the process architecture. One of the examples of spacer formation
is presented below. It includes TEOS oxide liner deposition forming a ∼ 20 nm
thermal oxide, followed by a ∼40 nm nitride deposition and addition of a 120 nm
thermal. Then, an anisotropic plasma etch of the oxide results in spacer formation,
followed by a nitride spacer chemical etch step (Fig. 3.6).
Thus, together with the poly region, the spacer masks subsequent NPLUS and
PPLUS implants.
After an NMOS source and drain mask, an n-type (arsenic) implant and graded
junction n-type implant (phosphorus) are made, followed by anneal. Then, the
corresponding PMOS source and drain mask and p-type (boron) implants form the
graded junctions.
Corresponding PPLUS and NPLUS implants are made to form well taps in the
CMOS devices.
n+ n+
p+ p+
STI STI p+ STI p+ p+
STI n+ n+
STI p+ STI
Nwell Pwell Nwell Pwell hpwell

hpwell
p+ p+ p+ p+ p+ n+ p+ n+ n+

hpwell

p+ p+ p+ p+ STI p+ n+ p+ n+ n+
STI STI STI STI STI STI STI STI STI
hpwell hnwell hpwell hnwell hpwell hnwell hpwell hnwell
PDeMOS NDeMOS PDeMOS NDeMOS

p+ n+ p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si 3N4 P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-pol y n-Poly SiO2 Si3N4
Fig. 3.6 Process flow cross-sections for spacer formation
The third spacer etch step consists of a dry oxide etch, which removes the spacer
footer oxide from the top of the poly and silicon, followed by cap oxide deposition.
3.1.1.9 Activation and Silicidation

Rapid thermal process anneal (RTP) is used to activate the completed shallow
implants. Then, a cap oxide is etched. After the etch, only the L-shape spacer is
left at the polygate edge of the CMOS devices (Fig. 3.7).
The silicide exclusion mask, SALEX, is developed to mask the area where sili-
cidation should be avoided. In standard devices, silicidation exclusion is used for
diffusion and poly resistors.
In ESD devices, the silicide exclusion region is absolutely necessary in order to
provide a drain ballasting region in snapback NMOS devices and a floating drain in
SCR devices.
In the case of cobalt material, the silicide is formed using cobalt precleaning for
the silicon and polysilicon surface in order to remove oxide. Then, cobalt deposition
is made (∼10 nm Co). Other additional layers, for example TiN, can be used as well.
After the first step of silicidation, RTP is completed. The cobalt that did not react
with silicon is removed, followed by the second silicidation anneal RTP step.
There are two important considerations for the silicidation process in ESD
devices.
p+ p+ n+ n+ p+ n+ n+
STI STI STI
Nwell Pwell hpwell
NMOS PMOS NMOSH
p-Epi Substrate
p+ p+ n+ p+ n+ n+
STI STI STI
hnwell hpwell
deep Nwell

p-Epi Substrate
p+ STI p+ n+ p+ n+ n+
STI STI STI STI
hpwell hnwell hpwell hnwell
deep Nwell
PDeMOS NDeMOS
p-Epi Substrate
p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4
Fig. 3.7 Process flow cross-sections for activation and silicidation
First of all, positioning of the silicide exclusion masking layer at the drain side is
critical for the performance of snapback ESD devices.
If the mask is aligned with the gate edge, then there is a risk of formation of a
very narrow silicide “sliver” region at the gate edge can result in low performance
of snapback NMOS devices, due to melting of the silicide at lower temperatures
during ESD stress and penetration in the active region of the device.
Such a narrow residual silicide region formed along the gate impacts local cur-
rent ballasting. Therefore, the maximum ESD current supported by such a device
is significantly lower. To overcome this issue, it is necessary to draw the silicide
exclusion mask layer overlapping the gate either completely or to the middle of the
gate. In ESD clamps with a grounded gate, this measure usually provides no dis-
advantage. However, an unsilicided gate region might be undesirable in high-speed
I/O where ESD and I/O functionality is shared within the same device structure. The
second option might be limited by the alignment tolerance.
Another ESD consideration for the silicidation process is related to the formation
of minimum composite regions as a part of the ESD devices, for example in ESD
diodes. In this case, artifacts from the nitride and oxide etches may reduce the actual
opening of the contact diffusion surface, thus significantly increasing the resistance.
To overcome this issue, ESD diodes might be optimized with a diffusion region size
larger than the minimum dimensions of the process.
3.1.1.10 Contacts and Backend

The contact formation steps include a self-aligned passivation layer deposition, a
pre-CMP deposition of 50 nm nitride, planarization by CMP, a contact mask, and a
contact etch (Fig. 3.8).
This is the last step of the front-end portion of the process flow.
p+ p+ n+ n+ p+ n+ n+
STI STI STI
Nwell Pwell
hpwell
NMOS PMOS NMOSH
p-Epi Substrate
p+ p+ n+ p+ n+ n+
STI STI STI
hnwell hpwell
deep Nwell

p-Epi Substrate
p+ p+ n+ p+ n+ n+
STI STI STI STI STI
hpwell hnwell hpwell hnwell
deep Nwell
PDeMOS NDeMOS
p-Epi Substrate
p+ n+
P-epi Pwell pldd p+ Dnwell Nwell nlld n+ p-poly n-Poly SiO2 Si3N4 a)
Metal2 Metal2 Metal2
W Plug
TFVIA
VIA1
TiW TiW
100A SiCCr
Metal1 b)
Fig. 3.8 Process flow cross-section for contact formation (a) and thin film resistor (b)
At the back-end, the ESD current density should be taken into account in order to
follow certain process-specific multiplication rules for the electromigration limits.
In the case of some processes, there exist limitations on stacking the vias and
contacts or vias on top of vias. ESD clamp design with backend consideration is
discussed in Chapter 4.
Another ESD aspect is the possibility of formation of ESD devices under the pad.
This usually requires a minimum number of metal layers on top of the device.
3.1.2 ESD Specific for BCD and BiCMOS Integrated Process Flow
3.1.2.1 Generic Process Flow
Similar cross-sections can be drawn for the BiCMOS and BCD process flow. To
reduce the volume of information presented in this chapter, these process flow
details will not be presented. However, most of these process steps can be under-
stood using the typical BCD devices’ cross-sections presented in Section 3.2.3, with
the combination of the brief discussion presented in this section.
From a cost perspective, BCD and BiCMOS processes provide a significant vari-
ety of different recipes, due to the effort to combine process performance with
optimized cost. Therefore, a typical way to integrate such processes is to share mask
layers steps between CMOS, BJT, and DMOS devices.
Another practical scenario that targets low cost and requires customized ESD
devices is the case of process application to the products that do not require all
mask steps. For example, some analog products in DGO BiCMOS process can be
implemented without process steps that create a low-voltage CMOS part, which
therefore can be removed from the process flow.
A somewhat generic integrated process BiCMOS flow is presented in Fig. 3.9.
In real cases, some steps can be different, depending on the specifications of the
required set of integrated devices and their parameters.
The starting epi-material usually has an orientation of <110>. Normally, the
P-type substrate is used with grown P-epi or N-epi. Alternative substrate can be
implemented using the silicon on oxide (SOI) process with a bonded wafer. For
0.5 μm analog processes, the typical epi thickness is 2–4 μm and the buried oxide
(BOX) thickness is ∼1–2 μm.
The BiCMOS process adds at least one vertical NPN BJT to the CMOS pro-
cess. In the BCD process, usually BJT, CMOS, and double-diffusion MOS (DMOS)
devices are formed.
Substrate selection for the process depends on the voltage and isolation spec.
Unlike in the CMOS process, one or more epi growth steps are required to form the
subcollector and base regions of the bipolar devices.
A typical set of available devices in the BCD process for high-voltage (HV)
power application includes power high-voltage devices NLDMOS, PLDMOS,
PMOS; complementary NPN and PNP devices; 0.5 μm 5 V CMOS devices; core
Starting Wafer Material
Buried Layer and Sinker Up

Formation
Epitaxial layer growth
Deep trench and Shallow

Trench Isolation
Well implants Dual gate oxide
CMOS
Selective growth Base Epi in
SiGe Processes
PNP Base BASE EMITTER COLLECTOR
Emitter Poly
STI STI STI
Silicidation, Schottky Diodes DT DT
Back End: Contacts,

Metallization, Passives a) b)
Fig. 3.9 Generic simplified integrated BiCMOS process flow (a) and cross-section for 0.24 μm
Si–Ge NPN BJT (b)
logic low voltage (LV) 0.5–0.13 μm CMOS; different HV and LV tolerant diodes;
optional thin film resistor (TFR), HV-isolated capacitor.
All devices must be substrate isolated above the high-voltage (HV) level on addi-
tional LV power supply level to enable boost circuit driver operation. A typical
advanced process for BCD process flow is presented in Fig. 3.10.
3.1.2.2 Subcollector and Substrate Isolation Regions

In order to support vertical BJT devices in the process and provide deep junction
isolation, there are several deep layers that are generated in the process.
Typically, a heavily doped N-type buried layer (NBL) and P-type buried layer
(PBL) are implanted prior to epi growth in the high-voltage process.
If the process needs to support high voltage, then the N-epi region thickness
is also high. In this case, to contact the buried layer, a stack of sinker layers is
required. For example, two sinker implants, NSINKERUP and PSINKERUP, can
be made prior to the epi growth or in the intermediate steps of the growth. Then,
after the epitaxial layer is fully grown, additional high-dose high-energy implants,
NSINKERDOWN and PSINKERDOWN, are made, in order to connect the buried
layers and the deep sinker region to the surface of the device.
To isolate the P-collector region in a PNP BJT device, an additional NISO region
is used, with the help of a high-energy implant. This implant completely isolates the
Starting Material CMOS Wells N+/P+/RTA
Buried Layers Dual Gate Oxide Salex/CoSi
Epi & Collector Base Implants Metal1

ThinFilm
Sinkers Poly Emitter Metal2
MiM Cap
STI and DTI LDD/Halo Implants Metal3
a)
N-Poly
Emitter Base
Collector
D
G
S
DT B
Deep
Trench
b) c) c)
Fig. 3.10 Example of process flow for BCD process (a) example cross-section of 20 V NPN
structure with deep trench isolation (b), and top view of deep-trench (DT) isolated 0.25 μm
CMOS (c)
PBL vertically. The NBL and NSINKER regions can perform lateral-junction-type
isolation. Alternatively, lateral isolation can be completed via the Deep Trench that
is available in the process.
An intelligent use of the buried layer and sinker implant is very productive in
implementing vertical and lateral self-aligned blocking junctions and avalanche
diodes suitable for ESD purposes. The vertical devices not only eliminate the effects
of the surface states but also provide advantageous in-depth heat dissipation in the
silicon region.
After completion of the N-epi growth, surface sinker regions are formed, fol-
lowed by an additional thermal anneal that results in diffusion of the implants in all
directions. The thermal processes provide not only a significant vertical diffusion of
the implanted species, but lateral diffusion as well. Therefore, for free ESD devices
with subcollector regions, it is important to take into account that the real bound-
aries of the diffused regions are significantly deviated from the originally drawn
mask layers.
Highly diffused NISO, NBL, and PBL implants can be patterned with the mini-
mum mask dimensions and thus control gain of the parasitic BJT device structures
and the breakdown voltage [42]. The last can be used as a reference for the triggering
voltage. The NSINKER can be used for current ballasting in ESD structures.
Depending on the actual process doping profiles and device spec, buried
avalanche ESD diodes can be formed using PSINKER-NBL or NSINKER-PBL
junctions.
The subcollector region provides for many possibilities of implementing different
SCR-type structures with in-depth current conduction, which results in withstanding
of much higher ESD power, due to heat dissipation in the bulk silicon material.
3.1.2.3 Isolation BCD Process Steps

Next step is similar to the CMOS process: the formed surface isolation can be a thick
field oxide or a shallow trench (STI), or both STI and DTI. Thus, the misalignment
of DTI and the deep layers must be accounted for.
Appropriate device isolation is a rather critical factor of the BCD process. At the
applied high voltage of ∼100 V to the regions on the top of the substrate, the injected
carriers can travel the distance of 300–700 μm. This effect can potentially initiate
such undesirable effects as latch-up, cross-talk, and interference between the ESD
devices and active circuitry, for example, with the switching power LDMOS array.
In the advanced BCD process, the trench, with a depth of ∼10–20 μm, is com-
bined with the buried and isolation regions. In a high-voltage process of 50–100 V,
an additional P-isolation region (PISO) is formed to connect the p-substrate, thus
providing high-voltage junction isolation in a thick epi-layer.
3.1.2.4 Collector and Initial CMOS Regions

Usually, the epitaxial region is of N-type and thus immediately provides n-collector
regions for vertical NPN devices.
For vertical PNP devices, the N-epi region is overcompensated by the P-collector
implant.
Taking into account the thermal budget, the CMOS module mask steps are per-
formed next, up to the level of well implants and dual gate oxide processing. The
major thermal budget contributors include the high temperature rapid thermal anneal
(RTA) required to form corresponding layers and regions.
Then, for example, in the SiGe process, selective base epi-growth is made,
followed by the formation of the polyemitter, remaining CMOS regions, shallow
implants and activation. Alternatively, the shallow internal and external base region
implants are made, followed by the formation of the polyemitter and polygates with
spacer processing.
3.2 Safe Operating Area in ESD Pulse Regime 87
The following steps include shallow implants used to form device contact
regions. Finally, different front-end modules are fabricated to support back-end
based passive elements.
The BiCMOS process provides lots of possibilities of engineering free ESD
devices. Required parameters within the ESD protection window range can be
achieved using a combination of different implanted regions in order to form
parasitic n–p–n, p–n–p, and p–n–p–n structures and appropriate blocking junctions.
Dual-direction devices based upon BJT base-emitter regions will be demon-
strated below. The presence of the subcollector region also allows the forming of
bipolar SCR devices (BSCR) [7].
Combination of the CMOS implants with the BJT part of the process might be
very useful as well. For example, the Nwell can be used to reduce the breakdown
and voltage of the standard NPN devices inside the ESD protection window.
The cross-section of the most typical CMOS and BCD devices, including para-
sitic structures responsible for save-operating area limitations, will be discussed in
the next section, thus proving additional insight into the BCD process flow.
3.2 Safe Operating Area in ESD Pulse Regime

Understanding of the pulsed SOA in ESD time domain is critical for successful ESD
design. Namely, control of the electrode-dependent voltage range between maxi-
mum operation regime and SOA determines the so-called ESD protection window.
In a simplified way, a certain voltage range is customary for ESD protection. The
range is confined between the absolute maximum operation voltage level of the ana-
log circuit pin at normal operation conditions and real absolute maximum for the
pins in the given ESD pulse regime. The last is practically determined by the pulsed
absolute maximum voltage of devices connected to the pin within specific ESD
test conditions. The test conditions may provide different coupling to the control
electrodes of the devices connected to the pins.
The definition of the absolute maximum limits assumes that the reversible oper-
ation of the devices is not guaranteed if the device operating regime exceeds the
limits. The product might be expected to fail due to electrical overstress (EOS)
if the absolute maximum limits for current and voltage at the pins are exceeded.
Respectively, in the ESD condition, some devices connected to the pin are expected
to fail if the pulsed voltage is over the limit, thus resulting in ESD failure.
Thus, if the ESD protection window is well understood, an ESD solution can be
engineered by implementation of the voltage limitation of the ESD device consistent
with the window.
In reality, understanding of the ESD protection window is much more complex.
For ESD pulse conditions, both lower and upper limits of the window are strong
functions of stress conditions. For example, ESD protection windows for the CDM
and HBM pulses might be totally different. This effect is related to the delay times
in coupling of the control electrodes and the time of the current transient itself.
Thus, the ESD protection window should be considered in the time domain
as signal waveforms that include many factors, even the history of previous ESD
events. Another critical aspect may be related to the fact that the ESD protection
window for the same pins can be very small and even negative depending on the
conditions of the internal circuit devices.
In general, the absolute maximum voltage of the supported device in ESD pulse
regime is limited by current instability phenomena, and the maximum level depends
on the control electrode conditions that control current incoming into the avalanche
multiplication area.
The conditions of the control electrode itself are transient in nature and dependent
on the internal circuit and ESD pulse waveform. They are often hard to identify even
when using ESD compact model analysis.
The next section highlights one of the methodologies adopted to overcome the
difficulties in defining the ESD protection window. At least for devices interfacing
with external pins, ESD pulsed safe operating area (SOA) can be and should be well
characterized in a wide range of control electrode parameters. This helps gain at
least a starting level of confidence in finding the correct ESD protection window.
Two methodologies measuring pulsed SOA in the ESD time domain are dis-
cussed below, after a description of the conventional understanding of the SOA in
the reliability field.
3.2.1 SOA and Current Instability Boundary in Reliability

To guarantee reliable long-term semiconductor device operation, a number of
constraints are usually imposed on operation regime.
In transistor structure, these constraints are dependent upon the electrical regime
determined by the applied voltage or current to the control electrode.
Usually more or less complete information about maximum and absolute maxi-
mum ratings for devices, as a function of their parameters and operation conditions,
is provided in electrical design rules (EDR) for integrated components, or in the
datasheet for discrete components. For example, the information can provide the
expected operation currents Imax , voltage Umax , dissipated power Pmax , and other
parameters specified for the given frequency, time domain, ambient temperature,
and other conditions.
The long-term reliability parameters of the device are guaranteed only within
maximum rating limits.
In the range between maximum and absolute maximum ratings, only the short-
term reversible operation of the device is guaranteed. Above the absolute maximum
rating, the device survivability is no longer guaranteed and the device may fail
irreversibly immediately after a limit violation in any time domain.
When an analog product is designed, the maximum and absolute maximum
ratings specified for each pin or pin groups, depending on the circuit design spec-
ification. In this case, the internal devices are used with appropriate margins that
may not even be related to the absolute maximum limits, but are limited by cir-
cuit performance: noise, linearity, functionally or simply a product spec that may be
way below the process limitation. Therefore, usually a certain margin in the voltage
domain can be expected when no ESD devices are connected. In real application,
these ESD devices might be the analog circuit components that provide physical
limitation of the absolute maximum voltage for the given analog circuit pins.
The reversible area of operation is represented by an area under a curve drawn in
I–U coordinates. This area is called the safe-operating area (SOA) for some given
conditions, for example, room temperature conditions.
Usually, SOA is plotted for guaranteed maximum ratings in order to provide the
customer with an SOA that is reliable for long-term operation. Similar plots can be
constructed for absolute maximum ratings and for different pulse and temperature
conditions.
The boundary of this SOA is determined by the limitations of the device in elec-
trical and thermal regimes. These limitations are determined by a broad spectrum of
electro-thermal physical effects.
The SOA for maximum ratings is limited in most cases by long-term reliability
parameters. Power dissipation may also provide the major limiting factor in a typical
device’s SOA. These factors are related to different degradation phenomena, for
example, mean time before failure (mtbf).
On the contrary, the SOA for absolute maximum ratings is based upon physical
limiting factors of a more rapid and instantaneous nature. These factors are related
to different electrothermal current instabilities that are realized in the conditions of
corresponding electro-thermal conductivity modulation effects [9].
In a simplified case, the SOA of output characteristics of an NMOS power
device at positive bias (Fig. 3.11a) is limited by three conditions: I≤Imax ; U≤Umax ;
I. U≤Pmax .
Fig. 3.11 SOA for dc (a) and pulsed (b) regimes: (1) I=Imax , (2) U=Umax , (3) P=Pmax (where tP
is the power pulse duration)
The current limitation I≤Imax is usually set by either backend electromigration

limits or the maximum saturation current that can be obtained from a device at
corresponding maximum gate bias (Fig. 3.11a, region 1).
On the contrary, in low power conditions, the voltage limitation U≤Umax
is usually determined by the isothermal avalanche breakdown and conductivity
modulation effects (Fig. 3.11a, region 2).
The third major limiting factor is the dissipated power limitation I. U≤Pmax .
Electrical, electro-thermal, or thermal instability effects initiated by Joule heating
usually determine this limitation. In dc operation, this limitation depends upon the
ability of the device to provide heat dissipation, for example, the thickness of the
semiconductor material and the efficiency of the heat sink.
A different scenario is introduced in pulsed operation regime. At shorter pulse or
a reduced duty factor, the thermal effects are reduced, providing a dominant role for
electrical instabilities [9]. In this case, the final irreversible breakdown or burnout
of the device is still related to local melting. However, the root cause of the local
melting is the electrically formed current filament state that subsequently provides
localized heat generation of such tremendous amplitude that it is capable of local
melting in the device, even in short pulse regime [9, 43–45].
The main concept of SOA makes sense if there is an operation regime with a
certain amount of time before failure. Apparently, SOA for device operation in dc
operation regime with, for example, tmtbf ∼ 105 h can be significantly different from
the SOA of the same device in pulsed regime, for example, with a pulse duration of
∼1 μs and low 0.1% duty factor, even if the same current–voltage regime is realized.
Then, SOA in a single nanosecond pulse domain can be significantly different from
that in a microsecond time domain. Thus, depending upon the regime, different SOA
can be constructed.
Furthermore, since SOA depends on the given operation regime, SOA extension
could be expected from most cases of device operation in the pulsed regime. This is
observed due to suppression of heat dissipation limitations (Fig. 3.11b). However, as
it will be shown below, this not always true in ESD pulse, due to electrical instability
and conductivity modulation that might be initialized by the dV/dt effect, result-
ing in significant displacement current in the device junction or control electrode
coupling.
3.2.2 Pulsed SOA for ESD Regimes

For normal operation regimes, long-term reliability parameters can be calculated
using accelerated test results. The physical mechanisms are related to different
degradation effects in the semiconductor structure and metallization; hot carrier
degradation, negative bias temperature instability (NBTI), electromigration, mutual
diffusion, and many others.
Such SOA limitation usually occurs due to various effects acting in different
time domains outside of the ESD pulse domain. Even in the case of electrothermal
instabilities, the typical time is comparable with the thermal heating time constant
in the device structure which falls in the microsecond time domain.
On the contrary, in the ESD test, the device is expected to withstand a limited
count of pulses in the 1–100 ns time domain. Moreover, after ESD test evaluation,
the components are no longer guaranteed for long-term operation parameters and
not shipped to the customers.
The goal of ESD SOA and instability boundary measurements is usually two-
fold: (i) evaluation of pulsed SOA for standard devices and (ii) measurement of the
instability boundary for ESD devices to help attain appropriate ESD clamp design.
3.2.2.1 Standard Devices

In standard package level spec ESD pulses (see Introduction), the rise time is usu-
ally within 0.1–10 ns and the pulse duration is in the time domain of ∼10–100 ns.
In addition, a significant time, ∼1 s, is provided between repetitive ESD zaps. Thus,
ESD stress can be considered as a single pulse event in terms of heat dissipation
in spite of the fact that some charge can be injected into the dielectric or inter-
face regions. Although, a more complex cumulative damage scenario can also be
observed (Chapter 4).
Thus, pulsed SOA in the ESD time domain (further as ESD SOA) is a SOA mea-
sured for specific pulse conditions. In principle, this SOA depends on the specific
pulse waveform and is expected to be different for HBM, MM, and CDM, as well
as IEC and CDE system level test pulses.
The most common characterization technique for ESD SOA evaluation combines
TLP measurements with constant gate bias. This technique is quite informative and
relatively widely available.
However, the level of confidence in these measurements is impacted by the cor-
relation between TLP and real ESD pulse conditions. A helpful consideration is that
the same TLP measurements are usually used for both ESD device evaluation and
pulsed SOA. Thus, the methodology provides rather informative practical results, as
demonstrated below.
The test setup for pulsed SOA evaluation using a standard TLP system is pre-
sented in Fig. 3.12. This technique applies TLP pulses to the device under test
(DUT), for example, NMOS device under constant gate bias on the gate electrode
(Fig. 3.12). The setup includes a power source to provide gate-source bias, base-
emitter bias, or current. Otherwise measurements are similar to conventional TLP.
DUT
TLP
System CGS VGS
Fig. 3.12 Setup for pulsed SOA evaluation using TLP measurements
BJT devices can be evaluated similarly with the initial conditions of first voltage
and then initial current for the base electrode. This mixed regime complicates a
possible extraction of the current gain, due to an unknown base current realized in
pulsed mode. However, as it will be demonstrated below, even in these conditions,

the pulse SOA can still be obtained, therefore helping the ESD design.
There are several limitations for this setup. First of all, only certain size devices
can be characterized, due to a limitation of TLP accuracy. Usually, these devices
are not less than 10 μm in total width, at the same time TLP voltage amplitude
is usually limited to ∼500 V. This limits both characterization of the power arrays
at a current below 10 Apms, and the ultra high-voltage device (UHV) within an
operating voltage above 500 V.
The system has internal load characteristics of 50 due to corresponding coaxial
cables used in its design.
Another problem of TLP SOA characterization is related to the control elec-
trode coupling. Usually, average TLP proportional voltage and current signals are
extracted from the TLP waveform in a range between 70 and 90 ns for the standard
100 ns pulse. However, in a relatively big device, for example, the NMOS array
parasitic gate-drain oupling capacitance is large enough to influence the test results.
Since the external gate bias voltage source is connected through the resistive
(∼0.1–1 ) wires with a parasitic inductance of ∼10 nH, the assumption of constant
gate bias during the TLP test is generally false. Before measurement in the 70–90 ns
time domain, the device can turn on in snapback. This results in an incorrectly low
measured critical voltage in comparison with true value.
To overcome this issue, it is necessary to prefabricate an additional gate-source
capacitor on the chip in order to hold the gate bias constant. All SOA presented in
this chapter are collected for a device with such decoupling (with an approximately
1 mm total width). The size of the capacitor can be obtained from a simple compact
model simulation. In CMOS devices, this size depends upon the ratio between the
parasitic gate-drain CGD and gate-source CGS capacitance in the device. In high-
voltage devices with long drift regions, the required on-chip capacitor is rather large
and comparable with the device size, if it is desired to keep gate voltage coupling
within half a volt.
3.2.3 ESD SOA for Typical Devices in BCD Process
Supported devices for analog circuits can be characterized using a pulsed SOA
approach.
The case of the 100 V BCD process is presented in Fig. 3.13, together with a
pulsed SOA measured using the TLP technique.
The exact set of the devices depends on the process spec. Examples of the process
in this section include the 7 V NMOS and PMOS device, the N-channel lateral
DMOS (NLDMOS) and PLDMOS devices, as well as the 20 V vertical NPN and
PNP devices.
In the NMOS (Fig. 3.13a) devices, a parasitic NPN is formed by the n+ -drain
acting as a collector, the n+ -source as an emitter, and the Pwell region with p+ -tap
diffusion as a p-base. Similarly, BJT regions can be identified in P-type devices.
a) b)
NMOS PMOS
GATE BIAS(V)
PULSED DRAIN CURRENT (A)

5 3
GATE BIAS(V)
0.3 4 0.3
2
0.2 0.2 3
2
0.1 0.1
1
1
0
0
0.0 0.0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
PULSED DRAIN-SOURCE VOLTAGE (V) PULSED DRAIN-SOURCE VOLTAGE (V)
c) d)
Fig. 3.13 Example of BCD process cross-sections and ESD SOA for LV NMOS (a, c), and PMOS
(b, d), respectively
TLP characteristics of the device at different gate bias are presented in

(Fig. 3.13c, d). In NMOS, the critical voltage for current instability is highly depen-
dent upon the gate bias. Due to avalanche multiplication of the channel current,
the instability condition αM>1 is true at much lower multiplication coefficients M,
which correspond to a lower drain voltage (Fig. 3.13c). In PMOS devices, the insta-
bility condition is much less dependent upon the channel current, due to the different
conditions for avalanche–injection at the reversed mobility ratio for majority and
minority carriers in the Nwell base (Fig. 3.13d).
In the case of NMOS device, the channel current determines the critical voltage
dependence upon the gate bias and controls the turn-on of the parasitic n–p–n struc-
ture in the conductivity modulation mode. Since the ballasting region is absent,
the device fails after snapback due to the possibility of unlimited current. The
destruction results in the local contact and semiconductor melting due to the current
filamentation effect.
Similar parasitic NPN and PNP structures can be recognized in high-voltage
devices.
The channel region on the top of PBODY diffusion forms a high-voltage

NLDMOS device (Fig. 2.14). Unlike the similar-in-purpose Pwell diffusion in
NDeMOS devices, this PBODY diffusion is self-aligned with the left gate edge and
thus provides stable characteristics for the MOS part of the device.
This second diffusion to the n+-source is essentially the reason for double-
diffusion MOS (DMOS) in the name of the device. The lateral device is further
combined with the drift region NDRIFT, which provides a similar drain exten-
sion function as the Nwell of the dedicated NEXT implant in the drain-extended
MOS device. The region is critical in providing the combination of a high break-
down voltage and low hot carrier degradation parameters, since at the applied high
voltage, the space charge region occupies most of NDRIFT. At the same time, the
most important characteristics for a power device, on-state resistance, is directly
related to drift region design. Different RESURF (reduced surface field) methods
[46] are used to improve device characteristics, starting with field electrodes, field
planes and junction geometry, to different multi-RESURF or super-junction [46, 47]
solutions.
Thus, parameters of the gate length, NDRIFT-POLY overlap (X), and drift region
length can control device SOA in a wide range.
Similar design considerations are true for the PLDMOS.
Unlike relatively short channel LV NMOS, NLDMOS has an embedded drift
region that is similar to a saturation resistor. That is, after avalanche breakdown, the
device may operate in avalanche mode with an additional voltage drop on the drift
region.
Typically, in high-voltage standard devices, irreversible failure immediately fol-
lows the triggering into snapback [48] (Fig. 2.14). Thus, the boundary line that
connects the critical voltage regime represents a current instability boundary. This
instability boundary limits ESD SOA and can be, respectively, applied as a reference
to devices of all sizes. However, the device topology may significantly change SOA.
The ESD SOA can be further used to establish the ESD protection window
as a function of the control electrode coupling provided by the internal circuit
during ESD stress. Thus, the voltage range between the maximum operation
voltage of the integrated circuit pin and pulsed SOA is defined with consid-
eration of the suitable ESD protection clamp that provides snapback in this
range.
One of the hardware-level difficulties for TLP-pulsed SOA measurements is the
functional test. In an ideal case, the conventional TLP leakage test can be automated
to allow a functional leakage test between zaps at zero gate bias.
The cross-sections and examples of SOA for representative 20 V BJT devices are
presented in Fig. 3.15 [49].
This particular example presents BJT with a polyemitter that is formed on the
top of a shallow-base implant.
Measurement of the pulsed SOA in BJT devices presents methodological diffi-
culties, since the voltage or current from the dc power source cannot be supported
in the pulsed regime. Thus, except in shorted and open base cases, it is difficult to
be sure about which conditions on the base correspond to a particular TLP plot.
LG LP
N+ Poly
TFO
P+ N+ N+
P Body X NDRIFT
Pwell PRESSURE Y
N-Epi
NBL
a)
0.05
GATE BIAS (V)
W = 1mm
0.04
12
0.03
10
7 5
0.02 4
3
2
0.01
1 0
0.00
0 10 20 30 40 50 60 70 80 90 100 110 120 130
PULSED DRAIN-SOURCE VOLTAGE (V) b)
Fig. 3.14 Cross-sections for 100 V NLDMOS (a) and measured pulsed SOA for 100 V
NLDMOS (b)
Nevertheless, the instability boundary is obtained according to experimental data,

and thus can be used to estimate the ESD protection window range (Figs. 3.13, 3.14
and 3.15).
3.2.3.1 Waveform SOA Measurements

An alternative method of evaluating SOA is based upon the HBM waveform capture
technique [33]. The tools for this technique have recently become commercially
available [32]. A single pulse waveform tester allows measurement of real ESD
pulse current and voltage waveforms.
This type of SOA measurement presents more complete information. As can be
seen, in addition to the information similar to TLP in the 70–90 ns domain, the phase
I–V characteristic represents both the overshoot and turn-off parts.
The main outcome from pulsed SOA characterization is accounting for the effect
of the control electrode coupling. Indeed, under general ESD test conditions, the
internal circuit may supply unknown conditions on the control electrodes of devices
connected to the analog circuit pin. For example, in the case of the NMOS device
a) b)
NEGATIVE COLLECTOR CURRENT (A/µm)

1.0 BASE CONDITIONS 1.0

BASE CONDITIONS
+5V
50mA
0.5
10mA 0.5
–0.7V
BVCEO
–5V 0V
0.
–50mA
1m
1mA –0.1mA
A
+0.7V 0V –10mA
0.0
–1mA
0.0
0 10 20 30 40 50 0 10 20 30 40 50 60
COLLECTOR PULSED VOLTAGE (V) NEGATIVE COLLECTOR-EMIT VOLTAGE (V)
c) d)
Fig. 3.15 Simplified cross-sections (a, b) and experimental pulsed SOA for typical 20 V NPN (c),
20 V PNP (d), respectively
(Fig. 3.13a) at zero gate bias, the maximum voltage is much higher in comparison
with the on state conditions.
3.2.4 Instability Boundary and SOA for ESD devices

In the case of standard devices, the ESD SOA is mainly required to ensure a suffi-
cient voltage range for local ESD protection implementation by choosing an ESD
clamp with the appropriate triggering voltage. In a different way, the SOA data can
be used to calculate an appropriate separation resistor for second-stage ESD protec-
tion, which allows use of the reversible current provided by the internal device for
internal voltage clamping (Chapter 4).
Excluding self-protection cases (Section 3.5), standard devices usually fail after
entering into snapback. This phenomenon is observed due to the absence of
current-limiting regions in most device structures. In some cases, if lateral cur-
rent transport at the surface is disabled, vertical NPN devices provide reversible
snapback operation due to internal subcollector ballasting.
Thus, in general, pulsed ESD SOA for standard devices supported in integrated
process technology is coincident with the current instability boundary. When the
electrical regimes reach the boundary, the positive feedback in the device structure
initiates a certain type of conductivity modulation, followed by the current filamen-
tation effect and creation of conditions sufficient for local heating and burnout.
This statement is not true for ESD devices. The ESD device is intentionally
designed to support reversible pulsed snapback operation in the conductivity modu-
lation mode. Therefore, in properly designed devices, the instability boundary does
not correspond to irreversible failure.
The way to achieve such operation is a direct subject of the ESD device (Sections
3.3, 3.4, and 3.5), topology, and clamp circuit design (Chapter 4). However, the
TLP or HBM waveform measurements are useful in establishing control of the ESD
device. This information can further be used to design ESD clamps with a biasing
circuit or to take advantage of dV/dt coupling of the control electrode. An example
of pulsed characteristics for NLDMOS-SCR is presented in Fig. 3.16.
a)
150 1.4 4
1.2 3.5 5kV TLP

Current 4kV
100
Voltage 1 3 3kV
Current [A]
Voltage [V]
Current [A]
50 2.5 2kV
0.8
2
0 0.6 1kV
1.5
0.4
1 500V
–50
0.2
0.5
–100 0 0
0 5 10 15 20 25 30 0 50 100 150 200 250
time [ns] Voltage [V]
b) c)
Fig. 3.16 Experimental set up for waveform SOA evaluation (a) and typical output for the wave-
form obtained for the 100 V NLDMOS device for the HBM pulse voltage waveform (b) and plotted
phase I–V characteristics (c)
Similarly to the standard devices certain understanding of the ESD SOA for
ESD devices is useful. After snapback in high-current state, device characteristics at
different gate bias are practically merged. This fact reflects a rather simple conclu-
sion that in high injection conductivity modulation regime, the device is practically
uncontrollable, due to in-depth current conduction. As a result, in MOS devices, the
effect of surface control electrode becomes negligible. A similar situation is realized
in BJT devices, due to high injection conditions.
For snapback ESD devices, SOA is usually confined in the vicinity of one critical
regime. This regime corresponds to the next stage of conductivity modulation.
Thus, reversible operation of ESD devices in secondary breakdown mode presents
a “normal” ESD operation regime.
Physical limitations of the operation regime for the ESD device itself are related
to the same physical effects as those that affect standard devices [9]. However, these
effects are usually realized at high current due to the conductivity modulation of the
contact regions, rather than the active and drift regions of the device. In this case,
the modulation is occurring between the contact metal and the semiconductor part
of the structure in high injection state. This creates a physical equivalent of parasitic
structures M-i-n+ or M-i-p+.
In case of semiconductor structure failure, the problem can be reduced to the
conductivity modulation of the contact regions. This scenario of evolution results
in much higher injection from the metal electrodes. The positive feedback results in
the exceeding of contact and metallization limits and a localized heating and melting
due to the current filamentation effect.
These phenomena are related to modulation of the contact region, fast elec-
trothermal instabilities in the pulsed regime due to high-power heating of the active
layers, and damage of the silicon structure by optical phonons and pieso-electric
effects [9]. In opposite, the irreversible failure scenario involves electromigration
limits of the contacts and metallization that may finally act similar to a fuse and
create an open circuit event. A similar effect might happen in the case of bond wire
failure. If backend limitations are removed, in principle, the physical limits of the
ESD device operation are related to the energy dissipation in the silicon material
itself, under the condition of typical heat generation within the few-microns deep
subsurface regions. Some aspects of such limitations are discussed below.
3.2.5 Physical Limitation of ESD Devices. Spatial Thermal

Runaway
A complete scenario of electro-thermal conductivity modulation that includes cur-

rent and spatial electrical and thermal instability for the case of ESD operation has
been demonstrated in [50]. The case of SiGe BJT devices in 0.24 μm BiCMOS
processes has been studied.
According to the experimental TLP collector-emitter I–V characteristic
(Fig. 3.17a), local burnout has been observed at some critical pulsed current
(Fig. 3.17b).
In order to explain phenomena that limit device operation in high-conductivity
mode in the case of ESD time domain, analysis has been completed using a ther-
mal coupled mixed-mode simulation for the finite element device with geometry
and implant profiles of a 0.24 μm BiCMOS process (Fig. 3.18a). A quasi 3D
simulation approach (Section 2.7) has been applied to a distributed multiple-cell
structure. The distributed structure has been “assembled” from segments of the
original BJT cross-section obtained from calibrated physical process simulation
10 0 BASE
TLP CURRENT (A)
EMITTER
10 –1
10 –2 DAMAGE
COLLECTOR
10 –3 1µm
0 2 4 6 8 10 12 14
COLLECTOR VOLTAGE (V)
a) b)
Fig. 3.17 TLP snapback curves for BJT pulse operation and SEM photos of the damaged device
Si-Ge SPACER POLYEMITTER

–0.5
–0.4
EX-BASE
0
Y(μm)
Y(μm)
–0.2
CONTACTS TO SiGe BASE
0 OXIDE POLYEMITTERS WITH TOP CONTACTS
0.5
0.2
N-EPI LAYER COLLECTOR REGION WITH BOTTOM CONTACT
2.8 3 3.4 2 3 4
X(μm) X(μm)
a) b)
Fig. 3.18 Cross-section of the original NPN BJT (a) and part of 9.6 μm distributed structure (b)
obtained by multiple reflections [50]
flow (Fig. 3.18b), thus providing a mathematical approximation for the general 3D
problem and representing the case of multifinger BJT array.
For visualization of the current stratification effect in transient conditions, the
maximum lattice temperature values have been extracted for the sequence of 1 μm
regions across the Z=10 μm structure width. These dependencies of the values
T1 . . .T10 upon time have been compared for the cut lines with Z1 . . .Z10 =1. . .10 μm,
respectively. At the uniform temperature distribution, all these dependencies were
coincident. The appearance of the difference between them indicated formation of
a spatially non-uniform solution.
The results of the thermal coupled mixed-mode simulation in the quasi-3D
structure have been compared to the 2D cross-section of the BJT device (Fig. 3.18a).
In the case of 2D simulation for the Si–Ge NPN BJT cross-section, the device
is triggered in the snapback regime after reaching a critical triggering voltage
CURRENT (A)
VOLTAGE (V)
6
4
4
VCE
2
2
0 IC
0
10–9 10–8 10–7
TIME (s)
MAX. LATTICE TEMP. (100K)
8
6kV
7
6
4kV
5
2kV
4
3
a) b)
Fig. 3.19 Transient characteristics for the ESD protection clamp with BJT structure (Fig. 3.18) at
area factor 200 for different HBM pulse amplitudes (a) and the lattice temperature distribution in
the 2D device after 100 ns from the beginning of the 6 kV HBM pulse (b) [50]
(Fig. 3.19). Due to a very short ESD time domain, the heat generation is local-
ized in a very narrow vicinity of the emitter-base region (∼1 μm) (Fig. 3.19b). With
increase of the ESD pulse amplitude, the peak temperature increases (Fig. 3.19a)
according to the applied power to the device and at some point can reach the crit-
ical level for local structure melting or other type of material damage. However,
in this 2D case, the current density along the imagined finger width is assumed
uniform. Thus, the results of this simulation present a significant overestimation in
comparison with the real effects that involve the short pulse electro-thermal current
filamentation effect. This effect has been revealed using a solution for the quasi 3D
structure.
In the quasi 3D structure, current filamentation is allowed and essentially proves
that real physical limitation occurs at a much lower level. The quasi-3D BJT struc-
ture (Fig. 3.17b) generally provides a uniform response to ESD stress only in the
range of HBM ESD pulse amplitude below a certain critical level. In this case, the
solution is uniform across the sample width and all extracted temperatures are the
same T1 =T2 =T3 . . .=T11 (Fig. 3.20a). Above some critical ESD pulse amplitude, a
different scenario is observed.
In this case, the extracted values of maximum lattice temperatures at different Z
coordinates in the different 1 μm parts of the device indicate splitting of the uniform
numerical solution (Fig. 3.20b), thus indicating a current stratification effect.
The current stratification effect results in the formation of a solitary hot spot
(Fig. 3.21). This hot spot is formed at a much lower HBM pulse level than in the 2D
8 8 8 8
7 7 7 7
CURRENT (A)
CURRENT (A)
VOLTAGE (V)
VOLTAGE(V)
6 VC 6 6 VC 6
5 5 5 5
4 4 4 4
3 3 3 3
2 2 2 2
1 IC 1 1 IC 1
0 0 0 0
10–9 10–8 10–7 10–9 10–8 10–7
TIME (s) TIME (s)
10

9 T7
9
8
8 T6
7 T2=T3=…=T11 7
T8
6 6 T5
T9
5 5
T10,T11
4 4 T2,T3,T4
3
a) 3 b)
Fig. 3.20 Mixed-mode transient ESD characteristics for the protection clamp with the equivalent
W=9.6 μm distributed BJT structure at area factor 42 for 2 kV (a) and 2.8 kV HBM pulse (b)
Fig. 3.21 The depth profiles for lattice temperature distribution in the distributed BJT equivalent
structure after 100 ns of 2.8 kV HBM ESD stress
simulation structure that assumes uniform current distribution (Fig. 3.19). A criti-
cal regime for such electrothermal current filament formation is observed (∼HBM
2.7 kV), which corresponds to the uniform distribution of the maximum temperature
of ∼ 700 K along the device finger (in the X-direction).
As soon as current filament is formed, a catastrophic event should be expected,
due to enormous peak lattice temperature of >1500 K generated in the filament
region (Fig. 3.21).
3.3 Low-Voltage ESD Devices in CMOS Processes

Under significant simplification, the ESD device design principle can be explained
in a very straightforward way. It consists of implementation of a certain “parasitic”
elementary structure that can support conductivity modulation (Chapter 2) inside
either the supported device structure, or in a “free” ESD device, in the conditions of
local current density limitation.
Referring to the previous SOA section, this practically means that the design
of a new pulsed semiconductor device is expected to contain both the reversible
instability boundary and desired instability boundary parameters.
The following sections of this chapter will discuss device-level design, while
Chapter 4 will address the topological solutions referring to device layout.
The problem of proper ESD device design is mainly confined within several
device engineering tasks
(i) provide an energy balance inside the device structure;

(ii) provide proper current balance at the clamp level that will enable width
scaling;
(iii) obtain a solution that will combine triggering and holding voltage parameters
and dc voltage tolerance, and produce an adequate voltage limitation wave-
form on the protected pin, while not interfering with the circuit during normal
operation regime.
Practically pulsed ESD SOA for standard devices is required both to identify the
spec for the desired ESD device and simultaneously to produce a starting point for
free ESD device design.
One of the primary goals of device-level ESD development is to experimentally
obtain a primitive ESD device to be used as a component in the future ESD clamp.
Essentially, this means to obtain an ESD device by adding new regions and making
transformations in the supported integrated device. This approach is demonstrated
in this section for low-voltage devices.
At the same time, there are no direct guidelines that define low-voltage and
high-voltage devices. For example, in the CMOS dual gate oxide process, low-
voltage devices with a thin gate oxide are usually called low-voltage, while the
devices with a thick gate oxide and corresponding high-voltage tolerance are called
high voltage. This definition has been used for CMOS process description in
Section 3.1.
The following section will use a different definition for high- and low-voltage
devices. Since the majority of current analog products are based on 0.18–0.5 μm-
process technology, we will rely on one of the most conventional definitions for the
LV and HV in this book. A voltage of ∼12 V will be taken as the dividing line
between the low- and high-voltage ESD devices. This voltage level approximately
corresponds to the n+-Pwell and p+-Nwell breakdown voltage in 0.13–0.5 μm
processes.
3.3 Low-Voltage ESD Devices in CMOS Processes 103
3.3.1 Snapback NMOS

Historically, the snapback NMOS ESD device was widely used in integrated CMOS
processes as a component of the grounded gate NMOS clamp (Fig. 3.22). This
device is based upon the avalanche–injection conductivity modulation mechanism
in the parasitic n–p–n structure that is already present in the NMOS device. This
n–p–n structure is formed by the source acting as an emitter, the drain acting as a
collector, and the Pwell acting as base with a corresponding p+ -tap contact region.
The base length and gain of the n–p–n structure is controlled by the gate length
LG , while the internal base resistance RPW is determined by bulk contact to source
spacing LSWS (Fig. 3.22).
LSWS LG LSB
Bulk Source Gate Drain
p+ STI n+ n+
RPW
npn Rn+
Pwell
NBL or DeepNwell
P-sub
Fig. 3.22 Cross-section for snapback NMOS device in a non-silicided CMOS process with a
parasitic n–p–n structure
In triple well CMOS process optional substrate isolation can be achieved by the
DeepNwell region with corresponding lateral Nwell isolation. In case of BiCMOS
process N-buried layer can be used for the same purpose.
In spite of simplicity, this device combines several important design features.
First of all, the gate electrode can be coupled to the source through an addi-
tional resistor RGS . During the fast rise time of the ESD pulse, the gate electrode
is biased above the threshold voltage by the current through the CGS /CGD capac-
itive divider that creates a voltage drop on the resistor, where CGS and CGD are
the gate-source and the gate-drain parasitic capacitances. In accordance with pulsed
SOA (Fig. 3.13a), due to the multiplication of the channel current, the conductivity
modulation is realized at a lower multiplication coefficient, thus providing a lower
triggering voltage in comparison with the dc breakdown voltage.
Thus, snapback NMOS in grounded gate configuration represents a transient
triggering solution. In addition the bulk coupling effect provides for the dV/dt
turn-on at a lower voltage level than expected from the supported similar NMOS
device with the same drain diffusion length. Due to the long ballasting region, this
effect is significantly stronger than in standard devices with symmetrical source and
drain regions. A gate resistor is required to disable the clamp at the normal oper-
ation region. However, the gate coupling in this clamp should be considered at a
high-speed I/O design to avoid parasitic current conduction.
The second feature is the drain ballasting region. The region provides satura-
tion resistor Rn+ with the parameters controlled by the ballasting region length LSB
(Fig. 3.22). This region is required to limit local drain current in the conductivity
modulation mode, thus providing a reversible operation in a wide range of the bal-
lasting region length (usually 1–8 μm). To obtain the region in silicided processes, a
silicide exclusion SALEX mask is usually used. It is, however, important to provide
appropriate mask alignment that will completely remove silicide at the poly edge,
as discussed in Section 3.1.
Finally, the base (Pwell) diffusion region placement defines the internal base
resistance of the parasitic n–p–n device. This parameter is critical in achieving
avalanche–injection instability conditions.
Thus, snapback NMOS ESD device operation is based on the parasitic n–p–n
structure, with additional current induced by the channel electron current.
NMOS clamp layout design specific will be discussed in Chapter 4.
3.3.1.1 Three-Dimensional Simulation of Current Instability in Snapback

NMOS Devices
The physical limitation of the snapback NMOS device is similar to the pulsed
electro-thermal effects discussed previously for the case of Si–Ge NPN BJT.
The original solution for physical effects that result in catastrophic failure of the
NMOS structure has been obtained using a full 3D simulation [50].
The case of the rather simplified 3D NMOS structure with a 10 μm contact
width (Fig. 3.23a, b) has been studied. Calculated ESD HBM transient character-
istics of this simulation object are presented in Fig. 3.23c. Starting from HBM pulse
amplitude of ∼ 6 kV, the maximum lattice temperature becomes non-uniform, rep-
resenting the current filament and corresponding hot spot formation phenomenon.
A cut-line along the drain junction under the gate demonstrates the spatial electro-
thermal current instability of the drain current evolving with time, as well as the
solitary hot spot in the structure (Fig. 3.23). Kinetics of current stratification is
presented in Fig. 3.24.
Both NPN BJT and NMOS devices have the same physical limitation of ESD
device operation regime. It is related to a rather fast thermo-electrical current fila-
mentation of the ballasting subcollector or drain regions. At the same time, before
the critical level, the avalanche–injection conductivity modulation is balanced below
the critical level by either the sub-collector or drain ballasting region.
Typically, the current limit of the ESD device operation itself is not a limiting
factor of ESD protection capability for a given circuit pin. Depending on design,
the ESD device can withstand rather high-current levels at high voltage. Therefore,
before the damage of the ESD device itself, the internal circuit can fail due to exces-
sive voltage at the pin. Thus, specifically the waveform of the ESD voltage limitation
produced by the ESD device is a major concern. Roughly, the waveform is projected
from the TLP measurement as a function of the triggering and holding voltages and
on-state resistance of the device in the high-current state after snapback.
POLYGATE
CE
UR
SO N
AI
DR
Z
Y X TEOS
–0.2 SPACER
POLYGATE
Y(μm)
SUBSTRATE 0
n+ -SOURCE n+ -DRAIN
0.2
Pwell
0 0.5 1 X(μm)
a) b)
972
10
Lattice Temperature (K)

Z(μm)
8
7
14 24
LATTICE TEMPERATURE (100K)
6 22
12 6
20
5
10 18
VOLTAGE (V)
1104
CURRENT (A)
534
748
4 V DS 16 432
8 4
14
3 6 12
ID
1206
Tmax 4 10
2 2
8
VGS 2
1 6
0 4
0 0
0 10 20 30 40 50 60
TIME (s) x(μm) 2 4 6
c) d)
Fig. 3.23 Three-dimensional NMOS structure (a) and the cut plane in the gate region vicinity
(b). Transient ESD characteristics (the drain voltage, the drain current, and the maximum lattice
temperature) of the 10 μm 3D NMOS structure with an area factor of 50 for 10 kV HBM (c) and
the cut plane representing hot spot formation at Y=2 μm (d)
3.3.2 FOX (TFO) ESD device

The voltage tolerance of the snapback NMOS is limited by the maximum rating of
the gate oxide. Therefore, in case of higher voltage tolerance, a different device-
level solution is required, depending on the voltage level. In some cases stacked
8e+6
X = 0.92μm; Y = 0.05μm
1500
ELECTRON CURRENT (units)

6e+6
TEMPERATURE (K) T
1000
4e+6
500
IE 2e+6
0 0e+0
0 2 4 6 8 10
Z(μm)
Fig. 3.24 Kinetics of current stratification obtained from the 3D simulation solution
clamps can be exploited. However, if I/O or power supply voltage is above the gate
oxide limits, the best suitable structure is the so-called field-oxide (FOX), thick field
oxide (TFO), or STI device. The device presents itself as a free lateral NPN, formed
by CMOS diffusion and the surface isolation region in the Pwell.
A typical example of such device application is the EEPROM write pins. These
pins usually require an elevated voltage level above the gate oxide maximum relia-
bility rating, for example, 6–7 V in 5 V NMOS devices. While the snapback NMOS
device is capable of providing an appropriate triggering voltage, it is no longer
useful, due to reliability constrains at normal operation regime.
In the case of the CMOS process, FOX is the next most logical ESD device
structure. The FOX device can be obtained through transformation of the snapback
NMOS device, by swapping the polygate region with a minimum dimension surface
isolation region, for example, in the shallow trench isolation with a length of LSTI
∼0.25–05 μm (Fig. 3.25). Often, names similar to original snapback NMOS are
LSWS LSTI LSB
Bulk Source Drain
p+ STI n+ STI n+
RPW
npn Rn+
Pwell
NBL or DeepNwell
P-sub
Fig. 3.25 Cross-section for snapback FOX device in CMOS process with indicated parasitic n–p–
n structure
preserved for the electrodes and regions of this device. In this case, the n-emitter
can still be called as source, the collector region as drain, while the P-base is called
a P-well or a Bulk.
Similarly to snapback NMOS, the device is self-aligned, since the n+-implant is
not penetrating through the surface isolation oxide. The remaining design features
are the same as in the snapback NMOS. To limit the current density increase in
the conductivity modulation regime, the device requires the n+ -ballasting region on
the local level. This region is usually selected based upon experimental results in
the same range of ∼1–6 μm, depending on the Pwell and n+-doping regions. The
well tap placement is designed with large spacing to the source (n-emitter) of the
structure to provide a high intrinsic Pwell (p-base) resistance, thus controlling the
triggering voltage.
In the case of SOI processes, the Pwell contact can be treated similarly to the
base of NPN. In this case, taking the advantage of substrate isolation, the device can
be connected in a common emitter circuit with an additional resistor and optional
voltage reference device (see Chapter 4).
There are two ways to tune the device turn-on in high-conductivity state: (i)
by using the dV/dt effect occurring due to displacement current in the p-base-n-
collector (drain-well) junction in ESD regime; and (ii) by adjusting the implant
profile at the collector region to control the breakdown voltage of the device.
The FOX device presents a more compact solution in comparison with the
stacked device. Often, a precise adjustment of the breakdown and triggering voltage
is required. One of the practical ways to implement such a solution is to use the
Nwell implant to create a graded junction (Fig. 3.26a). The major concern in con-
junction with this solution is non-self-aligned device architecture and understanding
of the triggering voltage variation upon Nwell position. It is important to understand
the impact of the Nwell–Pwell masks misalignment. An example of the experimen-
tal implementation of the HV FOX device with a graded Nwell (n+)-Pwell junction
is presented in Fig. 3.26b. Within standard misalignment for a process of ∼0.1 μm,
the 12 V voltage tolerance can reliably provide a deviation within ∼11–13 V.
Both the snapback NMOS and the FOX device have similar parasitic n–p–
n structures with relatively low-current gain due to a lengthy p-base (p-well)
of ∼0.5 μm. Therefore, the results and understanding of Chapter 2 is directly
applicable in describing the physical effects of avalanche–injection conductivity
modulation and spatial current distribution in these devices. To reduce the thresh-
old of instability, a large internal and often external base resistance is implemented.
To limit the current density on the local level, a distributed n+-ballasting region
is used.
3.3.2.1 Surface NPN

An alternative self-aligned device can be engineered as a surface NPN using a corre-
sponding n+-mask (Fig. 3.26c). For high voltage, a self-aligned n+–n+ spacing can
control the device triggering characteristics. For low voltage, if the process allows,
a PLDD blanket implant can be done at the surface.
LSWS LSTI LSB

Bulk Source X Drain
p+ STI n+ STI n+
npn
Pwell Nwell
NBL or DeepNwell
P-sub
a)
Nwell-STI Overlap X (μm)

b)
LSWS LSTI LSB
Bulk Source Drain
p+ STI n+ pldd n+
RPW
npn Rn+
Pwell
NBL or DeepNwell
P-sub
c)
Fig. 3.26 Cross-section snapback FOX ESD device with graded Nwell junction (a) and depen-
dence of the breakdown and triggering voltage as a function of Nwell overlap parameter X (b).
Implementation of the surface NPN device with PLDD (c)
The practical problem of such devices is the unstable characteristics that are very
dependent upon process variation, due to formation of the surface states. The surface
states can impact the device breakdown voltage by walkout effects. The walkout
effect can be realized as a result of charge trapping in the lightly doped surface
region. Another limitation is the minimum space between two n+- implants, which
might be too big for appropriate device parameters to be achieved.
A disadvantage of the device is the worse heat dissipation conditions, due to
confinement of the current at the silicon surface.
3.3.3 LVTSCR and FOXSCR

The next class of snapback ESD devices compatible with CMOS processes is the
silicon-controlled rectifier (SCR) devices. These devices combine a parasitic n–p–n
and parasitic p–n–p structure that form an internal parasitic p–n–p–n structure
(Chapter 2). Therefore, the numerical analysis and phenomenological description
are applicable to the analysis of the physical processes in these devices.
There are several ways to obtain SCR in the CMOS process. In the NMOS start-
ing structure, the device transformation requires implementation of the p+-emitter
region isolated by the Nwell (Fig. 3.27a). In the case of the PMOS starting structure,
the implementation requires additional substrate isolation by deep N-well or Niso
regions. In this case, the Pwell-isolated n+-emitter can be embedded into the device
in order to form the p–n–p–n parasitic structure with SCR properties (Fig. 3.27c).
The Pwell that is isolating the n+-emitter requires not only vertical isolation from
the p-substrate but also lateral isolation by corresponding Nwell regions.
10 –1
ANODE P+ REGION LENGTH LP ( m)
LSWS LG LF LFP LP LPD
ANODE CURRENT (A/um)
10 –2
Bulk Source Gate P-Emitter Drain 2.5 1 0.8 0.6
p+ STI
n+ n+ STI p+ STI n+ 10 –3
RPW n-p-n
0
10 –4
RNW
p-n-p
Pwell Nwell 10 –5
P-sub 10 –6
a) 0 2 4 6 8 10
DRAIN (ANODE) VOLTAGE (V) b)
LG LF LFN LN LND
LSWS LSTI LF LFP LP LPD
Bulk Source Gate N-Emitter Drain
Bulk Source P-Emitter Drain
n+ STI p+ p+ STI n+ STI p+
p-n-p
RNW p+ STI n+ STI n+ STI p+ STI n+
RPW
Nwell n-p-n Pwell RPW n-p-n
RNW
NBL or DeepNwell p-n-p
Pwell Nwell
P-sub P-sub
c) d)
Fig. 3.27 Cross-section for LVTSCR (a) in a 0.5 μm process and simulated I–V characteristics
at gate bias 1 V for different P-emitter length (b). The cross-section of p-LVTSCR device (c) and
FOXSCR (d)
Historically, one of the most popular implementations of the SCR device for
the CMOS process has been suggested in [51] with the name low-voltage sili-
con controlled rectifier (LVTSCR). This device can be obtained as a free device
in typical CMOS process technology using a simple modification of the snapback
NMOS within limitations of the process layout design rules. The LVTSCR device
incorporates additional p–n–p structure achieved by the introduction of an isolated
p+-emitter region. As a result of this modification, the new equivalent circuit of the
device includes both parasitic n–p–n and p–n–p structures with well resistors RNW
and RPW (Fig. 3.27a). In addition to the LSWS and LG parameters that control the
characteristics of the n–p–n structures, several new parameters control the charac-
teristics of the p–n–p structure. These parameters include the floating drain region
LF , the p-emitter length LP , and corresponding spacing LFP and LPD (Fig. 3.27a). An
example of the holding voltage control in the LVTSCR device by the anode region
length is presented in Fig. 3.27b. According to simulation results, both the trigger-
ing current and holding voltage can be substantially changed by the reduction of the
gain of the p–n–p structure.
The p-emitter isolated by Nwell can approach the n+-source rather close and
avoid punch through. At the same time, the NMOS structure provides the starting
current in the avalanche multiplication region at the drain junction. Therefore, to
combine both SCR and NMOS in a single device, a floating n+- region is required
to separate the Nwell from the n+-source.
The n+-floating drain region solves the misalignment problem as well. Small
variations of the Nwell position across the floating drain region within mask align-
ment tolerance provide only a minor variation of the breakdown and triggering char-
acteristics of the device. These characteristics are defined by the NMOS structure.
The principle of LVTSCR operation consists in turning on the device into the
double injection conductivity mode with a – typical for this mode – very low holding
voltage of ∼1.5–2 V (Fig. 3.27b). Thus, the device provides completely different
voltage waveforms in comparison with snapback NMOS or FOX devices. Due to a
much lower holding voltage, during the ESD pulse, the voltage is limited at a much
lower level.
The process of triggering into double injection conductivity mode depends upon
the device design and implant profiles. In properly designed devices, a domino-like
effect is expected and can be demonstrated by device simulation. This domino effect
can be explained as follows.
At first, the fast rise time of ESD pulse provides gate coupling above the threshold
voltage, which results in the formation of the accumulation channel and source-
floating drain current conduction through the Nwell region toward the n+-contact.
If the Nwell doping is low, this current can already lower the potential barrier at
the p+-emitter-Nwell junction and cause corresponding hole injection toward the
source electrode. If the Nwell isolation of the p+-emitter is high, then the avalanche–
injection in the parasitic n–p–n structure will occur, first providing a much larger
current to open the p+-emitter junction. Normally, for 0.5 μm CMOS processes, the
turn-on of the n–p–n structure is observed first, followed by p–n–p turn-on. When
both parasitic structures are in on-state, the critical condition for positive feedback
is defined by the (2.20) (Chapter 2).
This domino effect is typical for other SCR devices [7].
The final state of the device in the high injection mode is supported at very low
multiplication coefficients. Under these conditions, the avalanche current compo-
nent, generated in the high electric field, is completely replaced in the structure
by the injection of holes from p+-emitter region and electron injection from the
n+-emitter (source) region, generated in the low electric field.
The most valuable practical advantage of the device is the very high-current den-
sity and low clamping voltage level, in comparison with the snapback NMOS that
correspond to the holding voltage VH of the device. However, in the case of power
voltage VDD >VH , the possibility of transient latch-up should be accounted for, in
case of short term electrical overloads at the power supply pin.
Respectively, the FOX-SCR can be obtained using the same transformation
rules, by swapping the MOS part of the device with surface oxide isolation region
(Fig. 3.27d). Similarly to FOX, the FOX-SCR devices provide a high dc voltage
tolerance by elimination of the maximum rating limits for gate oxide reliability. The
breakdown voltage of the floating n+ drain to Pwell junction determines the turn-
on of the device. The same approach can be used for graded junctions to create
higher-voltage-tolerant devices, using Nwell grading of the blocking junction.
3.3.4 Low-Voltage Avalanche Diodes

Low-voltage avalanche diodes are rather useful components of ESD protection. If
the devices are not supported in the given CMOS process as standard devices, they
can be obtained freely and validated by the test data, taking into account process
variation across the wafer and from lot to lot.
At an appropriate doping level, the diode can be constructed as a self-aligned sur-
face device formed by n+-pldd, p+-nldd, or nldd-pldd implants (Fig. 3.28). The first
two diodes are practically self-aligned, since lightly doped drain (LDD) diffusion
can overlap with corresponding contact n+ or p+ diffusions.
10–2
10–3
EXPERIMENTAL
10–4
10–5
CURRENT (A)
10–6
SIMULATION (10x)
10–7
Cathode Anode 10–8
n+ pldd p+ 10–9
STI
10–10
Pwell 10–11
10–12
0 2 4 6 8 10
PULSED CLAMP VOLTAGE (V)
a) b)
Fig. 3.28 Simplified cross-sections for n+-pldd surface avalanche diode obtained in CMOS
process (a) and the experimental I–V characteristics (b)
A little technical difficulty in obtaining such a device is related to the CAD pack-
age automatic generation of LDD mask layers, and thus requires corresponding
manual editing of this layer.
A much bigger obstacle is availability of an appropriate implant dose in the pro-
cess that enables a truly low-leakage avalanche diode with a breakdown voltage
of ∼5–7 V. If the implant is too high, a tunneling component may significantly
impact the diode breakdown characteristics, providing a lower breakdown voltage
and higher leakage current at lower voltages. In an extreme case, a truly tunnel-
ing diode (or Zener diode [14]) might be formed, which can be inappropriate for the
desired voltage range. The experimental I–V characteristics of the successful surface
avalanche diode in 0.5 μm CMOS are presented in Fig. 3.28b.
As an alternative, in some cases the BiCMOS process may deliver an appropriate
low-voltage avalanche diode that is formed by base and emitter regions of the BJT
devices.
Building the non-self-aligned low-voltage avalanche diodes is usually impracti-
cal. Application of the low-voltage avalanche diodes in reference, stand-alone, or
second-stage components of the ESD clamps is discussed in Chapter 4.
3.3.4.1 Comparison of the Surface and Buried Avalanche Diodes

Unlike CMOS processes, the BiCMOS and BCD processes may deliver an opportu-
nity to build free low-voltage avalanche diodes using in-depth subcollector regions.
Comparison of the surface and buried avalanche diodes is presented below.
Surface avalanche diode Type-I (Fig. 3.29a) is formed by the n+-region on top
of the doping region DeepP. Another version of this diode, Type-II, is formed
by the same regions under the n+-region and connected by the Pwell region
(Fig. 3.29b). Due to having the same junction, both diodes deliver a similar break-
down voltage (Fig. 3.29c, d). However, the surface interaction effects provide higher
leakage and pre-breakdown current. According to TCAD analysis, the reason for
these effects is the breakdown close to the isolation region that results in involve-
ment of surface states and traps, as well as the reduced surface field (RESURF)
effect. In the Type-II avalanche diode, the breakdown is distributed more uni-
formly, due to the current flowing more in the vertical rather than the lateral
direction.
Comparison of the buried and surface avalanche diodes is presented in Fig. 3.30b.
The surface junctions provide a relatively low series resistance for the diodes, but the
impact ionization process is originated at the surface. This can potentially make the
breakdown voltage magnitude highly sensitive to the process-yield-related surface
effects. Similarly, surface diode Type-II has the lowest series resistance value; the
impact ionization process occurs at the edges of the N+ layer in the subsurface
region. The magnitude of the breakdown voltage depends on the proximity of the
breakdown region to the surface. A buried diode has an in-depth breakdown, but a
higher series resistance.
Cathode Anode Cathode Anode
n+
p+ n+
DeepP Pwell DeepP
Nepi Nepi
NBL NBL
PSUB PSUB
a) b)
10 –2 10 –2
EXPERIMENTAL
EXPERIMENTAL
10 –3 10 –3
10 –4 10 –4
SIMULATION (100x)
10 –5 10 –5
SIMULATION (10x)
CURRENT (A)
CURRENT (A)
10 –6 10 –6
10 –7 10 –7
10 –8 10 –8
10 –9 10 –9
10 –10 10 –10
10 –11 10 –11
10 –12 10 –12
0 2 4 6 8 10 0 2 4 6 8 10
PULSED CLAMP VOLTAGE (V) PULSED CLAMP VOLTAGE (V)
c) d)
Fig. 3.29 TCAD cross-section (a, b) experimental and simulated I–V characteristics (c, d) for the
Type-I and Type-II surface avalanche diodes
1.5
II
CURRENT (mA/μm)
e Type
1.0
ype
ce T
Surfac
fa
Sur
0.5
ner
d Ze
Burie
0.0
0 2 4 6 8 10 12 14
VOLTAGE (V)
Fig. 3.30 Comparison of buried diode experimental TLP characteristics with two types of surface
avalanche diodes
3.4 ESD Devices in BJT Processes

Apparently, ESD devices, discussed above, can be created in the CMOS part of
the BiCMOS process too. This section discusses ESD devices specific for the BJT
process.
3.4.1 Integrated NPN BJT Devices
Often, standard NPN BJT devices supported in the process can already provide
reversible operation in snapback mode. In the opposite case, the reversible operation
can be ensured by modification of the subcollector region parameters.
A typical NPN BJT cross-section in the 0.25 μm BiCMOS process is presented
in Fig. 3.31. The device combines the base-emitter region with a thin intrinsic
base and poly emitter, while the N-epi region and subcollector NBL and N-sinker
regions form the collector region. Namely the subcollector regions with appropriate
parameters, due to current saturation in these regions, are responsible for reversible
operation of the BJT device in the conductivity modulation mode.
a)
PULSED COLLECTOR CURRENT (A)
© 2007 IEEE
7 7
6 BASE–EMIT. BIAS (V) 6
5 5
4 4
3 1.7 3
2 2
1 1
0 10 20 30 40 10–11 10–10
COLLECTOR–EMITTER VOLTAGE (V) LEAKAGE (A)
b)
Fig. 3.31 Simplified cross-section for standard NPN BJT ESD device (a) and measured TLP
collector–emitter characteristics at constant base-emitter bias (b) [96]
3.4 ESD Devices in BJT Processes 115
Under normal operation conditions, the collector current in the N-epi region is
mainly vertical. In ESD conditions at high-current density, saturation in N-collector
is observed, followed by lateral current transport.
In these conditions, a part of the NBL and usually a highly diffused N-sinker
form an appropriate embedded ballasting region. The region locally limits current
density in the NPN device during avalanche–injection positive feedback.
In principle, the physical processes during conductivity modulation in n–p–n
structures which have already been discussed in Chapter 2 are directly applicable
to the discussion.
The output collector–emitter characteristics for NPN BJT at different conditions
in the base circuit are presented in Fig. 3.31b.
The turn-on voltage level for conductivity modulation mode of the first clamp
is based upon two physical effects. The effects are represented by either the
internal breakdown of the collector-base blocking junction or the dynamic effects.
The dynamic dV/dt effects is related to the displacement current effect in the
collector–base junction.
The collector–emitter 100 ns TLP characteristics have been measured according
to the methodology described in Section 3.2.2.
ESD protection window boundaries for standard BJT can be estimated from the
instability boundary (Fig. 3.31b), and corresponding parameters can be tuned for
the ESD BJT device.
The BJT clamp-level solution is discussed in Chapter 4. Adjusting BJT triggering
characteristics in a range that protects the standard device requires some alteration
of the structure. Essentially, these measures are related to the implementation of the
internal blocking junction, which has breakdown voltage lower than the collector–
emitter breakdown of the standard device.
One of the ways to achieve this is to reduce the collector–emitter spacing, and
thus create a different lateral junction profile using the tail of the N-sinker diffusion.
The disadvantage of this solution is that the N-sinker implant is not self-aligned
with base-emitter region, and significant misalignment can jeopardize the yield of
the analog circuit.
The isothermal current instability that results in negative differential resistance
in the collector–emitter I–V characteristics depends upon the product of the col-
lector thickness L and the collector doping level ND (Chapter 4). In the case of
the collector–emitter breakdown with the open base (BVCEO regime) and the base
current IB = 0 the collector current is simply equal to the emitter current and is
proportional to the value (1 − αM)−1 .
When the emitter and base terminals are connected to the ground (BVCES)
(UEB =0, Fig. 3.31a), the initial location of the avalanche breakdown is at the
collector–base junction. In this case, the avalanche breakdown is initially not
accompanied by injection from the unbiased emitter junction. Therefore, the collec-
tor current is provided by the multiplication of the thermal generation current IC0 ,
i.e., IC ∼ MIC0 . In this case, the breakdown voltage is close to the collector–base
breakdown voltage UCBBR .
With the increasing avalanche current, the generated holes flow out of the base.
In this negative base current regime, the hole current results in an additional voltage
drop across the p-base, increasing the base potential at the emitter junction by the
value rB IC , where rB is the internal base resistance. The value of rB can be extracted
from the simulation or from measured data. Thus, when rB IC is larger than or equal
to the potential of the emitter junction opening (approximately 0.7 V), injection
of electrons begins from the emitter and it is followed by the positive feedback
mechanism for n–p–n structures described in Chapter 2.
3.4.2 Bipolar SCR

In addition to the standard BJT clamps, bipolar SCR [7, 52] clamps can be cre-
ated. Simple device transformation rules require the formation of an additional
p+-emitter region connected to the collector. Since by its architecture, NPN has
an in-depth subcollector region, no additional regions are required to isolate the
p-emitter (Fig. 2.32a).
To enable low critical current, the p+-emitter is introduced between the base
region and collector contact region.
In principle, there is no architectural difference between the FOX-SCR and
BSCR. The different naming is due to historical reasons. However, the major differ-
ence between the two consists in that BSCR is designed such that it is essentially
composed of an original high-gain NPN device supported in process technology. In
this case, the device repeats the characteristics of the standard device until current
instability conditions are met.
The low-gain p–n–p structure embedded in BSCR by the introduced p+-emitter
is formed by the N-collector acting as a base and the P-base region of the NPNP
device acting as a collector.
The principles of ESD pulse operation for BSCR are similar to those of the FOX-
SCR, as discussed above. The BSCR device provides a low clamping voltage that is
typical for SCR, with triggering characteristics that are easily controlled by the base
current (Fig. 2.32b).
Due to high internal gain, the BSCR device can produce a rather efficient dV/dt
clamp that is triggered by the displacement current during rise time of the ESD
pulse.
For both NPN and BSCR, an alternative version of the self-aligned ESD clamp
with an internal avalanche diode can be obtained by stretching the PBASE diffusion
to create an overlap with the collector region. However, this solution requires a
significant deviation from the poly emitter NPN device architecture. Therefore, it
is not possible in self-aligned BJT processes and Si–Ge processes with an epitaxial
base.
A surface junction between the base and the N-sinker regions can be used to
change the breakdown voltage of the device to adjust an ESD protection window
targeting. Similarly, the internal avalanche diode structure can be implemented in the
3.5 High-Voltage ESD Devices in BCD and Extended Voltage CMOS Processes 117
© 2007 IEEE a) © 2007 IEEE c)

5 6
3 3 ESD CLAMP DESIGN
BASE-EMIT. BIAS (V)
ANODE (COLLECTOR)
ANODE (COLLECTOR)
4 5
CURRENT (A)
CURRENT (A)
2 2 4
3
2Fx50x3.5um
3
2
1 1 2
BVCER
1 1
0 0
0 10 20 30 40 –12 –11 –10 0 5 10 15 20 25 30 –12 –11 –10
10 10 10 10 10 10
COLLECTOR-EMITTER LEAKAGE (A) b) ANODE (COLL.)-EMITTER LEAKAGE (A) d)
VOLTAGE (V) VOLTAGE (V)
Fig. 3.32 Simplified cross-section (a) and TLP characteristics (b) of the original BSCR device
at different base-emitter bias conditions and BSCRZ device with internal base-emitter avalanche
junction (c) and its TLP characteristics compared to BSCR BVCER (d) [96]
BSCR, where an additional floating N-EMIT region is added between the PBASE
and the P-EMITTER (Fig. 3.32c, d).
BSCR devices are usually designed as free devices based upon high-gain NPN,
the current gain α NPN >>α PNP . The current instability criteria can be obtained, taking
into account the additional gain from the built-in low-gain PNP α PNP device. In this
case, the current instability criteria with the grounded base is presented by α NPN
MN +α PNP MP >1. For the case of the BSCR, this term can be simplified to α NPN MN
>1 due to a very low value of α PNP .
Thus, in a properly designed BCSR, the instability boundary can be controlled
similarly to the NPN BJT, making the design predictable through known yield
characteristics for supported devices.
Depending on the design parameters, BSCR snapback is induced by avalanche–
injection conductivity modulation, followed by the change of the modulation
mechanism to double injection at high injection levels, which eliminates avalanche
multiplication MN , MP ∼1 and thus provides for the condition α NPN +α PNP >1.
3.5 High-Voltage ESD Devices in BCD and Extended Voltage

CMOS Processes
ESD protection of fast high-voltage pins in analog applications is a complex
problem, due to the need to target several performance specifications. These spec-
ifications generally include appropriate voltage tolerance for the protected pins,
precise turn-on voltage control below the pulsed absolute maximum voltage of the
protected components during an ESD event, high-current operation for desired ESD
pulse specifications, and low sensitivity to the pulse rise time.
It is also critical to maintain protection against ESD events during system-level

operation. In this context, the most reliable way to guarantee system stability against
transient latch-up is to require a minimum on-state holding voltage above the power
supply voltage level. This requirement also maintains system stability against short-
term electrical overstress (EOS) events above absolute maximum voltage limits.
Finally, since obtaining a small footprint device is a cost-driven requirement, it is
obligatory for operation in the conductivity modulation (snapback) mode to achieve
high current density.
High-voltage tolerant ESD protection may have two major specification require-
ments. In the case of high-voltage control pins or slow power pins, snapback devices
with a holding voltage below power supply level can be used. However, in the
case of high-voltage fast transient pins, hot plug-in or other system-level require-
ments, the expectation from a high-voltage ESD device is to provide high-current
characteristics with the holding voltage above the power supply level.
In practical application, high-voltage ESD protection can be acomplished using
high-voltage NPN devices based upon avalanche–injection conductivity modula-
tion, SCR-type devices that provide high current with low holding voltage, and
lateral PNP devices that operate in avalanche–injection mode.
Attempts to attain reversible snapback operation in high-voltage NDeMOS,
NLDMOS, or VDMOS devices supported in the process are usually unsuccessful
due to the irreversible nature of the snapback operation in such integrated com-
ponents. The major reason for this is inappropriate current ballasting above the
relatively low saturation current density of the high-voltage drift regions. Attempts
to add an additional n+ ballasting region are also unsuccessful, due to deep current
transport in the relatively long channel devices. In this case, current limiting is inef-
ficient, since the current in the contact region is drawn from a much larger area,
compared to the distance between the drain and source contact regions.
Similarly, active clamps with monopolar current conduction are inefficient due
to the relatively high on-state resistance of high-voltage MOS devices. Therefore,
snapback solutions are the solutions of choice for local protection of the high-
voltage pins of analog circuits.
3.5.1 LDMOS-SCR and DeMOS-SCR Devices

As mentioned above, the most typical small footprint clamps are based on ESD pro-
tection devices operating in snapback mode. Either NPN BJT or silicon-controlled
rectifier (SCR) ESD structures are generally used as high-current clamp com-
ponents. Due to the nature of avalanche–injection conductivity modulation, both
types of structures provide low holding voltage when implemented in high-voltage
processes.
A typical cross-section of the DeMOS-SCR device implemented in a 0.5 μm
extended voltage process is presented in Fig. 3.33a with the experimental TLP
characteristics (Fig. 3.33b).
LSWS LG LP LPD
Bulk Source Gate P-Emitter Drain
p+ STI n+ STI p+ STI n+

n-p-n X
R
PW
RNW
p-n-p
Pwell Nwell
P-sub
a)
b)
Fig. 3.33 Simplified cross-section (a) and the experimental TLP I–V characteristics (b) of typical
DeMOS-SCR for 40 V ESD protection. [53]
The device combines a standard device supported by the process with parasitic
n–p–n structure and an introduced low-gain parasitic p–n–p structure. The embed-
ded p–n–p structure has a p+-emitter region with Nwell-region acting as a base and
a Pwell p+ contact acting as a collector.
The principle of device operation is similar to the previously discussed LVTSCR
device, with a corresponding change to the high-voltage blocking junction. Before
snapback, the avalanche breakdown corresponds to the same operation mode as in
the original LDMOS device. This is guaranteed, since the left part of the struc-
ture practically replicates the original device. After achieving avalanche–injection
conditions, electron injection is initiated from the source junction in the parasitic
n–p–n structure, with the n-collector formed by the Nwell drift region. At high volt-
age, the drift region is already almost fully depleted, according to the principle of
LDMOS operation. Therefore, the drift current along the extended drain region cre-
ates an opening potential for the p+-emitter–Ndrift junction. As a result, at some
critical current density, conditions for double injection conductivity modulation
are created.
Similarly to LVTSCR, in order to adjust device turn-on, either dynamic coupling

of the gate electrode or reduced drain-source (n-base to p-collector) spacing can be
used in the real ESD clamp, which is discussed in the following chapter.
Control of the triggering and breakdown voltages can be achieved by chang-
ing the drift region length by the RESURF approach, thus shortening the space
charge region area, for example, by using a drift region-poly overlap with a poly
gate extension.
The floating drain region can be added depending on the depletion of the
extended drain (drift) region the device. This can be done similar to NLDMOS-
SCR (Fig. 3.34). The region might be required to avoid premature opening of the
P-emitter junction, due to a small voltage drop across the drift region. Whether this
is necessary or not depends on the doping profiles.
Bulk Source Gate P-Emitter Drain

p+ STI n+ STI n+ STI p+ STI n+
Pbody
Pwell Ndrift
NBL
P-sub
Fig. 3.34 Simplified cross-sections for high-voltage NLDMOS-SCR
Triggering current characteristics of the device depend on the p+-emitter isola-

tion level. This can be controlled by the parameters of the floating n+- drain region
that is soft-connected to the drain contact region either by the NDRIFT (Nwell), or
by the n+-diffusion implemented by means of appropriate topology.
In the case of the BCD process, the device structure is practically similar to
above with the exception of the N-epi and NBL (Fig. 3.34), as well as the Ndrift
and Pbody, which play the same role as Nwell and Pwell in the NDeMOS-SCR. In
principle, the device provides similar ESD characteristics and snapback operation
with low holding voltage ∼2–5 V. However, specific ESD clamp design is different
in correlation with the differences in DeMOS and LDMOS layout (Chapter 4).
Applying similar transformation rules as in the case of a FOX-SCR device, a
high-voltage SCR can be obtained by swapping the polygate region with the avail-
able type of surface oxide isolation region (Fig. 3.35). The device eliminates the
MOS structure, but the poly region over the field oxide can still be used to create a
field electrode.
Versions of PLDMOS-SCR implementation using PLDMOS as a starting struc-
ture are presented in Fig. 3.36. In this structure, an n+-emitter is added to create the
parasitic components of the SCR device, while the blocking junction is supplied by
the PLDMOS device architecture.
Bulk Source Field Electrode P-Emitter Drain
p+ STI n+ STI n+ STI p+ STI n+

Pbody
Pwell Ndrift
NBL
P-sub
Fig. 3.35 Simplified cross-sections for high-voltage SCR
Bulk Source Gate N-Emitter Drain
n+ STI p+ STI n+ STI p+
Nepi and Nwell Pdrift
NBL
P-sub
Fig. 3.36 Simplified cross-sections for PLDMOS-SCR devices
The common feature of all the high-voltage SCR devices (Figs. 3.33, 3.34, 3.35,
and 3.36) is the embedded p–n–p and n–p–n internal structures which enable double
injection conductivity modulation. At the same time, the dc voltage tolerance of the
device is supported by the original standard device design that provides both the
blocking junction and the induced channel current in case of gate electrode coupling
or external gate control circuit. The topological control of the spacing of the device
regions provides the control of the breakdown voltage, the triggering voltage, the
triggering current, and even the holding voltage parameters of the composed ESD
device.
3.5.2 Lateral PNP BJT Devices

If the space-on-chip budget is not critical, high-voltage protection can be achieved
using large-size stacked clamps or by a high-impedance high-voltage avalanche
diode.
However, if a small footprint is required by the design, perhaps the only universal
practical alternative to low holding voltage SCR solutions is the lateral PNP device.
The first successful PNP ESD device implementation has been reported in [54–56].
In these studies, it has been demonstrated that a PNP-based protection structure

can be greatly improved to provide robust ESD protection for high-voltage I/Os in
advanced smart power technology. By using specific design guidelines and coupling
an open-base lateral PNP with a vertical avalanche diode, a low on-resistance (∼1
) can be obtained using a limited silicon area. This optimized structure was used
to protect specific high-voltage pins of a power-over-ethernet chip.
Later, implementation of the lateral PNP ESD device both in drain-extended volt-
age CMOS and in BCD processes with a supported lateral DMOS device has been
demonstrated [53].
In [57], the physical mechanism that underlies the experimentally observed con-
ductivity modulation in the lateral PNP has been studied using physical process and
device simulation.
The structures and transmission line pulse (TLP) characteristics of the lat-
eral PNP (Fig. 3.38a) can be compared to the DeMOS-SCR device (Fig. 3.33b)
implemented in the same 0.5 μm 40 V drain extended CMOS process.
Architecture of the 40 V PNP device (Fig. 3.37) is based on the DeMOS
device and can be obtained by transformation of the DeMOS-SCR, which involves
elimination of the former n+-region and replacement of the polygate region and
the n+-region with field oxide or shallow trench isolation. The extended drain
diffusion region and Pwell form the blocking base–collector junction. The field
electrode over the blocking junction (FE) (Fig. 3.37) is optional, but may be
very important for the elimination of the undesirable breakdown voltage walkout
effect [40].
This transformation removes the original parasitic NPN structure from the
device, since the n+-source acting as an n+- emitter is removed. Therefore, con-
ductivity modulation is now based upon avalanche–injection in the p–n–p structure.
This structure provides a different current balance and the resulting high holding
voltage (see Chapter 2).
P-Collector FE P-Emitter Drain
STI p+ STI p+ ST I n+
STI STI
Pbody (Pwell) Ndrift (Nwell)
NBL
P-sub
Fig. 3.37 Simplified cross-section of lateral PNP

a) b)
Fig. 3.38 The experimental TLP I–V characteristics for lateral PNP (device implemented in 40 V
DeMOS process (a) [53] and for 100 V BCD process (b)
In comparison to the DeMOS-SCR, this PNP structure presents a much closer

approximation of the desired clamping characteristics with a high holding volt-
age (Fig. 3.38a). The advantage of this device is the absence of the deep S-shaped
dependence with the 3–5 V holding voltage of the DeMOS-SCR (Fig. 3.33b).
At the same time, due to the change in the majority carrier, the on-state resistance
of the device is relatively high, and thus some negative differential resistance can
provide for a better approximation of the desired ideal case.
The focus of [53] is on physical explanation of the experimentally observed
phenomena for NDR in PNP, using Technology CAD analysis with calibrated phys-
ical process and device simulation tools. It has been demonstrated that specifically
modulation of the extended N-drift (or Nwell) region is responsible for negative
differential resistance of the lateral PNP devices, rather than the modulation of the
Pwell region. This fact is also confirmed by the absence of the NDR in vertical PNP
with a thin N-base region and corresponding long P-collector region.
Standard PNP devices show significantly different behavior in breakdown and
conductivity modulation modes relative to their lateral counterparts [53]. Simulated
3.39a, b for the example of a 0.5 μm 20 V BiCMOS process. The I–V charac-
teristics (Fig. 3.39c) of these devices are compared to simulated collector–emitter
characteristics of a standard vertical NPN with an N-buried layer and an N-sinker
subcollector.
The avalanche–injection conductivity modulation of the extended N-base region
can be understood via comparative analysis of the 2D profiles and cut-line depen-
dencies of the electric field magnitude at points α (switching voltage) and β (holding
voltage) (Fig. 3.39). Electric field reduction is seen in the base (negative differential
conductance) at holding voltage β in Fig. 3.39c. In the state at α, the electron den-
sity greatly exceeds the hole density in the base, while at β, the base is filled by a
C E B E B C
pSink
NBL PBL
(a) (b)
1.0E-01
β
1.0E-02
Collector Current [Amps/µm]
1.0E-03 α
1.0E-04
1.0E-05 vnpn
vpnp
1.0E-06 lpnp_notbutted
1.0E-07
1.0E-08
1.0E-09
1.0E-10
1.0E-11
1.0E-12
1.0E-13
1.0E-14
1.0E-15
0 10 20 30 40 50
Collector Voltage [V]
c)
Fig. 3.39 Process simulated cross-sections for lateral (a) and standard vertical (b) PNP devices
and their simulated collector–emitter I–V characteristics, in comparison with a standard vertical
NPN BJT device in a common emitter circuit at zero base-emitter bias
quasi-neutral electron–hole plasma and the concentrations of the two types of carri-
ers are approximately equal (Fig. 3.40d). This space charge compensation results in
reduction of the electric field in the base (Fig. 3.40a) in the state at β.
The difference between electron and hole mobilities leads to significant differ-
ence in the quasi-neutral area formation of NPN and PNP devices, and the resulting
decrease in the collector-emitter voltage.
A simple numerical simulation experiment can be used as a straightforward
illustration of the differences in NPN and PNP behavior. The cross-section of a
standard PNP is compared to the cross-section of an NPN, which is artificially cre-
ated by swapping the n and p-doping of the PNP. Thus, an “antipode” NPN device is
obtained from the PNP “ancestor.” Another physical numerical experiment involves
swapping all of the mobility coefficients between the electron and hole carriers in the
C E B C E B
1.0E+06
3.5
STI STI STI STI

2.0*105
1.0E+05
2.0*105
6.0*105 6.0*104 1.0E+04

1.5*105
1.5*104
α β
1.0E+03
α, y = –7.5
Electric Field [V/cm]

1.0E+02
β, y = –7.5
a) α, y = –7.7
β, y = –7.7
2.5*1 1.0E+01
0 1 2 3 4 5 b)
3.0*
Distance [µm]
2.0*
2.5*
3.0* 4.0* 1.0E+20
α 4.0*1 α 1.0E+19
1.0E+18
1.0E+17
1.0E+16
1.0E+15
8.3*1 8.3*1 1.0E+14
1.0E+13
5.5*1 1.0E+12 α, Electrons
2.5*1 5.5*1 1.0E+11 α, Holes
2.5*1 1.0E+10
β, Electrons
3.6*1 β β, Holes
1.0E+09
High-Voltage ESD Devices in BCD and Extended Voltage CMOS Processes
β Holes d)
3.6*1 1.0E+08
Electron, Hole Concentration [cm–3]
c) 0 1 2 3 4 5
Distance [µm]
Fig. 3.40 Comparison of the 2D electric field magnitude profiles (a) with the corresponding dependencies at cut-lines (b) (at y= –7.7) and electron and hole
density (c) with cut-lines (d) (at y= –7.7) for the states α (switching voltage) and β (holding voltage) indicated on the collector–emitter I–V characteristic in
125
Fig. 3.39c
simulation. Both the device with the altered species and the device with the swapped
mobilities demonstrate the same phenomenon – quasi-neutral region formation that
provides a low holding voltage.
In [53], a mathematical model of the conductivity modulation in the high-voltage
lateral PNP was proposed based on an equivalent circuit (Fig. 3.41), where an
additional parasitic NPN is included in the PNP base–emitter circuit. This NPN is
formed by the Ndrift region acting as a collector, P+-emitter as a base, and N+-base
as an emitter, including corresponding shared regions.
P-EMITTER
npn
rD
pnp
P-COLLEC.
Fig. 3.41 Equivalent circuit of the high-voltage lateral PNP with the parasitic NPN device formed
by PNP regions acting as Ndrift as collector, P+-emitter as P-base, and N+-base as N-emitter
3.5.3 High-Voltage Avalanche Diodes

High-voltage avalanche diodes are important reference components for dV/dt effect-
independent high-voltage clamps.
Due to the fast avalanche process as compared to ESD pulse domain, the
avalanche current can be used to provide bias or current for the control electrodes
of the devices. Respectively, ESD devices can be designed with rather low cou-
pling of the control electrode and injection junctions. The first can be achieved
using resistors of smaller value in the common emitter circuit. The lower displace-
ment current effect can be achieved by greater isolation of the injection regions.
The greater isolation might involve use of a heavily doped isolation region and
appropriate region dimensions.
As a result of such design, the high-voltage clamps with avalanche diode refer-
ence can provide dV/dt-independent turn-on, combined with high-current operation.
Similar to low-voltage design, depending on the options delivered by the pro-
cess, either lateral or vertical high-voltage avalanche diodes can be implemented.
In the case of the vertical diode, adjustment of the avalanche breakdown voltage is
practically absent and the design mainly relies on what is already available in the
process.
The lateral design presents a much larger degree of freedom in control of the
breakdown voltage, by a simple mask change of the lateral dimensions of the diode.
The disadvantage of lateral design is the non-self-aligned architecture.
3.6 Dual Direction Devices 127
Anode Cathode
STI p+ STI n+ STI
Pbody
RP RN Ndrift
Pwell
PRESURF Nepi
NBL
Psub
Fig. 3.42 Cross-sections for lateral high-voltage avalanche diodes in BCD process
A typical implementation of the avalanche diodes in BCD process reuses the

regions used in NLDMOS devices (Fig. 3.42). The lateral high-voltage blocking
junction is formed by the Pbody implant and Ndrift region. Since the Pbody implant
is aligned to the STI position, the structure’s sensitivity to misalignment is relatively
low. The breakdown voltage of the diode is controlled by the Ndrift region length
and optional PRESURF implant. The last is deeper than the N-drift region and cre-
ates a filed-plate effect, thus controlling the breakdown voltage of the device. The
avalanche current generated by the device is limited by the saturation resistances of
the Ndrift and Pbody regions RN and RP .
3.6 Dual Direction Devices

A small footprint and dual-direction functionality is required for a number of ana-
log applications, for example, column drivers in liquid crystal displays, RF inputs,
common mode voltage regulators, interface applications, and digital–analog con-
verters. Design of dual-direction ESD protection is one of the most sophisticated
device-level ESD problems, especially if system-level ESD protection requirements
are also specified.
A typical requirement states that the protected dual-direction pin should be tol-
erant to a certain positive and negative voltage range relative to the grounded epi
p+-substrate. Respectively, the ESD device is connected between the dual-direction
pad and the ground node and must provide a proper waveform under the ESD pulse
discharge.
Consequently, the device must include a corresponding double set of contact
diffusion regions to realize a conductivity modulation mechanism in both directions,
while at the same time, the regions should be separated by two blocking junctions.
3.6.1 Dual-Direction Device Architecture in CMOS Process

An absolutely necessary requirement for realizing a dual-direction ESD protection
device in the CMOS process is the availability of a Deep Nwell or similar deep
N-isolation layer in the process, which will vertically isolate the Pwell region. The
exception to this rule is, of course, an SOI process case.
In [49], test structures have been fabricated using a non-silicided 0.5 μm 5 V
CMOS process with a deep N-well, on a p+-substrate with a ∼4 μm thick P-epi
layer.
There are a number of options for arrangement of the contact diffusion and at
least two ways to engineer the blocking junction: symmetrically (Fig. 3.43a) and
asymmetrically (Fig. 3.43b).
a)
b)
Fig. 3.43 Typical dual-direction SCR structure; PN-n-NP structure with the n+-Pwell (a), the
compact asymmetrical version of the PN-n-PN device (b)
In the symmetrical case, the Nwell (NW) and deep Nwell regions (DNW)
(Fig. 3.43a) can fully laterally and vertically isolate all Pwell regions (RW). This
design minimizes the substrate effect described below.
In the asymmetrical case, only the pad side Pwell remains fully isolated
(Fig. 3.43b). This solution is more compact; however, the substrate effect might
significantly impact the operation of the device. The injection contact diffusions are
connected to GND and PAD device electrodes.
Another feature generally implemented to reduce the voltage tolerance in the
device is the middle n+ diffusion (Fig. 3.43). Since for a typical 5 V CMOS pro-
cess, the P-well to N-well breakdown voltage is usually rather high (∼18–25 V),
the additional floating diffusion region is required to bring the blocking junction
breakdown and corresponding triggering voltage closer to the 5 V core devices or
the 10 V stacked device, i.e., to the ∼10–14 V voltage range.
PAD
pnp1
npn1
GND PAD
pnp2
pnp1
p+ n+
p+
p+ n+ p+
p+
p+ n+
NW
PW
npn2
npn1
npn2
pnp2 GND
a) b)
Fig. 3.44 Typical compact dual-direction SCR structure with the n+- Pwell and p+-Nwell
blocking junctions (asymmetric PN-n-PN structure) (a) and its equivalent circuit (b) [36]
The device combines embedded parasitic n–p–n and p–n–p structures that can be
identified in the device cross-section and represented by the equivalent circuit. This
is presented in Fig. 3.44, for the case of asymmetrical device structure.
However, alternatively for each current direction, the injection regions form
corresponding SCR-type internal p–n–p–n structures that enable double injection
conductivity modulation in the device.
In [36], it has been shown that manipulation of the sequence of the injection
regions or creation of a more complicated topology produced an important benefit
of overcoming the undesired grounded substrate effect.
The problem consists in the embedded connection to the p-substrate by the guard
ring necessary to provide latch-up isolation for the dual-direction ESD cell. As a
result, the substrate acts as an additional p+- emitter in case of negative current
through the pad, while in case of positive current, additional negative feedback is
achieved.
3.6.1.1 Device-Level Positive and Negative Feedback

Thus, even symmetrical device (Fig. 3.43a) characteristics are significantly differ-
ent, depending on the conditions of the floating or grounded p+- guard ring. In the
case of a floating substrate connection, the device characteristics are symmetrical
and provide a low holding voltage typical of SCR structures, while in the case of
a grounded substrate, a significant asymmetry occurs (Fig.3.45). In the case of a
grounded substrate, the device provides lower resistance in the negative direction,
while in the positive direction, the holding voltage is rather high and the maximum
current is reduced according to leakage measurement data.
This effect has been explained in [36] through the positive and negative feedback
realized in the structure of the device. The asymmetry in the device operation can be
4 4
SUBSTRATE CONNECTION
3
2 2
TLP CURRENT (A)

1
0 0
–1
–2 –2
–3
–4 –4
–15 –10 –5 0 5 10 15 10–12 10–11 10–10 10–9 10–8
VOLTAGE (V) LEAK.CURRENT (V)
Fig. 3.45 TLP I–V characteristics of symmetric PN-n-NP structures for different substrate
connection conditions
understood on the phenomenological level as a result of the engaging of a parasitic

PNP. This PNP is formed by the p-substrate acting as a collector, deep N-well as a
base, and the upper Pwell acting as an emitter.
After the device turns on into the conductivity modulation mode, in the case of
positive current direction, the holes injected from the pad junction partially escape
through the parasitic PNP current path into the P+-substrate. As a result of this
effect, the level of the space charge neutralization between the electrons injected
from the ground junction is reduced. This negative feedback results in the formation
of a corresponding electric field distribution between the cathode and the anode of
the DD-SCR structure and leads to rather high values of the holding voltage and the
on-state resistance.
An opposite situation takes place in the case of the negative current direction.
In this case, the p-substrate may essentially contribute to the hole injection level by
acting as an additional P-emitter.
A similar effect can be observed with increase of the size of the symmetrical
device. The effective collector resistance of the PNP increases as well as the negative
feedback level is reduced, thus providing lower holding voltages for larger devices.
In order to compensate the device-level negative feedback provided by the par-
asitic PNP in the substrate, the diffusion sequence has been altered toward the
PN-n-PN structure (Fig. 3.46a), followed by a reduction in the contact-to-contact
spacing. With the positive current direction, the injection junctions are closer to
each other.
a)
2.5
CURRENT (A)
2.0
1.5
1.0
0.5
0.0
–18 –16–14–12–10 –8 –6 –4 –2 0 2 4 6 8 10 12 14 16 18
–0.5 VOLTAGE (V)
–1.0
–1.5
–2.0
b)
–2.5
Fig. 3.46 Results of positive feedback implementation (a); TLP I–V characteristics of PN-n-NP
structure (b)
This additional positive feedback mechanism compensates the negative feedback

due to the parasitic PNP. At the same time, this measure does not impact the charac-
teristics of the device in the negative current (Fig. 3.46b). For the case of NP-n-PN
DIAC, both asymmetrical and symmetrical architecture types of the device perform
similarly (Fig. 3.46b).
Reduction of the contact-to-contact spacing results in a further adjustment of the
holding voltage within the desired range (Fig. 3.47).
Thus, it becomes very important to utilize knowledge about physical processes
under conductivity modulation in ESD design, toward implementation of desired
high-current characteristics of the device and addressing both the desired voltage
waveforms and the potential dynamic latch-up issues.
3.6.2 High-Voltage Dual-Direction Devices
High-voltage versions of dual-direction devices are extremely sensitive to the

positive and negative feedback effects described above.
5
CONT-CONT SPACING
4
3
TLP CURRENT (A)

2
1
0
–1
–2
–3
–4
–5
–20 –15 –10 –5 0 5 10 15 20
VOLTAGE (V)
Fig. 3.47 Results of positive feedback implementation; TLP I–V characteristics of PN-n-NP struc-
ture created by scaling the device characteristics and the reduction of the contact-to-contact space
Usually, the challenging task is to implement both the high breakdown voltage
and the high holding voltage when using these devices.
Substantial progress in this case can be achieved by a topological solution that is
based upon formation of interdigitated contact regions. This approach is similar to
the two-stage SCR for hot plug-in application described above.
The general architecture presents itself as a dual-directional SCR (DIAC type)
that uses regions typical for BiCMOS process. The blocking junction of the device
is formed by a Pwell-Nwell-Pwell sequence with a floating Nwell, or a PCOLL-
Nwell-PCOLL sequence (Fig. 3.48).
Fig. 3.48 Cross-section for the DIAC
There are several design variations of the interdigitated device: (n)P–N–P(n),

where small n-regions are distributed mainly along the p+-diffusion, and (p)N–P–
N(p), where small n-regions distributed mainly along the n+-diffusion (Fig. 3.49).
For the given device spacing, the holding voltage is the function of the balance
of the diffusion regions for (n)P–N–P(n). An increase of the holding voltage from
p+
n+
Fig. 3.49 Topology of the DIAC structures with interdigitated diffusion
the original holding voltage of ∼5–7 V, up to 60–80 V (Fig. 3.50) can be achieved
by using interdigitated regions with different area ratio.
However, a major side effect of the increase of the holding voltage is the reduc-
tion of the maximum holding current (IT2 ) from IT2 =10–12 A (for VH = 7 V) down
to IT2 =4–6 A, which approximately follows the power law.
5
Pulsed Current (A)
4
B C
A
3
0
0 10 20 30 40 50 60 70 80 90 100 110
Pulsed Voltage (V)
Fig. 3.50 Control of the device characteristics using n+-p+ ration and corresponding I–V char-
acteristics. Device “A” – no interdigitation, solid n+ and p+ diffusions; device “B” interdigitated
device with 20% of p+ area and interdigitated device “C” with 10% of p+ area
Fig. 3.51 TLP I–V characteristics of the dual-direction device for positive (a) and negative (b) pad
current
The final level of complexity is related to the implementation of the high holding
voltage characteristics for both the positive and negative currents. An example of
such successful implementation is presented in Fig. 3.51. An appropriate combi-
nation of the subcollector and injection regions provides an adequate DC voltage
tolerance and holding voltage for both current directions at the device PAD electrode
relative to GND electrode (Fig. 3.51).
3.6.3 Dual Direction ESD Devices Based upon Si–Ge NPN BJT
Structure
Methodology implementing the positive and negative feedback compensation in

dual-direction device structures, described above, has limited application in CMOS
or BiCMOS processes below 0.5 μm. In this case, increased well doping levels sig-
nificantly reduce the gain of parasitic n–p–n and p–n–p structures. This reduction
results in an inadequate CMOS-based DIAC type structure.
An example of the best DIAC structure characteristics in the 0.13 μm process
is presented in Fig. 3.52. The device is implemented using the minimum spacing
dimension allowed by the process rules for 0.13 μm high-frequency analog Si–Ge
BiCMOS process. In comparison with the TLP characteristics of the DIAC structure
implemented in a non-silicided 0.5 μm 5 V CMOS process, target performance is
not reached.
In the case of the 0.13 μm process, the DIAC structures provide neither the dual-
direction ESD protection nor the desired high-current tolerance.
One of the possible solutions in this case can be realized on the clamp level using
isolated snapback NMOS devices. This is usually possible, since most of the scaled
CMOS processes typically have an available Deep Nwell layer.
–40 –30 –20 –10 0 10 20 1E-11 1E-10 1E-9

3 3
2 2 2
1 1
TLP Current (A)
TLP Current (A)

0 0 0
–1 –1
–2 –2 –2
–3 –3
–40 –30 –20 –10 0 10 20 1E-11 1E-10 1E-9
TLP Voltage (Volts) Leakage Current (A)
Fig. 3.52 Comparison of TLP current–voltage characteristics of the asymmetric DIAC structures
(PN-n-PN) for the 0.5 μm 5 V CMOS process (square symbols) and 0.13 μm high-frequency
analog BiCMOS process (round symbols)
A more aggressive alternative for small footprint solutions can be implemented

in the case of BiCMOS processes, where NPN devices with a polyemitter can form
a different type of the dual-direction device, as discussed below.
Similarly, at the clamp level, the dual-direction clamp can be composed of NPN
BJT devices available in the process, by back-to-back stacking of the n–p–n devices
(Fig. 3.53). The simple use of back-to-back isolated and stacked snapback NPN,
NMOS, or BSCR devices is a straightforward approach to solve the problem of dual-
direction ESD protection in standard CMOS and BiCMOS processes. This approach
has the advantage of utilizing standard snapback devices and permits the option of
using control electrode coupling to achieve a low triggering voltage. Nonetheless,
stacking of isolated unidirectional clamps consumes an excessive space and a better
solution is possible.
A more elegant device-level solution in proposed below. In order to provide
scalable ESD protection, new device architecture based on a novel merged collec-
tor dual-direction BJT-based (DD-BJT) ESD clamp is proposed and successfully
implemented in a 0.13 μm Si–Ge BiCMOS process.
A novel system-level solution is obtained by a merged collector DD-BJT clamp
based upon SiGe NPN BJT process technology. The device architecture is based on
stacked NPN BJT clamp solutions.
To achieve dual-direction ESD protection, two unidirectional NPN clamps can
be placed back-to-back to produce a stacked clamp, as shown in Fig. 3.53b. The
key to implementing this concept in a small footprint design is to partially elim-
inate and merge the subcollector regions of the stacked components. This gives
PAD
PAD
GND
GND
a) b)
Base Emitter Collector
STI
DT
NPN
c)
Fig. 3.53 Standard uni-direction NPN-based ESD clamp (a), dual-direction clamp obtained by
back-to-back stacking of two uni-directional clamps (b) and cross-section of Si–Ge BJT with 0.24
μm emitter (c)
rise to a novel DD-BJT device with a cross-section and equivalent circuit shown
in Fig. 3.54. At the device level, this new device is created by packing the base–
emitter regions of the NPN BJT devices into one cell and removing all individual
collector regions (Fig. 3.54a). This shared subcollector region approach allows for
a maximum packing density for the active region of the device.
The merged collector DD-BJT clamp device has four alternating groups of fin-
gers, represented by four circuit components in the equivalent circuit shown in
Fig. 3.54c. The base–collector junction of an NPN BJT (base–emitter junction
shorted) forms the reverse diode. The principle of device operation is based on
the physics of the BJT NPN snapback clamp. When stressed by a positive ESD
event, BJT2 remains passive and the current path is through BJT3 and the for-
ward biased diode is formed by the base–emitter shorted BJT1. This current path
is indicated in Fig. 3.54c with arrows. In the case of a negative ESD event, BJT1
is passive and the ESD current path is through BJT2 and BJT4 as depicted by the
arrows.
This proposed device architecture allows for variation of both the number of
fingers in each group and their arrangement. Two basic arrangements are shown in
PAD
GND
B E B E E E
B B
STI STI STI BJT3 STI BJT1 STI
BJT4 BJT2
N-epi a)
N+ Buried Layer
p-substrate
PAD GND
E E
B E B B E
B
STI STI STI BJT3 STI BJT4 STI
BJT1 BJT2
N-epi
b)
N+ Buried Layer
p-substrate
BJT2 BJT1
Merged
Collector
BJT3 BJT4
GND c)
Fig. 3.54 Cross-sections (a, b) and equivalent circuit (c) of the merged collector DD-BJT structure
Figs. 3.54a and 3.54b. In the first case, the ESD current flows between adjacent BJT
fingers; while in Fig. 3.54b, the ESD current flows between fingers separated by
another BJT finger.
3.6.3.1 Experimental Characteristics of the DD-BJT Clamp

Dual-direction protection capabilities of the proposed merged collector DD-BJT
ESD clamp (Fig. 3.54) have been demonstrated by standard TLP measurements. The
cell under characterization had a total area of 70 μm × 70 μm. The experimental
structures shown in Fig. 3.54a, b had 24 fingers (18 fingers with bases connected to
–4 –2 0 2 4
4 4
2 2
TLP Current (A)
0 0
–2 –2
–4 –4
–4 –2 0 2 4
TLP Voltage (Volts)
Fig. 3.55 Measured pulsed I–V (TLP) curve for device architectures shown in Fig. 3.54a (square
symbols) and 3.54b (round symbols)
ground through a resistor and 6 fingers with a shorted base-emitter junction). Half
of these fingers are connected to the pad and the other half to the ground, forming
a fully symmetric dual-direction structure. Pulsed I–V characteristics for this device
are shown in Fig. 3.55.
Both positive and negative branches of the TLP I–V curve of the merged
collector integrated dual-direction clamp are similar to those of a unidirectional
snapback NPN clamp. The triggering voltage of the dual-direction clamp is
about 0.7 V higher than that of the unidirectional, due to the forward-biased
diode in series with the BJT NPN in both directions. The relatively low trig-
gering voltage is an advantage of the proposed clamp over DIAC structures,
which typically have relatively a high holding voltage and a large dV/dt effect on
triggering.
The holding voltage of the dual-direction clamp is lower than that of unidirec-
tional clamp, due to the double injection (SCR effect) between the non-isolated
base-emitter short and snapback NPN fingers. The holding voltage in the Fig. 3.54b
configuration is higher than in that of Fig. 3.54a, because the SCR effect in the
Fig. 3.54b configuration is partially suppressed by the additional separation of the
fingers involved in ESD current path. However, due to the longer current path,
the on-state resistance of the Fig. 3.54b configuration is larger. Additionally, the
present device architecture allows control of the clamp holding voltage by varying
the distance between the “diode” finger and the BJT finger and by variation of the
value of the base–emitter resistor. Variation of these parameters allows the change
of operation from SCR to snapback NPN mode.
For the purpose of comparison, pulsed I–V curves were measured for a reference
unidirectional clamp based on the standard snapback NPN clamp. This clamp has
12 fingers (with collector regions) and occupies a total area of 70 μm × 50 μm.
As seen in Fig. 3.56, the current tolerance of both clamps is sufficiently large,
3.7 ESD Diodes and Passive Components 139
1E-11 1E-10 1E-9

4 4 4
TLP Current (A)
TLP Current (A)

2 2 2
0 0 0
0 5 10 1E-11 1E-10 1E-9
TLP Voltage (Volts) Leakage Current (A)
Fig. 3.56 Comparison of the positive TLP characteristic of the dual-direction merged collector
integrated clamp (square symbols) with the prototype uni-direction BJT NPN snapback clamp
(triangle symbols)
therefore dual-direction ESD protection does not require significantly more lay-
out area than the unidirectional clamp. This demonstrates the advantage of the
integrated dual-direction clamp in comparison with the back-to-back stacking of
isolated unidirectional cells.
A valuable advantage of the suggested device architecture is the enabled device-
level holding voltage control. Simple topological modifications resulting in change
of the device mode operation from snapback NPN avalanche–injection conductivity
mode with a high holding voltage to double injection mode in the formed SCR
structure.
3.7 ESD Diodes and Passive Components
This final section presents an overview of forward-biased ESD diodes specific to

analog ESD design to demonstrate common design methodology that could be
applied to such components. Passive components are also discussed.
3.7.1 Forward-Biased ESD Diodes

In spite of simplicity, forward-biased ESD diodes are very important components.
In the case of rail-based approach (Chapter 5), the diode supports the current path
and critically contributes to the total voltage drop. In the case of high-speed I/O, the
parasitics introduced by the diode structures are critical for the overall circuit perfor-
mance. In the case of low-noise amplifiers (LNA) or RF pins, the ESD requirements
may even have to be traded off to provide on spec chip performance.
In this section, several aspects of the operation of forward biased diodes are
emphasized within the limits of this book. This is done because the description of
analog ESD design will otherwise be incomplete without discussion of such a major
component as an ESD diode.
There are several important considerations that should be taken into account for
ESD diodes.
3.7.1.1 CMOS Diodes

In the CMOS process, the diodes are usually formed by n+ and p+ diffusions in
the Nwell or Pwell (Fig. 3.57). In the case of I/O, the Pwell diode can typically be
connected between the I/O pad and the GND or ESDMINUS bus, while the Nwell
diodes provide a “natural” isolation from the p-substrate and is thereby suitable
for connection between the I/O pad and higher supply (VDD, ESDPLUS bus) (see
Chapter 5).
C A C
STI n+ STI p+ STI n+ STI
Nwell
P-SUBSTRATE
A C A
STI p+ STI n+ STI p+ STI
Pwell
P-SUBSTRATE
Fig. 3.57 The cross-section for Nwell and Pwell CMOS base ESD diodes
From an ESD operation point of view, it is important to emphasize that the

ESD diode is acting as a lateral device at high-current level, while the parasitic
capacitance and leakage are defined by the total device area.
The lower high-current resistance, lower parasitic capacitance, and lower leakage
are usually desirable targets for the ESD diode spec. At first, the intuitive guess
brings an idea to implement the ESD diode using a minimum diffusion length for
p+ and n+ contact regions. This approach is reasonable, but not always correct.
Even in the case of non-silicided 0.5 μm processes, a more optimal diode struc-
ture may require different diode dimensions. Moreover, in the case of nanoscale
process technologies, the backend, rather than the silicon, regions can already limit
diode performance.
The minimum composite area (open silicon surface area without field oxide or
poly) may not be the most optimal choice, due to specifics of the silicidation process.
In some technologies, it can result in limited opening of the composite area due
residual nitride layers. In this case, a substantial resistance may be observed. This
effect may not be seen in standard devices at normal operation conditions due to a
much lower current density.
Simulation results for 0.13 μm demonstrate that a minimum diffusion length is
not optimal.
3.7.1.2 Gated Diodes

A practical alternative to CMOS diodes is gated diodes. The gated diodes have been
proposed in [58]. Their implementation in the 0.13 μm CMOS process is shown in
Fig. 3.58.
n+ p+ n+ p+
a) b)
Fig. 3.58 Comparison of the process simulated cross-sections for STI (a) and gate (b) p-well
diodes
These devices can provide a smaller footprint, due to lower on-state resistance
and reduced parasitic capacitance of the diode clamp.
According to simulation results (Fig. 3.59), the gated diode provides a much
lower (30–50%) on-state resistance, in comparison with the standard diode.
At the same time, a special effort should be applied to guarantee appropriate
voltage tolerance. Skipping either nldd or pldd does not affect the forward I–V char-
acteristics (Fig. 3.59). For composite diode the breakdown voltage is ∼8.5 V, for
gated it is ∼4.5 V, and skipping nldd implant improves the breakdown voltage up to
∼6.5 V (Fig. 3.60).
According to TCAD analysis data, the gated diodes are indeed a very interesting
solution for aggressive high-speed I/O or system-level ESD protection. For both
P- and N-Well-gated diodes with a 0.28 μm gate length, the forward current is more
than 50% higher than for the corresponding composite diodes.
2.0E-02
Comp, nldd, pldd
Gated, nldd, pldd
Gated, nonldd, pldd
Current (A/µm)
Gated, nldd, nopldd
Gated, nonldd, nopldd
1.0E-02
0.0E+00
1 1.1 1.2 1.3
Voltage (V)
Fig. 3.59 Forward bias I–V characteristics for different types gated diodes implemented in
0.13 μm CMOS process with different lightly doped drain combination
1.0E-01 Comp, nldd, pldd, no btbt

1.0E-02
Gated, nldd, nopldd, btbt
1.0E-03
1.0E-04 Gated, nonldd, pldd, btbt
1.0E-05
Gated, nonldd, nopldd, btbt
1.0E-06
Current (A\µm)
1.0E-07
1.0E-08
1.0E-09
1.0E-10
1.0E-11
1.0E-12
1.0E-13
1.0E-14
1.0E-15
1.0E-16
1.0E-17
1.0E-18
0 1 2 3 4 5 6 7 8 9 10
Source - Drain Voltage (V)
Fig. 3.60 Comparison of the reverse I–V characteristics of the gated diodes implemented in
0.13 μm CMOS process with different lightly doped drain combination with a regular (composite)
diode
The lower breakdown voltage of gated diodes is a concern, however, skipping

hnldd implant for P- and pldd implant for N-Well diodes leaves the forward current
unaffected while increasing the breakdown voltage by at least 2 V.
3.7.2 Passives
The robustness of passive components is also critical for analog ESD design. Poly
and diffusion resistors are often used to separate different inputs and control pins
from the pad, implementing two-stage ESD protection (Chapter 4). Thin film resis-
tors might interface with the pad directly, for example, in precision amplifiers. MIM
and poly capacitors may be part of both the ESD clamp and the analog circuit com-
ponents that interface with chip pins. This subsection highlights some aspects of
passive components performance specific to ESD conditions.
3.7.2.1 Saturation Resistors [59]

A saturation resistor can be formed using a layer available in the process both at the
surface (n+, nldd, pldd, Pwell, Nwell, etc.) and in the depth of the device by reusing
subcollector regions.
The resistor provides a close to linear I–V dependence up to the saturation current
level, where the density of the injected carriers becomes comparable with the doping
level. After the saturation region, the conductivity modulation phenomenon results
in device snapback. This effect was already discussed in Chapter 2 for n+–n–n+
structures.
For saturation resistors and many other passive components, TLP or dc curve
trace characterization is required to determine the SOA of the devices with specific
parameters.
Diffusion resistors should be always assessed according to their voltage toler-
ance. Other undesirable features are the parasitic capacitive coupling to substrate
and noise, besides the limited resistance range and the additional space required
on the chip. That is why preference is often given to unsilicided poly resistors as
devices of choice for ESD clamps and the ESD network.
Some experimental data for P-buried layer (PBL) and N-buried layer (NBL)
resistors are presented below. In the process, an n-epi is grown after the N-buried
layer. The doping in the NBL is approximately 3 × 1019 cm–3 , compared to the
5 × 1015 cm–3 doping level in the epi layer. The buried layer is contacted through an
n-type poly and N-sinker. The resistor is completely isolated from the substrate and
adjacent devices, due to the deep trench and bonded substrate (SOI). Figure 3.61c
is a 1D doping profile from the process simulation of the NPN transistor and shows
the doping profile of the Nepi, N-sinker, and NBL layers. Figure 3.61a shows a 2D
cross-section schematic of the NBL resistor.
The PBL forms the PBL resistor. The Nepi, which is grown after the buried
layer, is counter-doped using the Pwell. The doping profiles are shown in the 1D
cross-section of the PNP transistor, obtained from a process simulation. The 2D
cross-section schematic in Fig. 3.61b illustrates the structure of the PBL resistor.
A typical layout of a buried resistor is shown in Fig. 3.62. The L and W indicated
on the drawing correspond to the length and width of the resistor in the following
sections. Three rows of contacts have been drawn on both sides, with a maximum
number of contacts for the width. In the 10 μm-wide resistor, there are three rows of
three contacts, and in the 20 μm wide resistor, there are three rows of eight contacts.
The results of the curve trace indicate that there is not a linear scaling of resis-
tance with squares, as defined using the drawn W and L parameters from the layout
(Fig. 3.63). There is an additional resistance associated with the sinker, which has
not been accounted for. The scaling of the resistance with a number of drawn squares
Buried oxide
a)
silicide
STI Poly trench protect
p poly p poly
p sinker Pwell p sinker
Deep trench
p buried layer
Buried oxide
b)
Nsinker Psinker
Nburied Pburied
Pwell
Nepi
c) d)
Fig. 3.61 Two-dimensional cross-section schematic illustrating the construction of the NBL (a)
and PBL (b) resistors, and 1D doping profile of the NPN (c) and PNP (d) transistors from the
process simulation
Fig. 3.62 Typical layout of the resistor
clearly illustrates that the resistance cannot be based purely on the buried layer
between the sinkers. In fact, the drawn resistor at 2.5 squares has a lower resistance
than the resistor drawn at 2 squares. Thus, in general, implementation of saturation
resistors requires some characterization effort.
Fig. 3.63 Scaling of resistance with number of squares, as defined at layout
0.1 0.1
L = 20 µm; L = 50 µm;
L = 20 µm W = 20µm
W = 20 µm W = 20µm
0.08 0.08
Current (A)
Current (A)
L = 50 µm W = 20µm
0.06 0.06
0.04 0.04
L = 20 µm;
W = 10 µm
0.02 0.02
L = 20 µm W = 10µm
0 0
0 5 10 15 0 10 20
Voltage (V) Voltage (V)
a) b)
Fig. 3.64 I–V characteristics of the NBL (a) and PBL (b) resistors measured on the curve tracer
Breakdown voltage for the buried resistors varied between 10 and 12 V, with
the exception of the PBL L50 × W20 resistor that exhibited a higher breakdown
voltage at 17.5 V (Fig. 3.64). All the resistors measured exhibited a snapback-type
of characteristic, but this was not recorded in the curve trace, due to the destruc-
tive nature of this measurement. The resistors provide a region of current saturation
above 2–3 V in the N-buried resistors, and above 5 V in the PBL resistors.
The TLP curves measured on the n-buried resistors show a saturation region
between 5 and 12 V, but then the saturation region appears to collapse and a region
of lower resistance dominates (Fig. 3.65). However, in the p-buried resistors, no
saturation region is observed up to 20 V (Fig. 3.65).
The resistance is not defined solely by the geometry of the buried layer next to the
sinker. In addition to the contribution of the contact resistance, the doping profiles
of the sinker and buried layer must be accounted for in the calculation of resistance
scaling. This is a good topic for TCAD analysis and may require 3D analysis.
The second observation shows that the resistors do not behave as saturation resis-
tors at ESD conditions. Using the resistance extracted from the TLP measurements,
Fig. 3.65 I–V characteristics of the n-buried (a) and p-buried (b) resistors measured on the TLP
system at ESD conditions
70
TLP CURRENT (mA)

70 Thin Film Resistor Sample RC
60 RC 60
50 50
CURRENT (mA)
RD
40 40
RD
30 30
RA
20 RA
20 RB
10 10
RB
0 0
0 50 100 150 200 250 100 200 300
VOLTAGE (V) LEAK.CURR. (@1mV) (nA)
Fig. 3.66 TLP characteristics of the selected resistor structures with parameters indicated in
Table 2.1
it should be possible to use the resistors to limit current, if the element protected can
sustain the voltage drop.
3.7.2.2 Thin Film Resistors

Thin film resistors (TFR) are used in many analog applications, for example, in
precision amplifiers.
An example of ESD I–V characteristics for a thin film resistor with a sheet resis-
tance of ∼1 k/sq is presented below. A SiCCr film with precision matching (<0.1%
at 3 sigma deviation) forms the resistor.
The maximum ESD current depends on the aspect ratio of the TFR. Resistor
width plays the main role in the resistor’s ability to handle high-current values: the
3.8 Summary 147
wider the resistor, the larger current that it can handle. Even if inputs are separated
by large value of TFR, ESD clamps are always necessary, at least to protect the
pad dielectric. Typically, pad dielectric can withstand a voltage of ∼200 V, which
is significantly lower than the voltage amplitude during the ESD event, even in the
package-level test. At the same time, ESD protection may be needed for the resistor
itself, if under the ESD stress there is residual current flowing through the resistor,
as the ESD clamp keeps it biased at a certain voltage value.
Normal operating DC current density (per width) for these resistors is
0.05 mA/um. Typically, additional ESD protection for the resistor is deemed
necessary if the residual current value under ESD stress is 40× of the operating
current density, i.e., 2 mA/μm.
Maximal ESD current density is ∼3 mA/μm. The maximal operating current is
40 × 0.05 = 2 mA/um, which is very close to the max ESD value. For conservative
ESD design, this resistor has to be either made very wide or have an additional clamp
that fixes current at <2 mA/um. Otherwise, further reliability studies are needed to
more precisely find the boundary of the resistor robustness, i.e., neglect the 40× rule
and find the exact max operating DC current.
Both the critical power and the absolute maximum pulsed voltage of the resistor
depend upon the total resistor area. The power and voltage increases with increase of
the area at a given resistor value. Within measurement accuracy, the specific critical
power for the beginning of degradation is ∼10 mW per square micron (Table 3.1).
Table 3.1 Summary of the experimental data for TFR
Parameter Sample RA Sample RB Sample RC Sample RD
Resistor value (k) 10 10 4 4

Total area (width × length μm2 ) 3 × 30 2 × 20 12 × 48 6 × 24
Critical snapback voltage (V) 98.4 61.2 155 78.6
Critical snapback current (mA) 10.2 6.2 39 19.6
Critical pulsed power for degradation (W) 1 3.8 6 1.54
Critical current density per resistor width 3.4 3.1 3.3 3.3
(mA/μm)
Specific critical power for degradation 11.2 9.5 10.5 10.7
(W/μm2)
3.8 Summary
In this chapter, it is demonstrated that the phenomena of the same nature are respon-
sible for both limiting the absolute maximum voltage and determining critical
snapback voltages in standard devices. In each case, the snapback effect in the
device can be reduced to fundamental conductivity modulation mechanisms that
provide structure-level positive feedback.
The reversible operation of ESD devices in pulse conditions is achieved by imple-

mentation of the device-level negative feedback to limit current increase caused
by positive feedback. Both approaches are suitable for engineering ESD devices
with appropriate characteristics for optimal voltage waveform at ESD pulse. An
innovative use of device-level negative feedback provides an opportunity to create
self-protected devices.
A set of transformation rules is demonstrated to enable implementation of the
ESD devices by modification of the standard devices. The focus of the ESD
device material further illustrates that in most cases, the problem of ESD pro-
tection can be solved at the level of free ESD device implementation, which
avoids alteration of the process technology or implementation of additional mask
layers. Specific efforts are applied to design devices with low sensitivity to
mask misalignment. ESD device engineering requires a strong understanding
of integrated process technologies architecture and process-step specifics, while
also requiring substantial experimental verification and validation of the final
solutions.
At the same time, the ESD device can contain a number of parasitic structures
that provide alternative current path scenarios, internal reference, and components
for reverse path ESD current.
The physics of operation of the devices discussed in this chapter can be gen-
erally understood through the analysis of the 2D cross-sections. Thus, numerical
simulation tools are adequately accurate for comparative analysis and validation of
the new ideas.
The next chapter brings the ESD devices into 2.5D or 3D levels, where additional
dimension is involved to treat the ESD device as a distributed integrated object fab-
ricated on the surface of the semiconductor die. This aspect is one of the features
that separate the ESD device level from the clamp level. Another feature involved
in the clamp level is the subcircuit, which can be attached to the control electrode
of ESD device. The subcircuit is used to control the dc, triggering, and holding
voltage characteristics. Thus, the characteristics of ESD devices at different control
electrode conditions are a rather important part of the easily collected experimental
results. In addition the next chapter will demonstrate the significant advantage that
can be obtained from an active control of the triggering characteristics of the ESD
device by a driver.

DECIMMTM Simulation Examples for Chapter 3 149
Example 3.1 Standard Devices in BCD Process Technology

Library Name: Examples3 Default BCD and ESD Devices
Project Names: E3.1a_HNMOS; E3.1b_HPMOS; E3.1c_NPN; E3.1d_PNP
This set of examples provides fully parameterized device structures representing

devices most commonly used in typical 0.5 μm BCD process technology. Among
library devices, there are 5 V NMOS (Fig. E3.1a), 5 V PMOS (Fig. E3.1b), N-
well and P-well diodes (Fig. E3.1c), 24 V NPN and PNP BJT devices (Fig. E3.1d),
Fig. E3.1a Cross-section of fully parameterized NMOS structure and simulated family of drain
curves for different gate bias values
Fig. E3.1b Cross-section and simulated output simulation results for the default 5 V PMOS device
Fig. E3.1c View of fully parameterized cross-section for N-well and P-well diode structures
compatible with the CMOS process
Fig. E3.1d Cross-section for fully parameterized 24 V NPN BJT and family of simulated output
characteristics
24 V NLDMOS (Fig. 3.1e). In addition, vertical trench DMOS and IGBT devices
are also included (Fig. 3.1f).
The main goal of the numerical analysis presented here is simulation of the quasi-
static isothermal I–V characteristics, including the regimes of avalanche breakdown,
strong injection, and conductivity modulation. Utilization of fully parameterized
device structure geometries and implant profiles makes it easy to explore the impact
of changes in device structure on the device performance in different regimes.
Example 3.2 Typical ESD Devices in 0.5 µm BCD Process

Technology
Library Name: Examples3 Default BCD and ESD Devices Project Names:
E3.2a_SNMOS; E3.2b_NLVTSCR; E3.2c_PLVTSCR; E3.2d_FOXSCR;
E3.2e_BSCR; E3.2 g_NLDMOS_SCR
Fig. E3.1e Cross-section for fully parameterized 100–600 V vertical DMOS and IGBT structures
and simulated isothermal drain I–V characteristics for vertical DMOS for different N-epi depth
values and gate bias conditions
This set of examples provides fully parameterized device structures representing

typical snapback ESD devices most commonly used in local ESD protection clamps
in typical 0.5 μm BCD process technology. Among many library devices, there are
5 V Snapback NMOS (Fig. E3.2a), high-voltage bipolar SCR and CMOS based
field-oxide SCR (Fig. E3.2b), 5 V NLVTSCR, 5 V PLVTSCR, and 24 V NLDMOS
SCR (Fig. E3.2c). Suggested use of the numerical analysis includes simulation of
the quasi-static isothermal I–V characteristics, breakdown and holding voltage and
on-state resistance as a function of the device structure parameters. Utilization of
Fig. E3.1f Cross-section and simulated output drain I–V characteristics for fully parameterized
the default 20 V NLDMOS device
Fig. E3.2a Cross-section of fully parameterized 5 V snapback NMOS structure and simulated
output I–V characteristic
Fig. E3.2b Cross-section of fully parameterized high-voltage bipolar SCR and field oxide SCR
Fig. E3.2c Cross-section of fully parameterized 5 V tolerant NLVTSCR and PLVTSCR and high-
voltage NLDMOS-SCR with simulated I–V characteristics for 15 V version
fully parameterized device structure geometries and implant profiles facilitates study
of the impact of changes in device structure on the device operation in different
regimes.
Example 3.3 Ring Oscillators

Library Name: Examples3 Default BCD and ESD DevicesProject Name:
E3.3a_Ring_Oscillator_w_1V5_Compact_Model_DevicesProject Name:
E3.3b_Ring_Oscillator_w_FE_5 V_CMOS_Devices
These examples illustrate operation of ring oscillator circuits and highlight versa-
tility of the mixed-mode simulation approach, which allows employment of both
M1* M3* M5* V 0 $Vdd*(1...

M1 M3 M5 M7 M9 V 0 $Vdd*(1...
W=1 W=1 W=1 W=1 W=1 R0* W=1 W=1 W=1 R0*
M0 M2 M4 M6 M8 M0 M2 M4
1e-3+1e... 1e-3+1e...
W=0.5 W=0.5 W=0.5 W=0.5 W=0.5

W=1 W=1 W=1
Fig. E3.3 Mixed-mode simulation circuits for five- and three-stage ring oscillators built using
compact models and FEM devices, respectively. Voltage waveforms at different stages of compact
RO, and different PMOS/NMOS width ratios in the FEM RO circuit
highly physically accurate FEM devices and computationally efficient compact

models to achieve high simulation speed and accuracy.
Chapter 4
ESD Clamps
The previous chapter described ESD devices with emphasis on the positive and
negative feedback effects realized on the structure level of the devices. The positive
feedback effects have been identified through the conductivity modulation phenom-
ena in the parasitic n–p–n, p–n–p, or p–n–p–n elementary structures. At the same
time, ESD devices already include local and non-local current limiting on the device
level. This feature provides the negative feedback that is used to suppress the exces-
sive positive feedback and limit uncontrollable current density increase. Thus, the
ESD device structure combines the active device regions with blocking junctions,
the drift region, RESURF regions, control electrodes, and the contacts.
At the same time, a real implementation of the ESD protection device in the
ESD protection cell for integrated process technology involves many more different
aspects. These include device width scaling, lateral isolation, vertical isolation from
substrate, latch-up guard rings and, of course, all the clamp components for voltage
reference and dynamic coupling.
As an object, the ESD clamp (or ESD cell) presents the next level of hierar-
chy for ESD protection solution engineering in semiconductor integrated process
technology. Often, the ESD clamps are treated simply as analog circuit building
blocks. Thus, intellectual property (IP) for the clamps is usually released in the
form of ESD libraries for the given CAD environment. Clamp IP development is
usually completed according to a specification for the envisioned product pins spec
in correlation with ESD protection network design.
In general, a clamp combines a distributed ESD device with appropriate width
scaling according to the ESD current specification and additional circuit compo-
nents. The additional circuit components are optional and are used to bring the
triggering characteristics of the device within the ESD protection window. Often,
for the ESD device with internal voltage reference, the clamp component count can
be relatively small, containing, for example, only the reverse path diode. However,
in some cases demonstrated in this chapter, rather complex drivers can be used to
enable and disable the ESD clamp depending on operation conditions.
This means, in principle, that the clamp can be designed based upon a single
ESD device structure, if the ESD device itself provides appropriate characteristics
that already meet the protection clamp specifications. The following parameters of
the clamp are usually major figures of merit: the voltage triggering the device into a

156 4 ESD Clamps
high-current state, the holding voltage, the transient response (turn-on time, voltage
waveform), dc voltage tolerance, the breakdown voltage at a given current, the leak-
age current for the maximum operation voltage, dimensions of the clamp footprint,
used mask layers count for active layers and metallization, RF parasitic, reverse
current clamping, recovery time, and some other parameters more specific to the
particular application.
In most cases of package level requirements, the triggering characteristics of the
clamps should be guaranteed within normal operation temperature conditions, while
the clamps’ parasitic characteristics should be provided for the whole operational
temperature range. In the case of system level ESD protection, the specification
may include much broader requirements.
Often, the clamp requires additional components to incorporate all the necessary
capabilities.
The most typical clamp components include the following:
(i) Triggering circuit: a dynamic coupling, reference voltage, or reference current

sub-circuit. This subcircuit is used to provide clamp turn-on inside the ESD
protection window voltage range.
(ii) Holding voltage control components.
(iii) High-current ballasting and current crowding prevention design.
(iv) Reverse path diode for voltage clamping at negative current direction when the
pad is biased at negative voltage.
(v) Latch-up isolation guard rings.
(vi) Secondary and following stages components.
The clamp may also include rather sophisticated subcircuits with control, shut-
down, or enable functions selected depending upon the internal circuit state. At
the same time, one of the trends in creating small footprint solutions is the use of
internal device-level features that integrate the clamp components inside the device
structure.
The other major difference between the ESD clamp and the ESD device structure
is the specific clamp layout design for both drawn and generated process mask lay-
ers. This difference is essentially between the 2D and 3D ESD devices, or at least
the 2.5D physical description required for the clamp.
Layout implementation of the ESD device is a very critical part of the clamp. It
brings a number of important peculiarities. Often, minor changes on the layout level
may significantly change the triggering and other characteristics of the final clamp.
This is why a basic principle of ESD design is verification of the re-designed clamp
characteristics on the test chip prior to application of the clamp in the product.
Usually, ESD protection clamps are placed near the bond pads or micro-SMD
(surface mounting device) bumps to avoid the excessive voltage drop on the met-
allization routing and to reduce the impact of the injection current in the clamp on
the internal circuit latch-up. Thus, the pad and pitch dimensions across the pad ring
usually determine the width of the ESD clamp that supports a product-independent
ESD pad ring design. In most cases, the clamp width in the layout is a specified
4 ESD Clamps 157
parameter. For example, at a pad and pitch spacing of ∼120 μm, the drawn clamp
layout width should not exceed this size. At the same time, in order to collect a ∼1.5
A current from an ESD device with a current density of ∼3 mA/μm, the required
total device width is ∼300 μm. Thus, in most cases, the ESD device should be
designed in the form of an array. The total amount of fingers N multiplied by the
finger width w in the array is usually within the range W=N×w=100–1000 μm.
In such a multifinger distributed ESD device, it is either hard or simply inefficient
to make each part of the device with identical internal parameters. Due to truly 3D
design, each different part of the finger and the different fingers themselves may
have a different metallization resistance, different effective spacing to the blocking
junction connection, or different control electrode reference.
Thus, the device cross-sections for each segment in the distributed layout can
only roughly be considered congruent. For example, in a snapback NMOS cell
(Section 4.2), the distance between the well tap and each finger’s source diffusion
region is variable. This results in a difference in the effective base resistance of the
parasitic n–p–n structure. As a result, the cell turn-on can hardly be expected to
be uniform. Moreover, perhaps the most typical scenario is non-uniformity of the
middle or corner of the distributed ESD device in the cell.
Nevertheless, it is usually expected that in a properly designed ESD cell, each
micron of its width can contribute to the total ESD device current. In other words,
linear width scaling is assumed upon device width. This is possible because of the
propagation of conductivity modulation across the cell during turn-on, due to mutual
coupling of the device segments.
This simple understanding brings another important principle of ESD cell design.
The principle requires implementation of all the fingers of the ESD device within the
same shared well, epi, or other body layer of the device. This is done specifically to
enable sharing between the distributed device segments in opposite to, for example,
composing the cell by paralleling fully isolated fingers to collect the appropriate
level of current. Not following this principle can create a passing current level issue
due to multifinger turn-on [30] or cause very non-linear width scaling.
On the other hand, the topological degree of freedom is often advantageous in
engineering appropriate characteristics of the clamp itself. Thus, the topology of
ESD devices inside the clamps is the first aspect addressed in this chapter.
The second major subject is the reference subcircuits and techniques that can be
used to implement appropriate clamp characteristics during both ESD operation and
under normal operation conditions.
There is a tremendous variety of different ESD clamps and solution variations
covered by numerous original papers in the ESD field. The major source of up-to-
date information is the on-chip and device physics sections in the Proceedings of the
EOSESD Symposium [8]. One of the most recent and complete reviews for different
solutions is presented in [4–6].
The purpose of this chapter, however, is not to present most of the available find-
ings in the ESD clamp design field. Instead, similarly to all the following chapters,
only the most practical examples are used for in-depth discussion of the major
principles and methodologies involved in design.
158 4 ESD Clamps
This is done due to the following two reasons. First of all, this information is
limited in order to avoid overloading the book with excessive number of examples.
The second reason consists in the fact that in most cases, practical implementation
of ESD protection solutions requires experimentation and verification specific to the
particular process technology.
Based upon the authors’ experience, it is rather rare that solutions directly based
on publication can be expected to provide target performance for the new process.
The printed space of this book is used to provide understanding of ESD clamp
design rather than to serve as an encyclopedia or reference manual for all possi-
ble clamp implementations. Such an approach is expected to be more useful for
practical ESD design of clamps in new process technologies.
The material in the chapter will be presented in two steps. First, the most typical
ESD protection clamps are discussed. Then, several more complex examples are
discussed to illustrate the potential for innovation in the ESD design field. This
basic material is followed by more advanced sections focused on complex clamp
subcircuits, self-protection, and system level protection. The final classification of
the solutions is presented at the end of the section.
4.1 Active NMOS Clamp

Despite the focus of this book being on clamps based upon ESD devices operating
in conductivity modulation regimes, this section discusses the alternative solution
first. This alternative solution is the so-called active clamp that does not operate
in snapback. The first such clamp solution has been proposed in [60], and since
then became the most widely used solution for high pin count products, as well as
digital and low-voltage domains. There are many variations of implementations of
the active NMOS and PMOS clamps [61–70] that are used by industry leaders (see
Chapter 5 too).
Originally, the major principle of such clamp design was the rise-time-dependent
turn-on of the big distributed NMOS device in normal current conduction mode.
The dependence upon rise time provides turn-on during ESD pulse, but suppresses
clamp turn-on during a much slower powering voltage ramp.
In spite of certain application-related limitations, this clamp-level solution should
be considered, perhaps, first in any new design.
The major advantage of this clamp is that it does not require ESD snapback
devices. The solution enables a complete ESD waveform simulation using compact
models for the clamp components which are usually available.
A typical circuit diagram and layout of the clamp cells is presented in Fig. 4.1.
The clamp is composed of an RC-timer cell and a number of “slave” clamps.
The slave clamps contain an NMOS device with a low on-state saturation resistance
that provides monopolar channel conductivity during ESD operation.
In a simple case, the RC-timer is designed for typical time constant of ∼4–8 μs.
This guarantees a low current state of the slave clamp NMOS device during possible
transient in normal operation conditions and provides relatively low parasitic current
4.1 Active NMOS Clamp 159
RC Timer Layout
6us RC Timer Schematic
ESDP
R1
600K
VTRIG
C1
10p
ESDM
Clamp Layout Slave NMOS Clamp Schematic
ESDP
DN MP C2
80/2 20/0.4 0.7pF
I/O PAD VTRIG MN MN

DP 5000/0.4
5/0.4
80/2
ESDM
Fig. 4.1 Circuit diagram for the active clamp with RC-timer (a) and slave clamp cells (b)
during the slow powering sequence. At the same time, the time constant is sufficient
to completely discharge the HBM ESD pulse to safe voltage levels with a decay
time of 150 ns.
The clamp has two distinctive regimes of operation: the normal operation regime
and the ESD operation regime.
In the case of ESD event, fast transient voltage at the pad pin results in a cor-
responding rising voltage on the drain of the slave NMOS devices. Due to the
drain-gate capacitive coupling, the transient gate bias of the NMOS device exceeds
the threshold voltage which switches it into the “on” state with low channel resis-
tance. To improve the coupling voltage, speed-up capacitance CN is used in every
slave clamp. Due to the on-state condition, an NMOS device with an appropriate
total width provides a discharge current path with a low-voltage drop. In a properly
designed clamp, this results in a voltage limitation below the 1–2 V level. After
the first ∼300 ns of fast transient operation, the slave clamp inverters regain con-
trol of the NMOS device’s gate potential which is now determined by the RC-timer
state. The RC time constant for the RC-timer is usually chosen ∼6 μs. After the RC
timer capacitor CT is sufficiently charged through the resistor RT , the slave inverter
switches the NMOS device into a low leakage “off” state.
In the case of a normal powering up sequence, the power supply voltage ramp
is usually relatively slow. This results in a rather small parasitic current through the
clamp due to the relatively fast response of the RC-timer.
A relatively large width NMOS device of ∼4–10 mm is required to support an
ESD current level. The RC-timer components are space consuming as well since
160 4 ESD Clamps
large area capacitor and resistor are required. The ESD protection network (Chapter
5) includes several slave clamps composed with fewer RC-timers.
The implementation of active clamps in the ESD protection network will be
discussed in Section 5.1.
A more complex approach is proposed in [71]. This proposed circuit fulfills a
dual function by detecting the slew rate and controlling the duration of the turn-on
time of the active clamp. Detection of the slew rate has the advantage of recognizing
input pulses as ESD or other, such as hot plug-in or power on.
There are several limitations related to the active clamp design:
(i) Clamp size: usually, ESD protection with clamps becomes efficient in analog
circuits only in the case of 8–10 pins that belong to the same voltage domains,
for example, in a digital interface.
(ii) Low-voltage NMOS: the clamp has a reasonable size only in case of low-
voltage NMOS devices. Clamp design using NLDMOS is in most cases
impractical. Similarly, the voltage tolerance of the clamp is limited by
the hot carrier reliability limitations on the drain-gate bias of low-voltage
NMOS.
(iii) System-level and high ESD level spec solutions: this is again an issue due
to the limited size of the NMOS. While in standard packaged level require-
ments for 2 kV HBM and 200 V MM the 750 V CDM protection level can
be achieved with a reasonable clamp size of ∼1,00,000 μm2 , the system-level
requirements for power-on and ESD current level tolerance ten or more times
higher typically make active clamp design impractical.
(iv) High-speed limitations: in the case of fast transient pins, the clamp cannot
support normal operation conditions and will drain excessive current through
the protected pin.
(v) ESD PLUS and ESD MINUS bus: organizing an ESD network with an active
clamp requires a low-resistive ESD bus structure.
(vi) Cross-talk: in the ESD protection network, pins with shared active slave clamps
will detect some signal due to parasitic coupling in diodes and the finite resis-
tance of the metal interconnects. Thus, this clamp solution is not applicable to
the multiple channel low-noise amplifiers with multiple inputs and outputs, for
example.
Thus, in spite of the elegance of this solution and the provided low clamp-
ing voltage in analog and high-speed designs, high-voltage and multiple domain
designs utilize snapback clamps that employ conductivity modulation mecha-
nisms (discussed in Chapter 2) and are based on ESD devices (discussed in
Chapter 3). These snapback clamps are discussed in the following sections of
this chapter.
Other types of active clamp solutions based on NPN BJT and PMOS devices,
as well the ESD protection network with active clamps, are discussed in the first
section of Chapter 5.
4.2 Low-Voltage Clamps with Internal Blocking Junction 161
4.2 Low-Voltage Clamps with Internal Blocking Junction

Reference or dV/dt Turn-on
4.2.1 Snapback NMOS Clamps

4.2.1.1 Ground-Referenced Snapback NMOS
Snapback NMOS is one of the most popular solutions for local ESD protection
that can be implemented in any CMOS process as a “free” ESD device. The alter-
native name for this clamp often used in publications is grounded gate NMOS
(ggNMOS).
Depending on the technology used, the snapback NMOS device is usually
capable of supporting up to a 2–5 mA/μm current density in high-current mode.
Respectively, the size of the snapback NMOS ESD device in the clamp usually
varies from 200 to 600 μm. At the pad dimension of ∼60–100 μm, the specified
clamp width with all guard rings is expected to be within ∼100 μm. Therefore,
the layout is implemented as an array with multiple fingers of 40–70 μm width
each.
An example of the 5 V snapback NMOS clamp schematic and layout view for
a 0.5 μm process is presented in Fig. 4.2. The circuit for the clamp includes a gate
resistor and an optional reverse path diode.
In principle, operation of the clamp is based upon parasitic NPN avalanche–
injection (Chapter 2) controlled by the channel current incoming into the multipli-
cation region (Chapter 3). Therefore, there are three controlling factors for clamp
turn-on:
(i) drain avalanche breakdown

(ii) dynamic coupling of the gate
(iii) displacement current in the parasitic n–p–n structure
4.2.1.2 Gate Coupling

A gate resistor of a sufficiently high value can be selected to enable turn-on of
the clamp due to the gate capacitive coupling effect by the fast voltage rise on the
protected pad. This dV/dt effect is the result of the triggering voltage dependence on
the level of channel current.
In addition to the gate coupling effect the lengthy drain ballasting region already
results in a, respectively, higher internal drain-bulk capacitance, in comparison with
the bulk-source capacitance, the internal divider creates a displacement substrate
current. Since the critical voltage for current instability in a silicon NMOS device
usually depends upon both the channel current and substrate potential (Chapter 3),
the triggering voltage in transient conditions is usually achieved at significantly
lower level than the dc breakdown voltage. For example, for the clamp in Fig. 4.2,
the dc breakdown voltage is ∼13 V.
162 4 ESD Clamps
ESDP
Double Deep
Trench P- and
N- Guard
Rings
Multifinger
Snapback
NMOS
ESDM
Polyresistor
Reverse Current Path CMOS Diode with Pwell Contact
a) b)
Leakage Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
4
3 SWS=20 μm
Pulsed Current (A)
SWS=2 μm
1
0
0 1 2 3 4 5 6 7 8 9 10
Pulsed Voltage (V)
c)
Fig. 4.2 Layout view (a) and circuit diagram (b) of the snapback NMOS clamp and measured
TLP characteristics (c) for two different well-to-source parameters SWS (c)
4.2.1.3 Displacement Current Effect

The second factor controlling the clamp turn-on is the intrinsic base resistance of
the parasitic n–p–n device. In the clamp, the source and well tap p+-region are con-
nected at the metal. Therefore, the internal base resistance of the parasitic n–p–n
structure is the critical parameter for the avalanche–injection instability boundary
of the final distributed array. In principle, in the case of RG =0, an additional com-
ponent of the dV/dt effect can be observed as a decrease in the triggering voltage
with the decrease of the pulse rise time. This effect could be the result of a pure
displacement current effect in the parasitic n–p–n structure. However, in the case of
long finger NMOS devices formed in non-silicided processes or a silicide blocked
gate region, the gate coupling can still be a dominant factor. In these conditions, the
distributed resistance of the gate allows significant gate potential increase in tran-
sient conditions, even if the polygate has a metal connection to the source at the end
of each finger.
Exclusion of the silicide region from the gate is usually a side effect of the imple-
mentation of a silicide blocking region. One of the practical implementation issues
is the formation of a narrow silicide stripe region along the gate edge. This is possi-
ble if the silicide blocking mask is aligned with the side edge of the polygate drain.
The most practical measure to avoid such a silicide sliver formation is to align the
silicide blocking mask to the middle of the gate.
Since parasitic base resistance in the n–p–n structure has a significant impact on
the instability boundary, there are many practical ways to make clamp layout design
advantageous. The clamp is usually designed as an array of non-butted NMOS fin-
gers with spaced well tap contact regions to accumulate the necessary total width
(Fig. 4.2a).
There are several ways to implement the Pwell connection to increase the base
resistance of the internal NPN. In the most simplistic cases, a p+ ring that is
already required for latch-up isolation can be drawn at some distance from the
NMOS source-drain regions. In this case, the middle part of the array will have
a significantly larger spacing, in comparison with the corners of the array, with a
corresponding increase in base resistance. The increase of the base resistance will
reduce the triggering voltage (Fig. 3.2c) in accordance with avalanche–injection
conductivity modulation in n–p–n structures (Chapter 3).
In spite of the concern of possible current crowding, the total finger width
may still be utilized in high-current operation of the structure. Additional current
ballasting can be achieved using comb ballasting region design [72].
It can be shown that in such design, turn on of the middle of the device in the
conductivity modulation state will cause the remaining part of the device to engage.
Thus, the device will work as a single entity. In the case of isolated CMOS processes,
BiCMOS processes, or SOI processes, either a local or soft-well connection can be
realized.
Another important note can be made for dV/dt-triggered clamps. The transient
gate-coupling effect might become a problem in high-speed I/O circuits. At a data
rate of ∼10 Mb/s, the rise time of the signal at the pad is already faster than in
standard HBM and MM ESD pulses. In spite of the high slew rate, the voltage
amplitude cannot trigger the ESD clamp into snapback mode but the coupling of
the gate may result in significant current through the clamp, thus deteriorating the
high-speed performance of the I/O pin. Compact modeling with extracted parasitic
resistance and capacitance from metallization is helpful but cannot guarantee the
end result due unaccounted for parasitic n–p–n structures in typical NMOS models,
for example, in the widely used BSIM3v3.
164 4 ESD Clamps
Alternatively the reduced triggering voltage can be achieved by decrease of the

breakdown voltage with additional ESD implant [73].
4.2.1.4 Reverse Path Protection

For every local clamp, the most typical requirement is dual-direction protection.
Usually, the voltage specification at negative bias is low and therefore a simple diode
can provide the desired result. An advantage of the snapback NMOS clamp is the
option to integrate the diode inside the clamp. The diode can be easily formed by
adding an n+ -diffusion region near the p+ -diffusion in Pwell. The only important
consideration is to ensure appropriate metallization for such a diode.
The reverse path diode is a typical internal clamp component. In principle, the
diode is already presented in the clamp as a body diode and is formed by the p-tap
p+ region and the n+-drain diffusion region. However, due to the large distance
between these two regions, the diode resistance is rather high. Thus, an additional
n+-diffusion region is usually connected to the pad inside the cell in the proximity
of the p+-diffusion to form the diode. In the clamp (Fig. 4.2a), this region is placed
below the NMOS fingers. In this case, appropriate metallization should be provided
for the high-current level.
Respectively, there is the option to separate the p+ -tap region from the drain-
source area by the n+ reverse path diode region. This brings a new topological
degree of freedom for control of the base resistance of the parasitic n–p–n struc-
ture on the device level. The degree of impact for such a design approach depends
on the Pwell profile. The diode n+ -region provides additional depletion in the Pwell,
thus controlling the equivalent resistance to a higher degree.
Often, the reverse path diode can be implemented as separate fully isolated struc-
ture. Depending on the process, a 50–100 μm width is usually sufficient for a
standard minimum package level ESD spec (2 kV HBM, 200 V MM).
4.2.1.5 Isolated Snapback NMOS

The snapback NMOS clamp is usually designed to target the ESD protection win-
dow using the transient triggering effect. In this case, the triggering voltage is
realized below the expected triggering voltage for internal circuit devices with low
gate potential during ESD stress. However, if the coupling of the internal device
is high, protection of, for example, an open drain circuit creates a significant chal-
lenge. In this case, a device of the minimum gate length may provide the triggering
voltage at a rather low level, thus taking over the ESD current path. Usually, CMOS
process provides no degree of freedom in changing the blocking drain junction in
the snapback NMOS.
In the case of isolated CMOS process, the isolated snapback NMOS clamp can
be designed. For substrate isolation, a deep N-region (NBL, NISO, DeepNwell,
Chapter 3) is used in combination with appropriate spacing for lateral isolation. This
clamp is required for non-ground-referenced pins or high-side pins, for example,
BOOST PIN in DC–DC converters (Chapter 6).
The main design principles of isolated snapback NMOS clamps are the same as
those of regular snapback NMOS clamps. However, in the case of isolated devices,
an additional advantage can be derived from the reduced well resistance (Fig. 4.3).
The middle part of the clamp is the same as in non-isolated snapback clamp layout
design (Fig. 4.2a). An additional isolation N-ring and an additional N-epi region
connection are required.
ESDP to ESDM Pulsed Current (A)
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
4
V TLP
3 I leak (at 5 V)
0
0 2 4 6 8 10 12
ESDP to ESDM Pulsed Voltage (V)
Fig. 4.3 Measured TLP characteristics of the isolated snapback NMOS clamp
In BiCMOS processes, the buried layers can be used to control characteristics of

the NMOS-based clamps.
As a result, sheet resistance of the Pwell isolation can be higher if either the
P-buried layer is not used or the N-buried layer is used instead. In this case, the
triggering voltage of the clamp can be achieved at a much lower level. This can
be illustrated by a comparison of the TLP I–V characteristics for the non-isolated
clamp (Fig. 4.2c) and its isolated version (Fig. 4.3).
Both snapback NMOS and isolated snapback NMOS rely on the dV/dt effect in
order to bring the triggering voltage of the clamp inside the ESD protection voltage
range.
In addition to the already mentioned drawbacks of the transient triggered clamps,
the dV/dt effect also has a strong dependence upon the load realized by the internal
circuit (Chapter 5). The lowering of the rise time for the ESD pulse is simple to
understand, taking into account a possible additional capacitance of the internal
circuit which is effectively connected in parallel to the ESD protection structure.
This capacitance can be significant compared to the pulse forming capacitance, in
which case, the rise time of ESD pulse can be significantly increased.
The alternative method to adjust the triggering voltage involves a voltage refer-
ence component in the drain-gate circuit or alteration of the NMOS device toward
lower breakdown voltages.
In the case of the typical CMOS process, options for these changes are quite lim-
ited and very process dependent. It appears the only practical self-aligned approach
166 4 ESD Clamps
is to use longer gate length in the I/O components and minimum gate length in the
ESD NMOS clamp. Other possibilities include partly blocking the nldd implants
from core low-voltage devices in the case of the DGO CMOS process.
In the BiCMOS process, the PBODY mask and other shallow implants that sup-
port the bipolar part of the process may be used to reduce the breakdown voltage to
bring it inside the ESD protection window.
All these measures require experimental validation to confirm compatibility with
the given process. This statement is true for any snapback device solution.
Some I/O libraries are designed using snapback NMOS for the dual purpose
as both an I/O buffer component and an ESD protection clamp element, thus
introducing self-protection capability.
Various styles of NMOS clamp design can be implemented based on designer
preference or the process specifics. For example, based upon comparison of the
experimental results in Figs. 4.2 and 4.3, no difference is found between vertical or
lateral finger directions in the clamp, as long as a number of backend design rules
are followed. Another typical use of the isolated NMOS clamp is in the stacked
clamp for higher-voltage tolerance [74, 75].
4.2.1.6 The 40× Rule for Backend

For the ESD condition, the “40×” rule for metallization is usually followed. The
rule simply reflects a correlation factor between the critical density for metallization
burnout in ESD pulse conditions and the experimentally measured electromigration
rules of the process. The exact correlation factor can be experimentally determined
for the given process and, perhaps, may not be exactly 40. The rule simply suggests
that the safe current density level through the contact, via, and metal layer region can
be estimated by multiplying the electromigration current density maximum rating
by the number 40.
According to design experience, this rule is validated for at least 0.18–0.5 μm
process technologies, it may require additional verification for 90–32 nm process
technologies with new materials [76].
In an ideal design, clamp metallization is expected to provide the same metal
resistance to each point of the device. This is usually achieved by using comb-like
metallization, applying an approach similar to the metallization of power arrays.
Power arrays will be discussed later in this chapter.
One of the assumptions of snapback NMOS cell design is that device turn-on may
happen in part of the cell, but eventually all the fingers are switched into conduc-
tivity modulation mode. In other words, a turn-on domino effect is realized inside
the clamp. If a local part of the clamp or a part of a finger turns-on into the high-
conductivity state, the lateral diffusion of the carriers provides significant initial
current density in the avalanche multiplication region. Thus, spreading of the con-
ductivity modulation process occurs along the whole finger width and the ESD cell.
However, if each clamp finger is fully isolated, then there is no coupling between
the fingers. In this case, the first finger that turns-on takes over all the ESD cur-
rent, thus blocking turn-on of the other fingers. This effect may produce multiple
S-shaped regions that correspond to the contribution of the active fingers [30]. This
effect depends on the ratio between the triggering voltage VT1 and the critical volt-
age for maximum current VT2 . If VT2 is exceeds VT1 , then the additional fingers
may turn on after current saturation. However, it should be taken into account that
the secondary turn-on may occur in conditions of much higher rise time. Therefore,
at the same equal conditions, the triggering voltage VT1 for remaining fingers can be
substantially higher. Nevertheless, if VT2 < VT1 , then the burnout of the first finger
can be expected.
A proper design of both backend metallization routing and well tap connection
and gate coupling are important to avoid the multifinger turn-on effect. If the process
does not allow any alternatives, additional backend ballasting has been suggested in
[30].
An additional design parameter is contact enclosure in the diffusion region. One
important thing to consider is that process design rules are usually engineered to
support the current density for normal device operation. Taking into account process
variability and lateral diffusion, it is logical to expect that the minimum contact
enclosure might not be optimal for a high-current density ESD device. If at the
ESD current level the contact region is not quasineutral, then an excessive power
generation can cause an elevated local temperature at the contact which may result
in irreversible processes in the metallurgical contact structure. One of the adjacent
aspects of pertaining design solutions is implementation of the appropriate contact
enclosure in the lateral regions of the finger.
4.2.2 Transient-Triggered PMOS Clamp

Similar design concepts can be applied to the PMOS-based clamp [77]. Unlike in
the snapback NMOS clamp, the parasitic p–n–p structure in a low-voltage PMOS
clamp provides either absence of or moderate S-shaped I–V characteristics. This
phenomenon is a direct consequence of the different level of space charge compen-
sation in the p–n–p structure versus the n–p–n structure determined by the physical
difference in the carrier mobility of the silicon semiconductor material.
Respectively, the higher holding voltage is realized due to the ratio of nobilities
of the majority and minority carriers (Chapter 2) under the low gain of the parasitic
PNP structure.
However, the basic avalanche–injection process in PMOS is the same as that
in the NMOS, so the negative differential resistance effect is a strong function of
the implant profiles of the device. Usually, negative differential resistance can be
observed at a low current level in short channel PMOS devices with a gate length
below 0.18 μm, thus revealing the general nature of the effect.
At the same time, absence of negative differential resistance provides an advan-
tage. In the case of the PMOS clamp, current stratification is absent, eliminating the
need for the drain ballasting region. As a result, the clamp can be composed within
the same area, but with a much higher total width of the array.
168 4 ESD Clamps
Double Deep
Trench P-
Guard Ring
Multifinger
Snapback
BJT
Polyresistor
Reverse Current Path BJT Diode
a) b)
Leakage Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
5
4
Pulsed Current (A)
PBL
3
NBL
2
0
0 2 4 6 8 10 12
Pulsed Voltage (V)
c)
Fig. 4.4 Topology of the 5 V PMOS clamp (a), circuit diagram (b), and measured TLP char-
acteristics (c) for the clamp based upon a PMOS device with an NBL layer and PBL layer
The design and TLP characteristics of a W=1 mm total width 5 V PMOS clamp
are presented in Fig. 4.4. This clamp is implemented in the same BCD process as
the previously discussed NMOS clamps. The ring around the source-drain PMOS
arrays forms the N-well contact region.
With the same footprint as snapback NMOS, this clamp provides similar
high-current operation. The turn-on point is controlled by the high Nwell resistance
path and the gate-coupling resistor RG is connected to the pad.
The potential use of the clamp depends on the ratio between breakdown voltages
which can be achieved in NMOS and PMOS structures. In terms of transient effects,
the clamp has the same peculiarities as the transient triggered snapback NMOS
clamp.
In principle for the CMOS the clamp characteristics are dependent upon substrate
type due to internal base resistance impact. The impact of the substrate resistivity is
studied in [76].
4.2.3 10 V FOX Snapback Device

The next simple self-aligned ESD clamp for the CMOS part of the process is based
upon the field oxide device (Chapter 3). In principle, this clamp eliminates the gate
regions and subsequent gate-coupling effects, leaving only the free lateral NPN-BJT
structure.
The clamp can reuse the same topology and backend metallization as the pre-
viously designed snapback NMOS clamps. Thus, a simple transformation rule
replaces the former polygate region with the minimum dimension thick field oxide
isolation region that can also be applied on the cell level. An example of the ESD
clamp layout and schematic view is presented in Fig. 4.5.
In principle, the device in the clamp presents an implementation of a free lateral
NPN by reusing CMOS diffusion regions. This NPN has a long P-base and thus
a rather low current gain before breakdown. The high internal base resistance is
controlled by the Pwell-source and active region spacing (usually, naming conven-
tions for device contacts reuse the names of corresponding regions in the ancestor
snapback NMOS).
The major change, in comparison with the snapback NMOS (Fig. 4.2c), is the
higher triggering and holding voltage characteristics (Fig. 4.5c).
The major advantage of this clamp is based upon the device structure itself.
The device eliminates limitations on the voltage tolerance related to gate dielectric
breakdown and hot carrier degradation, combined with a predictable self-aligned
blocking junction formed by the n+-diffusion and Pwell.
Thus, this clamp is an efficient solution for a case where voltage tolerance of the
analog circuit pin exceeds the CMOS voltage level and thus protection cannot be
achieved by the snapback NMOS cell.
Most typical cases are the EPPROM “write” and “erase” pins or stacked I/O
protection.
The same topological control of the internal base resistance can be implemented
using local, partly isolated, or n+-diffusion blocked options for Pwell tap placement
acting as a base for the parasitic n–p–n structure.
For the CMOS process, device turn-on will be based on the given n+-Pwell break-
down voltage. Presence of the poly or metal field electrode on top of the isolation
region can be used for turn-on voltage control. However, such a solution is usually
rather marginal due to process design rules. Thus, the basic method in targeting the
170 4 ESD Clamps
ESDP
Double Deep
Trench P-and
N- Guard Ring
Multifinger FOX
Snapback
Device
Reverse Path
ESDM Diodes and
Pwell Taps
a) b)
DC Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02

6
5
V TLP
I leak (at 5 V)
4
0
0 2 4 6 8 10 12 14
c)
Fig. 4.5 Topology of the FOX clamp (a), circuit diagram (b), and measured TLP
characteristics (c)
ESD protection window is to utilize the dV/dt effect due to the displacement cur-
rent. An alternative method is to alter the base–collector (Pwell-source) blocking
junction (Chapter 3).
The base (Pwell) resistor to the ground enables additional lowering of the turn-on
voltage. In the case of non-isolated CMOS processes, the Pwell is softly connected
to the p-substrate and thus shunts the resistor. In this case, the substrate pump
method is realized [78]. In the isolated clamp version, the well diffusion connected
to the source through the resistor acts similarly to the external base resistor in the
NPN clamp in reducing the triggering voltage (Chapter 2). The Pwell (base) contact
can also be used to inject the voltage-referenced base current.
4.2.4 LVTSCR and FOX-SCR Clamps

Similarly to snapback NMOS and FOX devices, the LVTSCR and FOX-SCR clamps
can be implemented using the same footprint, topology, and design features to adjust
the triggering characteristics and provide an appropriate level of width scaling.
Transformation rules for the device can be applied on the cell topology level, sim-
ply replacing the internal snapback NMOS device array with corresponding SCR
devices, as described in Chapter 3.
An example of the LVTSCR clamp is presented in Fig. 4.6. Since the SCR device
usually provides an on-state current approximately one order of magnitude higher,
the clamp size can be significantly reduced (ultimately down to a 25–40 μm-width
ESD device).
Polyresistor
ESDP
Double Deep
Trench P- and
N- Guard Rings
Multifinger
LVTSCR
Device
Reverse Path
Diodes and
Pwell Taps
ESDM
a)
1.0E-07 1.0E-06 1.0E-05 1.0E-04
5.0
4.0
TLP CURRENT (A)
3.0
2.0
1.0
0.0
0 2 4 6 8 10
ESDP TLP VOLTAGE (V) b)
Fig. 4.6 Topology of the LVTSCR clamp (a) and measured TLP characteristics (b)
172 4 ESD Clamps
In the case of the LVTSCR clamp [51], the gate coupling is used to trigger the
LVTSCR into high-conductivity state similarly to triggering in snapback NMOS. In
field oxide SCR, the advantage of displacement current in the parasitic NPN and
PNP devices is used to turn the clamp into high-conductivity mode.
In spite of the advantage in high-current conduction, the disadvantage of the SCR
clamps is the rather low holding voltage. This is already discussed in Chapter 3
on the device level. There are several solutions for implementation of the higher
holding voltage levels using both device and subcircuit components.
In general, the low holding voltage of the clamp creates a major problem for pins
sensitive to transient latch-up. To eliminate this problem, several solutions have been
suggested that involve implementation of certain high-current positive feedback on
the clamp in order to increase the holding voltage above the power supply level.
This is discussed in Section 4.2.5.
According to the original design concept of the device, the triggering of the
clamp is first based upon the snapback NMOS turn-on. Then, when a certain current
density level is achieved during turn-on, the p+-emitter region becomes positively
biased over 0.8 V, and direct injection from the p+-emitter region initiates double
injection conductivity modulation (Chapter 2).
However, in principle, the clamp can be designed to fulfill the double S-shaped
characteristics. The first snapback region corresponds to the NMOS avalanche–
injection operation. Then, the condition for SCR double injection conductivity
modulation is achieved only at a rather high snapback current level. This case is
usually encountered at excessive p+-emitter isolation.
Alternatively, with p-emitter isolation reduction, the parasitic p–n–p structure
can become dominant and the device may turn-on directly into the double injection
mode. Repositioning the p+-emitter closer to the gate region or disconnecting the
drain can achieve this. In the last case, the device will be turned on by the punch
through effect even if it skips the avalanche breakdown stage.
Another typical challenge, especially for FOX-SCR devices, is the trigger-
ing voltage control. Section 4.2.5 presents some practical methods of triggering
characteristics control using a special control circuit connected to the clamp.
4.2.5 High Holding Voltage LVTSCR Clamps

Two methods of increasing the holding voltage have been suggested: circuit and
topology-based device solutions.
4.2.5.1 High Holding Voltage Cell Topology

In the LVTSCR device, positive feedback in the injection current from the emitter
and the source junction results in the formation of a quasi-neutral region, as a result
of mutual space charge neutralization of the injected electrons and holes. The final
holding voltage after triggering is an integral function of the resultant electric field
distribution in the discharge region.
The method of increasing the minimum holding voltage as a result of the change
in carrier balance on the device level has been proposed in [37, 79]. Carrier balance
control has been achieved by implementation of the different levels of emitter junc-
tion isolation on the device topology level. To achieve the controllable imbalance
in the injected charge, the p-emitter length is reduced down to a minimum feature
dimension while maintaining proper device spacing. The hole injection current is
limited, thus defining the holding voltage.
Similar effect has been reported in [80]. The effect was achieved by using
N-isolation region to change the carrier balance in the device.
A similar effect can be achieved on the clamp layout level by changing the emitter
isolation topology. One of the examples includes implementing the p+-emitter by
interdigitated islands surrounded as the n+-diffusion and increasing the distance
between the p+-emitter and the source, or reversing the positions of the drain and
the p+-emitter.
In an extreme case, if the p+-emitter is over-isolated, the snapback NMOS
operation current density level may not produce a p+-emitter voltage drop suf-
ficient to overcome the potential barrier of ∼0.6–0.8 V. Thus, no SCR-specific
double-injection effect will be produced in the device. Such an SCR device will
be physically equivalent to a snapback NMOS with an additional Nwell saturation
resistor in the wide current range.
This topological method has been validated by numerical simulation and fol-
lowed by pulsed measurements in [81]. The original device structures of the
LVTSCR were created using a calibrated 0.18 μm CMOS TCAD process simulation
flow. The 2D structures were generated using process [82] and device simulators
[83]. Cross-sections for the LVTSCR and NMOS structures used in the simulation
analysis are presented in Fig. 4.7a, b. The emitter isolation effect is presented in
Fig. 4.7c. A comparison between snapback NMOS and SCR with a reduced emit-
ter region demonstrates a higher current level from the SCR device optimized for
3.3 V-tolerant ESD protection.
The measured data confirmed simulation results. A strong dependence was
observed of the holding voltage on the structure and the length of the emitter region
(parameter LN, Fig. 4.7a). The on-state current is dependent upon the length of
the emitter. As the emitter region is reduced, the holding voltage increases and the
saturation current decreases. Stable triggering characteristics (at ∼10 V), with an
intermediate value for the holding voltage, are obtained in a 0.5–3 μm range of
the length of the emitter region. The holding voltage range achieved is between the
conventional LVTSCR holding voltage value (∼1.5–2 V) and the snapback NMOS
holding voltage value (∼5–6 V) (Fig. 4.8). A comparison of the measured snap-
back characteristics for the grounded gate snapback NMOS clamp and the LVTSCR
clamp with an emitter region that delivered similar holding voltages is presented in
Fig. 4.8b.
Thus, the high holding voltage in the LVTSCR provides both higher current in
the saturation region after triggering and, at the same time, higher critical power
174 4 ESD Clamps
Pwell Nwell
a)
Pwell
b)
0.03 EMITTER 0.05

LENGTH [LN ] (μm)
DRAIN CURRENT (A/um)
DRAIN CURRENT (A/um)
T = 300 K 0.04
0.02
0.03
R
S
C
0.02 MO
TS
-N
LV
0.01 ck
ba
ap
5
0.01
5
1.
Sn
0.
8
0
3
0.
1.
0.
0.00 0.00
2 3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 10
DRAIN-SOURCE VOLTAGE (V) c) DRAIN-SOURCE VOLTAGE (V) d)
Fig. 4.7 Simulation cross-section for LVTSCR (a) and NMOS structure (b). Isothermal I–V char-
acteristics for different emitter area lengths (c) and comparison of the LVTSCR structure with a
0.5 μm emitter region LN and the snapback NMOS structure (d)
(that corresponds to the final point on the TLP curve in Fig. 4.8b). Similar results
have been demonstrated for FOX-SCR clamps.
4.2.5.2 Clamp-Level High Holding Voltage Using P-Emitter De-biasing

An alternative and a more elegant way to solve the holding voltage problem on the
clamp design level in SCR structures has been proposed in [81]. The emitter injec-
tion control approach has been suggested and implemented using a new structure
(Fig. 4.9). This approach involves the high-current de-biasing diode structures.
DRAIN CARRENT (A) [W = 56um]
DRAIN CARRENT (A) [W = 56um]

3.5 3.5
EMITTER LENGTH (μm)
3.0 3.0
2.5 2.5
2.0 2.0
R
1.5 1.5
C
TS
1
S
LV
1.0 1.0
MO
6
0.
N
0.5 0.5 GG
0.0 0.0
0 2 4 6 8 10 12 0 2 4 6 8 10 12
a) b)
Fig. 4.8 Measured TLP I–V characteristics of LVTSCR: (a) for different emitter region lengths;
(b) comparison of NMOS and LVTSCR with the same holding voltage; the last point of each curve
corresponds to the beginning of soft-leakage degradation
The diode structures can be implemented as stand-alone or as internal compo-

nents inside an N-well. They create an internal circuit that provides an additional
voltage drop in the p+-emitter circuit (Fig. 4.9a, b).
Due to the additional forward biased junctions, the emitter injection is limited
after triggering by the voltage drop on the diodes, thus cutting off emitter injection
at the corresponding voltage level (Fig. 4.9c).
In a conventional LVTSCR, after the ESD triggering, the balance between injec-
tion from the emitter and the source junctions creates a quasi-neutral electron–hole
plasma region that produces a low-voltage drop in the on state. The charge balance
of the injected carriers determines the voltage drop. Therefore, by changing this
balance in high injection conditions, an impact is expected on the holding voltage.
To implement this physical principle, the LVTSCR emitter has been designed
with a separated contact. This emitter contact was connected not directly to the
protected pad, but to a string of forward-biased diodes instead. In this case, the
holding voltage decrease in SCR mode is limited by closing the emitter junction
below the reference voltage drop on the emitter diodes. Consequently, this effect
results in a shift in the balance between injected carriers and a change in the level
of the carrier space charge neutralization, increase of the electric field and, finally,
creation of a higher source-drain voltage drop.
Thus, an intermediate, mixed regime can be obtained with a holding voltage
range between double injection and avalanche–injection. In general, selecting the
proper number of diode structures in the emitter circuit can control the holding
voltage.
In [81], has been shown that even at the same holding voltage as that of snap-
back NMOS, the high holding voltage LVSCR is still several times more efficient
due to its low on-state resistance (Fig. 4.9c). This makes the diode component a
176 4 ESD Clamps
HBMPULSE VDD
SCR
D1
D2
DRAIN(n+)
EMITTER(P+)
Pwell Nwell
Nwell/n+
Sub
Pwell
SOURCE(n+)
RG
a) b)
8 – ORIGINAL LVTSCR
– ONE EMITTER DIODE
7
– TWO EMITTER DIODES
TOTAL TLP CURRENT (A)
0
0 2 4 6 8 10 12 14
10–2
LEAK. CUR. (A)
10–3
10–4
10–5
10–6
10–7
10–8
10–9
10–10
0 2 4 6 8 10 12 14
VOLTAGE (V) c)
Fig. 4.9 Cross-section of the LVTSCR drawn using a physical process simulation (metallurgical
junctions are shown) (a), its simplified equivalent circuit (b), and calculated isothermal triggering
I–V characteristic (c). Experimental TLP data for different LVTSCR design with 100 μm total
contact width, presenting the reference LVTSCR structure with a common emitter/drain contact,
LVTSCR with one and two diodes in the emitter line inside the N-well (c)
very valuable protection solution for different system applications, including charge
cable specified interface applications.
4.2.6 Triggering Characteristics Control in SCR Clamps

In addition to holding voltage control, another important clamp parameter is the
triggering voltage. This section presents the most advanced method to provide such
control, based upon the mixed device-circuit concept. In this section, the approach
is applied to a generic SCR p–n–p–n clamp, which is based on devices similar
to corresponding structures discussed in Chapter 2. The design of such a device

assumes the presence of two emitters, p+ and n+, isolated by corresponding n-base
and p-base regions. Both bases form the blocking p–n junction that provides the
high-voltage tolerance of the device. The bases have p+ and n+ contact regions with
corresponding control electrodes.
4.2.6.1 Mixed Device-Circuit Concept

The mixed device-circuit ESD solution has been first proposed in [85]. Similar prin-
ciple has been used in recent study [84]. It is based upon a different principle than
conventional ESD clamps. Instead of solving the problem of targeting the ESD pro-
tection window in both the voltage and rise-time domains, the mixed device-circuit
approach provides for different triggering characteristics of the device under ESD
and normal operating conditions.
The snapback ESD device is first modified to provide access to a terminal that
the active circuit will control. The snapback device is required to have (i) a very
low triggering voltage in the case of a positive bias at, or forced current through,
the control electrode, and (ii) a triggering voltage that is higher than any expected
voltage in the case of a grounded control electrode. Requirements for the active
control circuit are as follows: (i) to provide control electrode biasing or current
forcing during the first part of an ESD event (∼20 ns), in order to turn-on the ESD
device; (ii) to ground the control electrode after the ESD device is operating in
snapback mode (after ∼100 ns); (iii) always turn-off (ground) the control electrode
if VDD (∼5 V) is generated.
4.2.6.2 Practical Implementation of the Concept

for the Case of 130 nm Process
If voltage tolerance above the limitations of the NMOS gate oxide is required, an
SCR can be used, thus excluding the gate oxide region. However, in this case, the
triggering voltage is rather high, especially in CMOS processes of 0.13 μm and
below.
With the scaling of CMOS processes down to 60–45 nm the voltage tolerance
becomes rather low. One of the logical solutions is to use stacked forward-biased
diodes to protect the I/O pins. However, the high temperature dependence and high-
frequency I/O pin requirements for low-capacitance ESD protection circuits demand
a more advanced approach.
For applications in which the dual-diode protection circuit cannot be used, SCR-
based solutions [86] provide the highest level of protection per unit capacitance.
Since the blocking junctions in the SCR device have a rather high breakdown
voltage of ∼5–7 V, trigger circuits are needed to turn on SCR’s quickly and at
sufficiently low voltages.
The diode-triggered SCR (DTSCR) [87] (Fig. 4.10a) and grounded gate NMOS-
triggered SCR (GGSCR) [38] (Fig. 4.10b) are suitable for protecting low-voltage
I/Os. The DTSCR is favored over the grounded gate SCR GGSCR because its
178 4 ESD Clamps
IO Pad IO Pad
RNW
RNW
Q1
M1
Q2 Q1
Q2
RPW
RPW
a) b)
Fig. 4.10 Circuit schematic of the DTSCR (a) and GGSCR (b)
trigger voltage, VT1 , is easily adjusted by altering the number of diodes in the trig-
ger circuit. Additionally, in advanced technology nodes, special process steps, e.g.,
silicide-block, may be required for the GGNMOS embedded in the GGSCR. The
downside of the DTSCR is its significantly higher leakage current.
A different approach has been proposed in [88], featuring the dual-base triggered
SCR. The proposed solution combined the best features of the GGSCR [38] and the
DTSCR [87]; it has low off-state leakage current and an easily adjustable trigger
voltage. The turn-on time was virtually identical to that of an optimized DTSCR
designed in the same technology.
A major step in the philosophy of new design was to use the informative VDD
signal produced by the internal circuit, this approach was proposed in [85] and elab-
orated further in [89, 90]. Although this solution hierarchically partly belongs to the
ESD protection network level presented in Chapter 5, it will be discussed in this
chapter to allow comparative analysis.
The core VDD is used as a control signal to prevent triggering and reduce leakage
current while the chip is powered, this is a successful application of the high-
voltage ESD protection concept introduced in [85] to a low-voltage ESD protection
scenario.
The trigger circuit permits one to design for low VT1 values during ESD without
compromising resilience against false triggering under normal operating conditions,
because the circuit’s VT1 automatically increases when the chip is powered up.
4.2.6.3 Principle of Dual-Base Control Operation

The dual-base triggered SCR clamp is shown in Fig. 4.11. It consists of an
11-transistor trigger circuit and an SCR [91]. The trigger circuit is composed of
a control circuit (M0–M3), a voltage divider (M4 and R0), and a chain of three
Inverter Inverter Inverter I/O Pad

1 2 3
M1 M3 M9 M6 M7
M4
M0 M2 M10 M5 M8
VDD R0
VSS
Fig. 4.11 Circuit schematic for dual-base triggered SCR clamp
inverters (M5–M10). The trigger circuit consumes a relatively small layout area as
all of its transistors are nearly of minimum size except for M6 and M8. Thick-oxide
I/O transistors are used throughout the circuit to reduce leakage current and increase
the ESD robustness of the trigger circuit.
The Pwell of the SCR is connected to the output of the second inverter and its
Nwell is connected to the output of the third inverter. These connections and the
resulting dual-base triggering action are similar to that of the RC-triggered SCR
[92]. However, the triggering voltage of the RC-triggered SCR is not adjustable,
unless significant design changes are to be made to the trigger circuit.
Furthermore, the dual-base triggered SCR uses VDD as a control signal to pre-
vent triggering and reduce leakage current while the chip is powered. In principle,
any power supply bus can provide the control signal; however, application-specific
considerations (e.g., power-on sequence, bus routing) dictate whether core VDD or
I/O VDD should be used.
While the chip is powered, VDD is high and M0 pulls the gate of M4 down to
ground. Consequently, the output of the first inverter is high, the output of the second
inverter is low, and the output of the third inverter is high. The SCR N-well is pulled
up to the I/O pad voltage and the P-well is pulled down to ground, thereby shorting
out the base–emitter junctions of the PNP and NPN transistors. This keeps the SCR
in the off state and minimizes its leakage current.
During ESD conditions, VDD may be closely coupled to VSS, depending on
the topology of the ESD protection network and the amount of decoupling capaci-
tance. The design presented here is targeted for applications in which VDD is low
during ESD.
The voltage at the I/O pad increases during the rising edge of an ESD pulse,
causing the PMOS transistors M1 and M3 to begin pulling up the nodes connected to
their drains. The drain of M1 is connected to the gate of M3 and vice versa, forming
a feedback loop with contention during the initial stage of the turn-on process. M1
is larger than M3 so that it pulls up its drain node more quickly. Once M1 raises the
180 4 ESD Clamps
voltage at the gate of M2 above the NMOS threshold voltage VT , M2 turns on and
forms a positive feedback loop with M1. The gates of M2, M3, and M4 are pulled
up to the pad voltage by M1 and the voltage at the gate of M1 is pulled down to
ground by M2.
Once the voltage at the gate of NMOS transistor M4 increases above VT , M4
turns on and forms a voltage divider with the resistor R0. When the output of the
voltage divider reaches the switching threshold of the first inverter, the first inverter
will switch to low, the second inverter will switch to high and pull up the P-well of
the SCR, and the third inverter will switch to low and pull down the N-well. Current
is injected into the N-well and P-well, turning on both BJT’s simultaneously and
triggering the SCR. The trigger voltage is highly dependent on the voltage divider
formed by M4 and R0; VT1 can be increased by increasing the channel resistance
of M4 (RM4 ) and/or by decreasing the value of R0. In practice, reducing its channel
width WM4 increases the channel resistance of M4.
The voltage divider’s effect on the trigger voltage is two-fold. First, as the ratio
RM4 /R0 increases, the output of the voltage divider decreases and the I/O pad voltage
must be raised higher so that the divider output can reach the switching threshold
of inverter 1. However, as the pad voltage increases, the switching threshold of each
inverter also rises. The dependence of the switching threshold of inverter 1 on the
I/O pad voltage increases the sensitivity of the trigger voltage to changes in the
voltage divider.
4.2.6.4 Pulsed Characterization of Dual-Base Control (DBC) Clamp

In [88], circuit simulations were used to illustrate the effect of the voltage divider
on the trigger voltage of the protection device for 90 nm logic CMOS technology,
this section presents experimental results.
Dual-base triggered SCR’s with an SCR width of 50 μm were fabricated in a
0.13 μm BiCMOS technology (gate oxide thickness = 20 A) with a core VDD
of 1.2 V and an I/O voltage of 2.5 V. Devices were characterized with a Kelvin
transmission line pulse (TLP) system using 100 ns pulses [93].
During an ESD event, the chip is unpowered and thus VDD was left floating
during this set of TLP tests. Figure 4.12 shows the pulsed I–V characteristics of
three different dual-base triggered SCR’s. The pulse rise time was 10 ns.
The three SCR’s are identical, but their trigger circuits contain different sizes of
WM4 and R0, thereby altering the voltage division ratio between M4 and R0 and,
consequently, the trigger voltage of the SCR. VT1 values of 3.1, 3.7, and 4.2 V
were obtained. These circuits were designed to protect a 2.5 V I/O pad. They were
designed to have a VT1 greater than the I/O voltage of 2.5 V, because they were
designed for an application where a power-on sequence cannot be assured. For other
applications, the trigger voltage could be varied over a wider range.
The core VDD line is used as a control signal to reduce leakage and increase the
trigger voltage when the chip is powered on. The trigger voltage, VT1 , was measured
under power-on and unpowered conditions using pulses with 2 and 10 ns rise times.
Three different designs were characterized. In the three designs, the transistor sizing
2.5
Current (A)
2 Low Vt1
Medium Vt1
1.5 High Vt1
1
Low Medium High
Vt1 Vt1 Vt1
0.5
0
0 1 2 3 4 5 6
Voltage (V)
Fig. 4.12 Pulsed I–V of three dual-base triggered SCR’s designed to have different trigger voltages
Table 4.1 Trigger voltages of three designs with varying inverter sizing ratios
Design 1 Design 2 Design 3
WM6 /WM5 13.6 27.3 41

WM8 /WM7 11.4 22.7 34.2
VT1 (VDD=0, tR = 10 ns) 3.13 V 3.74 V 3.34 V
VT1 (VDD=0, tR = 2 ns) 3.13 V 3.76 V 3.56 V
VT1 (VDD=2.5 V, tR = 10 ns) 7.18 V 7.18 V 9.91 V
VT1 (powered, tR = 2 ns) 7.00 V 5.71 V 4.00 V
ratios for inverters 2 and 3 were varied. In all designs, WM6 > WM5 and WM8 > WM7 ,
which promotes quick triggering action.
Summary of the experimental validation for different design parameters of the
control circuit presented in Table 4.1 for the different conditions on VDD node. The
data presented for two different rise time parameters of the TLP pulse tR . VT1 is vir-
tually independent of the pulse rise time when VDD is disconnected. Furthermore,
VT1 is always higher when VDD is powered on than when it is unpowered.
However, the value of VT1 under power-on conditions is affected by the pulse rise
time. This is a dV/dt triggering effect that can be mitigated with proper design. The
magnitude of the dV/dt triggering effect is dependent on the transistor sizing ratios
in the second and third inverters, WM6 /WM5 and WM8 /WM7 .
4.2.6.5 DC Leakage of DBC Clamp

One of the most critical advantages offered by the solution is the low leakage. To
minimize power consumption, small DC leakage is desirable for ESD protection
circuits.
Figure 4.13 shows DC I–V curves for the second design of the dual-base trig-
gered SCR, with currents normalized per unit SCR width (A/μm) measured at three
182 4 ESD Clamps
1.E-08 25C
1.E-04 3-String DTSCR
75C 1.E-05 4-String DTSCR
1.E-09 125C 5-String DTSCR
1.E-06 Dual-Base Triggered SCR
Current (A/µm)
Current (A/µm)
1.E-10
1.E-07
1.E-11 1.E-08
1.E-09
1.E-12
1.E-10
1.E-13 1.E-11
1.E-14 1.E-12
0 0.5 1 1.5 2 2.5 3 0 0.5 1 1.5 2 2.5 3
Voltage (V) Voltage (V)
a) b)
Fig. 4.13 Normalized DC leakage currents for second dual-base triggered SCR design (a) and
normalized DC leakage currents for three DTSCR’s and a dual-base triggered SCR at a temperature
of 125◦ C (b) [88]
temperature points. When VDD is supplied, as would be the case under normal
operating conditions, the dual-base SCR exhibits flat, low-leakage characteristics
throughout the tested range (0–2.5 V). At 125◦ C, the leakage current is equal to 0.2
nA/μm at a pad voltage of 2.5 V.
Figure 4.13b compares normalized DC I–V curves for three DTSCR’s with vary-
ing diode string lengths fabricated in the same technology with that of a dual-base
triggered SCR at a temperature of 125◦ C. At elevated temperatures, the DC leak-
age of the dual-base triggered SCR is lower by orders of magnitude than that of
the DTSCR’s. At a pad voltage of 2.5 V, the normalized leakage of the dual-base
triggered SCR is 200 times lower than that of even the 5-string DTSCR with a much
higher triggering voltage VT1 .
4.3 Voltage and Current Reference in ESD Clamp

Dependence of the ESD clamp performance on the ESD pulse timing characteristics
is not always a desirable feature. In case of high capacitive or current load at the pin,
the rise time characteristics of the ESD pulse can be significantly altered.
To avoid load-dependent clamp operation, the avalanche breakdown voltage ref-
erence can be implemented. The implementation is based on the fast avalanche
multiplication process due to a practically local dependence of the multiplication
coefficient upon the electric field in Si material. If the process permits change in
the blocking junction using the non-self-aligned approach or by additional available
implants, then the internal blocking junction can be altered in the snapback device.
This measure may provide a desirable effect if the dynamic coupling of the junctions
and control electrodes is suppressed on the device level. However, this is not always
possible, especially in the case of high-voltage devices, due to a conflict between the
dimensions of the required drift regions, minimum process design rules and mask
alignment tolerance.
4.3 Voltage and Current Reference in ESD Clamp 183
Therefore, in the opposite case, additional voltage reference components can be

added to the clamp to provide the desired turn-on level. In this case, understanding
snapback ESD devices as three or more terminal components is important. Practical
value of the corresponding measurements comes from the dependence of the trig-
gering voltage of the ESD device for different conditions on the control electrode.
This information is further used to design appropriate triggering circuit components.
Depending on the type of ESD device, the voltage reference can be implemented
relative to the highest potential of the clamp (protected pad) or the lowest potential
of the clamp (ground).
The most typical voltage reference component is an avalanche diode. In the case
of field electrodes, for example, the polygate in a CMOS structure, the avalanche
diode will provide a voltage drop at the breakdown voltage level. In the case of
junction-based control electrodes, the avalanche current is used to inject carriers
into the electrode.
4.3.1 Low-Voltage Clamps in BiCMOS process technology

Clamp design with a high-side avalanche diode requires an isolated diode structure,
which can be used with both the N-type MOS device and NPN BJT ESD devices
(Fig. 4.14a, b, c).
ESDP ESDP ESDP ESDP
D D C
PE C
G B G B B PE
B
S S E
RG RG RB RB E
ESDM ESDM ESDM ESDM

a) b) c) d)
ESDP ESDP ESDP ESDP

PE
RG RG
D RB E NB
D
B
G B G B
NE NE
S S
C PB
NE
ESDM
ESDM ESDM ESDM
e) f) g) h)
Fig. 4.14 Typical schematics for high-side (a–d) and low-side (e–g) breakdown voltage refer-
ence clamp implementation using avalanche diodes for the case of NMOS and NLDMOS (a);
NLDMOS-SCR (b), NPN (c), BSCR (d), PMOS or PLDMOS (e), and PLDMOS-SCR (f), PBSCR
(g) and the SCR with low and high-side double reference diodes (h)
184 4 ESD Clamps
In the case of the P-type device, the low-side avalanche diode can have a p-region
that is not isolated from the grounded substrate (Fig. 4.14e, f).
In the case of the SCR device, a similar approach can be applied through decou-
pling of the SCR blocking junction base region contacts from the injection regions
(Fig. 4.14d, g, h).
The principle of operation of the clamp is based upon utilization of the inverse
dependence of the triggering voltage upon either the voltage of the field control
electrode or the current through the base electrode. In practice, this condition is
satisfied rather often (Chapter 3), but not always. Therefore, in order to properly
engineer the clamp, a three-terminal characterization of the device (similar to pulsed
SOA) is rather important. The purpose of such characterization is to collect infor-
mation about the instability boundary of the device in order to calculate the clamp
parameters.
In the case of CMOS electrodes, this calculation is simple. The current path
can be assumed through the clamp resistor during the most part of ESD. Thus, the
desired voltage drop on the resistor can be easily estimated. In the case of base junc-
tion structures, the total current can be roughly calculated, assuming a base potential
of ∼0.7 V.
Experimental results for implementation of the avalanche diode reference
(Fig. 4.15a) in the NMOS and PMOS clamps discussed above are presented in
Fig. 4.15c, d. This design approach provides the lowering of the triggering volt-
age to a level that corresponds to the sum of the reference voltage and the threshold
voltage of CMOS devices. The results provide a practical benefit in adjusting the
ESD protection window for these 5 V CMOS devices much closer to the absolute
maximum limit of ∼6 V.
In principle, any other active device or a stack of components can be used to
produce the voltage or current reference for the control electrode of the ESD devices.
In the previous section, GGSCR and DTSCR have already been mentioned. The
BVCEO-referenced BJT clamp includes a small reference NPN with an open base
that is the source of the base current for the main snapback ESD NPN device in their
common emitter circuit.
Finally, the avalanche diode can form a useful clamp by itself, as in the example
of the avalanche diode I/O and power clamp is presented in Fig. 4.16. However, the
capability of such a solution is significantly limited by the ∼0.1 mA/μm current
density that can be obtained in avalanche breakdown conditions.
This limitation is especially severe in high-voltage devices, where the lengths of
the blocking junction regions result in additional current reduction. An example of
the avalanche diode clamp with a 25 V voltage tolerance is presented in Fig. 4.17.
While the clamp itself can withstand relatively high current, the useful voltage ref-
erence and clamping can only be achieved in a rather narrow voltage range with a
current of 1–3 mA.
These results, however, are not negative. The major practical application of the
avalanche diode as a clamp belongs in two-stage protection circuits. This approach
will be discussed in Chapter 5.
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
0.06
D1
Gnd (Vleakage 5V)
0.05
Pulsed Current (A)

0.04
0.03
0.02
0.01
0
0 2 4 6 8 10 12 14 16
Pulsed Voltage (V) a)
Leakage Current (A) Leakage Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02 1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
4 5
4.5
Pulsed Current (A)
4
Pulsed Current (A)
3
3.5 D5
C18
3 Gnd (Vleakage 5V)
2 2.5 Gnd (Vleakage 5V)
D2
C14
Gnd (Vleakage 5V)
2
Gnd (Vleakage 5V)
1.5
1
1
0.5
0 0
0 2 4 6 8 10 12 0 1 2 3 4 5 6 7 8 9 10
Pulsed Voltage (V) b) Pulsed Voltage (V) c)
Fig. 4.15 I–V characteristics of the avalanche diode (a) and comparison of the triggering char-
acteristics for the transient triggered and voltage referenced NMOS (b) and PMOS (c) snapback
clamps
ESDP
I/O
ESDM
Fig. 4.16 Example of utilization of the avalanche diode as an input and power clamp
4.3.2 NPN Clamps with Voltage Reference

This section compares NPN BJT clamps with different triggering techniques. The
overall purpose of this particular design is to implement an 8 V ESD protection in a
low-cost isolated junction BJT process [94, 95].
Comparative analysis between transient triggered and different voltage refer-
enced ESD power clamps is presented in [96]. Several architectures of ESD clamps
186 4 ESD Clamps
a)
Cathode to Anode Pulsed Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02

0.04
0.03
V TLP
I leak (at 20 V)
0.02
0.01
0.00
0 5 10 15 20 25 30 35 40 45 50
Cathode to Anode Pulsed Voltage (V) b)
Fig. 4.17 Example of the layout (a) and pulsed I–V characteristics (b) for 25 V reference voltage
lateral avalanche diode
a) b)
Fig. 4.18 Negative and positive base current directions in the BVCER (a) and enhanced avalanche
(b) BJT ESD clamps
with both external and internal breakdown voltage reference techniques are com-
pared. In particular, the grounded base or BVCER BJT clamp (Fig. 4.18a) is
compared to an enhanced Zener clamp (Fig. 4.18b). In spite of a similar com-
mon emitter circuit, the operation of these clamps is different, relying on different
direction of the base current (Fig. 4.18).
The turn-on voltage for conductivity modulation mode of the first clamp is gener-
ally based upon either the breakdown of the internal collector-base blocking junction
or the transient dV/dt effect related to the displacement current in the collector-base
junction.
The second clamp is specifically designed to be less sensitive to the dV/dt effect
due to the fast avalanche breakdown in a relatively small avalanche reference diode
in the base–collector circuit. Thus, turn-on of the BJT into conductivity modulation
mode occurs at negative and positive base currents for the first and the second cases,
respectively.
The avalanche breakdown voltage reference can be achieved by using avalanche
diodes formed by the PBASE-NEMITTER and PBASE-NSINKER lateral junctions
(Fig. 4.19).
a) b)
@ 2007 IEEE
DEVICE TYPE (SIZE)
4 4
PBASE-NEMITTER
17F×70×3.5 um
3 PBASE-NSINKER 3
2F×70×7 um
2 2
1 1
0 10 20 30 10
–12 –11
10 10 10
–10 –9
COLLECTOR-EMITTER VOLTAGE (V) LEAKAGE (A)
Fig. 4.19 Simplified cross-sections and experimental TLP characteristics for the PBASE-NEMIT
(a) and PBASE-NSINK (b) lateral avalanche diodes [96]
Similar experiments have been performed with ESD clamps using bipolar SCR
(BSCR) [7, 97] ESD devices.
Another version of the ESD clamp device has been designed with the inter-
nal avalanche diode. This internal junction is formed by stretching the PBASE
diffusion to create an overlap with the collector region, thus forming a surface junc-
tion between the base and the N-sinker regions (Fig. 4.20). Similarly, the internal
avalanche diode structure is implemented in the BSCR, where an additional floating
N-EMIT region is added between the PBASE and the P-EMITTER.
Comparison of TLP characteristics for different NPN BJT clamps is presented
in Fig. 4.20. In the case of the clamp with external avalanche reference, the
188 4 ESD Clamps
a) c)
© 2007 IEEE © 2007 IEEE
ANODE (COLLECTOR) CURRENT (A)

4 ESD CLAMP 4
5 6

DESIGN ESD CLAMP DESIGN
4 5
3 3
4
3
2 2 3
2Fx50x3.5um
2
2
BVCER
1 1
T
1
1
0 0 0
0 5 10 15 20 25 30 0 5 10 15 20 25 30 10–12 10–1110–10
10 –12 10–1110–10
b) ANODE (COLL.)-EMITTER VOLTAGE (V) LEAKAGE (A) d)
COLLECTOR-EMITTER VOLTAGE (V) LEAKAGE (A)
Fig. 4.20 Avalanche breakdown voltage reference effect in NPN BJT and BSCR clamps.
Cross-sections for NPN (a) and BSCR (c) with internal voltage reference; measured TLP I–V
characteristics for the BVCER, enhanced Zener clamp and internal avalanche NPN clamp (b) and
for internal Zener NPN and BSCR clamps (d) [96]
breakdown voltage of the device is significantly reduced, the critical current required
for snapback is rather large, and device operation is not optimal in terms of provid-
ing the proper waveform in the desired voltage range (8–10 V for the particular
application). In contrast, the device with the internal blocking junction provides
excellent triggering characteristics that provide both the low triggering current and
the low holding voltage.
Similar snapback characteristics, but with better high-current tolerance, were
observed in the BVCER and internal Zener BSCR clamps (Fig. 4.20). Also, in the
case of the base and N-sinker overlap, the corresponding blocking junction was
formed with a lower breakdown voltage of ∼7 V.
In spite of the similarities in design and operation, the internal and the external
voltage reference clamps have quite different triggering mechanisms. These mecha-
nisms are defined by the direction of the base current and the interplay between the
avalanche and the injection effects. These effects were already discussed in Chapters
2 and 3.
It should also be taken into account that from the structural point of view, the
NPN BJT is composed of vertical and lateral NPN devices. At lower current values,
the vertical BJT is involved in a conductivity modulation mechanism, thus providing
the initial part of the snapback characteristic. The lateral surface NPN is engaged
after subsequent current increase.
4.4 High-Voltage ESD Devices
As was mentioned before in this book, a conventional submicron process subdivi-

sion between low and high-voltage pins is assumed at the level of ∼12 V. In this
4.4 High-Voltage ESD Devices 189
section, clamp design specifics for high-voltage applications are discussed for the
BiCMOS and extended MOS processes in the 20–100 V range.
4.4.1 20 V NPN with Blocking Junction Internal Reference

An example of the 20 V NPN clamp with an internal voltage reference is presented
in Fig. 4.21. In this cell, a low-voltage Nwell implant has been used to reduce the
breakdown and triggering voltage from the original standard device level of ∼35 V
down to the 28 V. This margin is appropriate for pins with an operation voltage of
20 V, taking into account the temperature dependence and the statistical deviation
of standard device SOA parameters.
The clamp has practically no dV/dt sensitivity in the 2–10 ns rise time domain.
Double Deep
ESDP Trench P-
Guard Ring
ESDM Multifinger
Snapback
BJT
Polyresistor
Reverse Current Path BJT Diode
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
8
7
6
TLP
5 I leak (at 20 V)
4
3
2
1
0
0 10 20 30 40
Fig. 4.21 High-voltage avalanche breakdown voltage reference NPN BJT clamp layout and
schematic view and measured TLP characteristics
190 4 ESD Clamps
To enable automatic voltage tolerance for the high-voltage reverse-path diode

component, an additional shorted base NPN structure is used. In this case, the
collector–emitter diode provides the reverse pulse protection in negative ESD pulse.
An alternative way to build a more compact cell is to short some base–emitter junc-
tions in a few selected fingers of the snapback BJT. In this case, it is expected that
the fingers with shorted base, dedicated to reverse current path protection, will also
contribute into the total collector current in the snapback mode.
One of the important considerations of clamp design is to take into account the
high-voltage isolation of and interaction with the guard ring structure.
4.4.2 NPN Clamp with External Lateral Avalanche Diode

Reference
An alternative example of the external avalanche diode 20 V NPN clamp imple-

mented in the same 20 V BCD process with external voltage reference is presented
in Fig. 4.22.
In this case, the reference avalanche diode component is the same as the one
Similarly to the clamp in Fig. 4.21, this clamp also has practically no dV/dt
sensitivity in the 2–10 ns rise time domain.
4.4.3 SCR-Based High-Voltage Clamp
An alternative clamp can be created using SCR-type ESD devices (Fig. 4.23). For
example, a high-voltage BSCR can be obtained from the NPN device. Similarly, an
NLDMOS-SCR can be obtained by a corresponding transformation of the standard
20 V NLDMOS supported in the process. Both structures require embedding of the
p+-emitter region in order to form a parasitic p–n–p structure, in addition to the
n–p–n structure already present in the original device (Chapter 3).
The parasitic PNP can be formed between the p+ guard ring, p+ -emitter, and
NDRIFT or collector epi isolation (acting as a base).
A similar clamp can be composed for the NDeMOS-SCR implemented in high-
voltage extended-drain devices.
4.4.4 Lateral LPNP Clamp

High holding voltage protection using lateral PNP structures has already been
discussed in this chapter. The implementation of the clamp mainly targets the
high-density layout to achieve maximum possible total width of the device
(Fig. 4.24). For this purpose, base contacts are assembled using the minimum
diffusion region at the finger side.
Reference Avalanche Diode

ESDP
Multifinger
Snapback
BJT
Reverse
Current Path
BJT
Polyresistor ESDM
Double Deep Trench P-Guard Ring
a) b)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
4 V TLP
I leak (at 20 V)
3
0
0 5 10 15 20 25
ESDP to ESDM Pulsed Voltage (V) c)
Fig. 4.22 High-voltage NPN BJT clamp with collector-emitter avalanche diode reference: layout
view with clamp components (a), schematic view (b), and measured TLP characteristics (c)
One of the specifics for clamp implementation is the n-base connection. The n-
base connection is necessary to eliminate excessive base-collector coupling that can
be undesirable for pins with high transient voltages.
4.4.5 Mixed Device-Circuit Dual Mode Solutions

The basic principle of the mixed-mode device-circuit solution has already been
demonstrated for the dual-base-controlled SCR in Section 4.2.5.
In this section, a similar approach is illustrated for the high-voltage case, using
the example of a 50 V clamp [85, 89, 90].
192 4 ESD Clamps
Reference Avalanche Diode

Multifinger
ESDP BSCR
Reverse
Current Path
Diode
ESDM
Polyresistor Double Deep Trench P-Guard Ring
a) b)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
9
8
7
6
5 V TLP
I leak (at 20 V)
4
3
2
1
0
0 5 10 15 20 25 30
Fig. 4.23 Example of the 20 V NLDMOS-SCR layout, schematic view, and TLP characteristics
The trend in improving switching capabilities of high-voltage devices usually

results in a rather small or even negative ESD protection window, as a side effect of
the technology improvement.
Thus, application of both voltage referenced and transient triggered clamps
becomes undesirable and rather challenging, from the point of view of achieving the
required parameters. In this section, an alternative way to implement ESD protection
is discussed with the purpose of demonstrating the physical concept of using ESD
devices as active structures with snapback characteristics that are actively controlled
by the circuit driver.
4.4.5.1 Example of Circuit Design

A simplified schematic of the active control circuit is presented in Fig. 4.25. The
circuit contains a driver that controls the gate of the snapback device MESD and
Polyresistor
ESDP
Multifinger ESDP
BSCR
RB E
Double Deep
Trench P- B
and N-Guard
Rings C
ESDM ESDM
a) b)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
1.6
1.4
1.2
1.0
V TLP
0.8
I leak (at 20 V)
0.6
0.4
0.2
0.0
0 5 10 15 20 25 30
Fig. 4.24 Example of the 20 V lateral PNP layout, schematic view, and TLP characteristics for
the NLDMOS-SCR
VIN
MP2
MP1
MN3 MN1
VDD MESD
R2 R3
C1 MN2
MN4 C2
\GN © 2007 IEEE
Fig. 4.25 Active driver circuit with snapback SCR device (MESD) controlled from the gate
terminal [85, 89]
194 4 ESD Clamps
operates depending on the signal from the VIN and VDD pins. This circuit has been
studied using circuit simulation.
In the case of a positive ESD event on the VIN pin, MP2 will be on until the
capacitor C1 is fully charged (∼20 ns). During this time, MP2 and MN3 directly
apply a positive bias on the MESD control electrode (gate). This creates the condi-
tions for fast triggering of the MESD into snapback at rather low VIN pad voltages.
Significantly, the circuit does not only depend on the capacitive coupling to pull the
gate high, as in some alternative designs. MP2 also turns on MN1, which sources
current for a turn-off RC-circuit composed of resistor R2 and capacitor C2. After a
corresponding RC-delay of ∼200 ns, the transistors MN2 and MN4 turn on. MN2
shorts the control electrode of the ESD device to ground, while MN4 turns off both
MN3 and MN1. MESD remains in high-current mode until the ESD discharge is
completed. The low-voltage avalanche diodes are used to limit the gate voltage of
the MOS devices. During an ESD event, the VDD potential is near ground due to
the high capacitance and leakage of internal circuitry of the voltage regulator.
In the case of normal operation, during the relatively slow (determined by the
total equivalent capacitance of the chip and the external devices) powering sequence
of VIN above ∼5 V, the VDD signal is generated in the internal circuitry; MN2 is
then turned on through R3. MN2 holds the control electrode low. This maximizes
the trigger voltage of MESD. The VDD signal will also turn on MN4. MN4 will
keep MN3 and MN1 off, regardless of VIN, eliminating the chance of contention
between MN3 and MN2. Holding these transistors off also provides for low leakage
in the clamp circuit.
This ESD solution has been experimentally validated in [89] for a 0.35 μm
CMOS process with 20 V extended drain MOS devices (DeMOS). For pulses with
a 10 ns rise-time, VABSMAX is 27 V. Given VMAX = 22 V, the protection win-
dow is 22–27 V (Fig. 4.26a). With a 200 ps rise time, the upper limit is lowered
to ∼23 V. The DeMOS-SCR lateral ESD protection device used is similar to the
NLDMOS-SCR device (Fig. 4.26a).
The TLP I–V characteristics for two different rise-times and under two differ-
ent operating conditions, at VDD=0 V (the results were the same as for a floating
VDD node) and at VDD=5 V, are presented in Fig. 4.26b. Under the condition
that corresponds to ESD stressing of the non-powered circuit (VDD=0 V), the trig-
gering voltage is very low (∼5 V). This guarantees reliable ESD protection not
only for high-voltage devices but also for the low-voltage devices in the internal
circuit that are connected to the protected pin. However, the leakage measured
at VIN=20 V remains the same as in a grounded gate ESD device (Fig. 4.26b,
left plot).
In the case that corresponds to normal operation (VDD=5 V), the trigger voltage
of the clamp is high (∼29 V) and equivalent to that of an NLDMOS-SCR device
with a small gate-source resistor. Evidently, the clamp may be designed to have a
triggering voltage that, under normal operation, exceeds VABSMAX . This is desir-
able, since a high trigger voltage correlates to unreliable snapback operation and
can result in damage of the devices. ESD operation of the clamp is fully controlled
1.0
GATE-SOURCE VOLTAGE (V)
Vabs max
Vmax
0.8
PULSED CURRENT (A) 0.6

5
0.4
4
3
0.2
2
0
0.0
0 10 20 30 40
PULSEDVOLTAGE (V) a)
10 1 © 2007 IEEE
VDD = 0 6
10 0 (ESD TEST)
VIN = 20V
TLP CURRENT (A)
TLP CURRENT (A)
10 –1
10ns VDD=5V 4
10 –2 (SWITCH OPERATION)
200ps
10 –3 200ps 2
10ns
10 –4
10 –5 0
0 5 10 15 20 25 30 10–2 10–1100 101 102
TLP VOLTAGE (V) LEAKAGE (nA) b)
Fig. 4.26 Pulsed output characteristics and ESD protection window for the 25 V absolute maxi-
mum voltage DeMOS device (a), and measured TLP results for the mixed device-circuit solution
clamp showing both modes of operation (b) [89]
by the channel current, rather than the avalanche breakdown current, thus always
providing reversible snapback operation.
It has been shown that both overshoot and turn-off voltages can be controlled
by this solution (Fig. 4.27). This example demonstrates a cross-disciplinary aspect
of future advanced physical ESD design. The ESD device, as a component of
the dual mode clamp, should be designed with a proper dc voltage tolerance in
order to provide low sensitivity to the fast rise time and, at the same time, afford
sufficient control of the triggering characteristics (reversible instability bound-
ary) by the control electrode, with reversible high-current operation and a small
footprint.
196 4 ESD Clamps
4kV 30
30
VOLTAGE (V)
VHBM
20
25
2kV
VOLTAGE (V) 10
20
0
1kV 0 1 2 3
15
TIME (µs)
0.5kV
10
0
–5 –4 –3 –2 0 0 1 2 3 4 5
TIME (ns)
CURRENT (A)
3.0
2.0
1.0
0.0
0 500 1000 1500
Fig. 4.27 Comparison of the measured voltage waveforms for the active gate control circuit [89]
4.5 The Concept of Self-Protection

Self-protection capabilities have been already mentioned indirectly in the discussion
of SOA for standard devices. In the most simplified case, if the devices connected to
the output pin are of adequate size, the pin will not require an additional clamp. In a
more general case, the self-protection capability can be achieved at different levels:
the device level, circuit level, or array level.
4.5.1 Device-Level Self-Protection

A number of changes can be made in the architecture of the standard device in order
to implement self-protection capabilities.
One of the most common examples of self-protecting solutions is the addition of
the drain ballasting region. With simultaneous increase in the well tap spacing, I/O
NMOS devices can reversibly withstand the snapback operation similarly to ESD
clamps. In normal operation mode, the side effects of this design are an increased
space on the chip, parasitic drain capacitance, and a higher on-state resistance.
4.5 The Concept of Self-Protection 197
p+ n+ p+ n+ p+ n+ p+ n+ p+ n+ p+ n+ p+
PO n+ – DRAIN
LY
SOURCE
DRAIN GA
TE
n+
n+ – SOURCE E
p+ RAT
ST
UB
ell/S
Pw
a) b)
HBM 2kV, W = 500um

12
LPG [m]
11 –0.1
DRAIN VOLTAGE (V)
10
9
0.2
8
7
0.5
6
0.8
5 1.3
4
10 –10 10 –9 10 –8 10 –7
TIME (s)
800
LAT. TEMP.(K)
600
400
c)
200
Fig. 4.28 Power array finger with interdigitated source/P-well diffusions (a), 3D equivalent struc-
ture (b), and the results of 3D numerical simulation analysis (c) of the waveforms of mixed-mode
HBM simulation upon the distance between the PPLUS region and the edge of polygate (LPG )
198 4 ESD Clamps
Another way to implement an additional self-protection capability or at least

push the current instability level to a higher voltage and current range is to use an
interdigitated well tap connection solution (Fig. 4.28a). In this case, the parasitic
NPN bipolar has the minimum possible parasitic base (bulk)–emitter (source). This
measure provides the lowest current gain in the internal parasitic bipolar structure
for NMOS devices. Since the avalanche–injection conductivity modulations occur
at a much higher critical avalanche current, both the instability boundary and the
ESD protection window are changed.
This solution is especially practical in fully silicided processes that provide a
reliable connection to the interdigitated regions with no special contact placement.
From the physical point of view, operation of this device is based upon reduction
of the hole density in the discharge region due to a reduced gain of the para-
sitic NPN BJT, and effective extraction of the carriers by the p+-contact regions.
Respectively, the average electric field amplitude under the gate in the triggered-on
state is increased, due to a lack of space charge compensation between the elec-
trons injected from the source junction and the holes generated in the avalanche
multiplication drain region. In particular, this results in an increase of the holding
voltage. TCAD-level validation for this solution is presented in Fig. 4.28b, c. The
major parameter for comparative analysis in this case is the distance LPG between
the introduced PPLUS region and the source side of the polygate region.
The mixed-mode signal waveforms (Fig. 4.28b) HBM ESD operation demon-
strates that the holding voltage realized for the structure can be even bigger than
the triggering voltage. Analysis of the 3D electron and hole current distributions
exhibits the conductivity modulation process and the holding voltage increase in
greater detail.
A similar solution has been proposed for high-voltage lateral DMOS devices,
where the P-buried layer is used to reduce the corresponding internal base–emitter
resistance and common base current gain of the parasitic lateral BJT, thus improving
the SOA of the device.
4.5.2 Array-Level Protection

A more sophisticated concept is proposed in [98], where self-protection capability
of the arrays is achieved by embedding the additional parasitic SCR structure inside
the array. In this case, an embedded ESD device is engaged to take over ESD current
conduction, before limitations in the parasitic NPN device of the standard array
occur due to avalanche–injection. Contrary to a conventional local clamp approach,
this concept provides the self-protection capability within the array itself, since the
area of the embedded ESD device also contributes to the on-state current during
normal operation.
This solution was experimentally demonstrated for an NLDMOS power array
in a 50 V BiCMOS process. Series of distributed p+-emitter regions and diffusion
regions have been embedded into some array fingers in order to form an additional
parasitic SCR structure with reversible snapback capabilities.
4.5 The Concept of Self-Protection 199
The self-protecting array (SPA) capabilities have been achieved by transforma-

tion of small parts of the NLDMOS array into an NDMOS-SCR that provides
current conduction during the ESD event. At normal operation, the NDMOS-SCR
portion of the device contributes toward the useful array current.
The equivalent circuit of the self-protecting array device is presented in
Fig. 4.29d, where the left part represents the fingers of the NLDMOS and the right
part represents the NLDMOS-SCR.
GND PAD GND PAD

G G
B S D B S P D
p+ n+ n+ p+ n+ n+ p+ n+
PBODY Nwell PBODY Nwell
N-epi N-epi
a) b)
Drain SCR-Drain
Gate n+
p+ VDD PAD
LDMOS-
NLDMOS
n+
SCR
n+ n+ p+ n+ p+
n+
Internal
p+ Circuit
n+
GND
c) d)
Fig. 4.29 Simplified cross-section for NLDMOS (a), local clamp NLDMOS-SCR devices (b),
and topology (c) for the self-protecting device with the left finger formed by a standard supported
device and right finger formed by the embedded SCR device; (d) simplified circuit diagrams for
the self-protecting array [98]
Simplified cross-sections of the NLDMOS and NLDMOS-SCR devices and the

constructed topology of the self-protecting device, based upon the principles listed
above, are presented in Fig. 4.29a, b. The topology is utilized by additional inter-
digitated P-emitter regions added into the drain composite regions of one or more
device fingers, depending on the array size (Fig. 4.29c).
These experimental SPA devices provide different pulsed I–V characteristics
when compared to standard control NLDMOS array devices (Fig. 4.30). Unlike
standard devices, the SPA’s are able to sustain a larger current level suitable for
reversible ESD protection. At the same time, major figures of merit associated with
the array’s normal on-state operation have been changed negligibly with acceptable
change in the pulse ESD SOA of the device itself (Fig. 4.31).
200 4 ESD Clamps
10 1 © 2006 IEEE 10 1
SP-NLDMOS

100 100
10–1 NL 10–1 NLDMOS

DM
SP O
S
-N
LD
10–2 MO 10–2
S
10–3 10–3
0 10 20 30 40 50 60 10 –1110–10 10–9 10–8 10–7 10–6
PULSED DRAIN-SOURCE VOLTAGE (V) LEAK.CURRENT (A)
Fig. 4.30 Comparison of TLP characteristics for the standard and self-protecting 50 V NLDMOS
array with a 400 μm width in the grounded gate configuration [98]
10 1 © 2006 IEEE
GATE BIAS (V) 10 1 © 2006 IEEE

GATE BIAS (V)
10 0
10 0
10 –1 5
3 10 –1 5
4
3
2
10 –2 10 –2
1
0 1
0
10 –3 10 –3
0 10 20 30 40 50 60 0 10 20 30 40 50 60
PULSED DRAIN-SOURCE VOLTAGE (V) PULSED DRAIN-SOURCE VOLTAGE (V)
a) b)
Fig. 4.31 Comparison of the pulsed SOA using output pulsed drain-source TLP characteristics
measured at different gate bias for standard NLDMOS (a) and SPA arrays (b) of similar footprints
[98]
This solution is essentially made possible by the implementation of the instability

boundary for NLDMOS-SCR below the pulsed SOA limits of the NLDMOS device.
This implementation principle generally requires a precise embedded SCR design,
where the snapback voltage is a much stronger function of gate bias. At the same
time, one of the objectives of the solution is to preserve the SOA of the whole array
at the original level.
4.6 ESD Protection of Ultra High Voltage Circuits

Evolution of certain power applications creates the necessity of integrating very-
and ultra-high-voltage (UHV) level components on the chip [99–101], up to the
4.6 ESD Protection of Ultra High Voltage Circuits 201
range of 500–1000 V. This voltage level is required for compatibility of the different
power sources with the power grid. Meantime the experience in such high-voltage
protection is very limited [102, 103].
Recently, an interest has emerged in implementing integrated grid-tolerant cir-
cuits. In this application, integrated circuit pins have to be tolerant to 700–1000 V,
with some galvanically isolated pins requiring additional system level protection
with the voltage tolerance level up to ∼5 kV. One of the typical applications is the
solar energy conversion. In this case, the dc voltage level generated by the solar
panels needs to be converted to the standard ac current of the grid. These actively
developing photovoltaic (PV) applications require circuits with voltage tolerance of
up to 800V at current levels of ∼1 A.
Other ultra high-voltage applications include factory automation, electric motor
control, and LED lighting.
An example of a photovoltaic system is presented in Fig. 4.32. The left part of the
circuit presents a DC/DC converter deriving input voltage from the PV cell in the
range of 23–38 V and producing a maximum output voltage of ∼400 V. The DC/DC
converter includes a high-frequency DC–AC converter block, a 1:10 transformer
providing the AC voltage increase up to 400 V, a diode rectifier, and an LC filter.
Then obtained high DC voltage is used by unfolding the DC/AC inverter to create
a grid-compatible AC output. The key components are four discrete devices used
in the T-bridge with a voltage tolerance over 700 V. The 700 V voltage tolerance
is required to achieve the 220 V root mean square (RMS) voltage level. It can be
assumed that future systems will require ESD protection of pins with such a high-
voltage tolerance.
Perhaps one of the first impressions is that UHV pins may not require ESD
protection at all, since the voltage tolerance is already rather high. Nevertheless,
according to a more accurate evaluation, ESD protection is still required, especially
at the system level.
In spite of their high-voltage tolerance levels, such pins may require ESD
protection.
Fig. 4.32 Power circuit of the photovoltaic inverter with UHV components
202 4 ESD Clamps
However, even in such “exotic” cases the basic principles still apply. The most
straightforward solution is the implementation of the LDMOS-SCR structure with
the p+-emitter isolated by the drain region.
An example of the layout of the 500 V device is presented in Fig. 4.33 with
the UHV pad located in the middle of the structure. The typical device topology
supports a significantly different scaling for the drift region length (∼20–50 μm) in
order to control the triggering voltage.
The measured device demonstrates reversible snapback behavior (no failure).
The device demonstrates switching into “SCR-mode” after turn-on. Experimental
evaluation of the UHV devices presents a serious challenge. It requires significant
modification of the high-voltage high-speed voltage probe for waveform evaluation.
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06
0.2 D1 VGS = 0V
0.18 D1 Vgs = 10V
Pulsed Current (A)
0.16
D1 Vgs = 20V
0.14
0.12 D1 Vgs = 36V
0.1 D1 Vgs = 0V
leakage
0.08
0.06
0.04
0.02
0
0 200 400 600 800 1000
Pulsed Voltage (V)
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06
0.2
N500 VGS = 0V
0.18
N500 Vgs = 10V
0.16
Pulsed Current (A)
N500 Vgs = 20V

0.14
N500 Vgs = 36V
0.12
N500 Vgs = 0V
0.1 leakage
0.08
0.06
0.04
0.02
0
0 200 400 600 800
Pulsed Voltage (V)
Fig. 4.33 Example of the 500 V device layout with middle pad, TLP characteristics of the standard
device, and ESD clamp at different gate bias levels
4.7 Summary 203
Another problem is related to a significant voltage and current overshoot

observed which increases with precharge level (Fig. 4.34).
1200 5
500 V 500 V
1000 1 kV 4 1 kV
Voltage [V]
800
Current [A]
3
600 1.25 kV 1.25 kV
2
400
200 1
0 0
–20 0 20 40 60 80 100 –20 0 20 40 60 80 100
Time [ns] Time [ns]
Fig. 4.34 Voltage and current waveforms for 0.5, 1, 2, and 3 kV for 400 V grounded gate
NLDMOS-SCR clamp
4.7 Summary
In this chapter, the knowledge of ESD devices and structures (Chapters 2 and 3)
has been applied to describe ESD clamps. Implementation of ESD devices, studied
through their cross-sections, into distributed layout objects on the semiconductor
wafer is not always a straightforward task.
Clamp implementation includes combination of a broad range of aspects: device
width scaling, lateral isolation, vertical isolation from the substrate, latch-up guard
rings and, of course, all the clamp components for voltage reference, and dynamic
coupling.
An emphasis is placed on the methods and possibilities used to achieve the
desired voltage waveforms for ESD pulse operation. This has been accomplished
in two ways: by the clamp circuit design and by the clamp topology design.
Features similar to the positive and negative feedback effects realized on the
device structure level can be engineered at the clamp level as well, in order to control
the waveforms produced by the clamp during ESD event.
One of the most effective clamp designs creates voltage and current reference
circuits using clamp components in the fast avalanche breakdown mode, as well as
more complex circuits for the control of clamp triggering-on and triggering-off.
Layout implementation of the ESD device is a very critical part of the clamp.
The topological degree of freedom is advantageous in engineering appropriate
characteristics of the clamp itself.
The complexity of the aspects involved in clamp design requires verification and
monitoring of the clamp characteristics using extensive test chip experimentation.
204 4 ESD Clamps

Example 4.1 Snapback NMOS Clamp Operation Analysis

Library Name: Examples4_Snapback_Clamps
Project Name: E4.1_5 V_Snapback_NMOS_HBM
This example explores basic operation and optimization of one of the most popular
ESD solutions – the snapback NMOS clamp. Comparative analysis of the HBM
pulse waveforms can be performed for a range of parameters of the Snapback
NMOS device, gate-source resistor R2 values and different HBM pulse ampli-
tudes. In particular, the increased gate-coupling effect due to the increase of gate
resistor (R2) value results in a reduction of the voltage overshoot (Fig. E4.1).
Reduction of the snapback device (M3) width results in an increase of the turn-
off time (Fig. E4.1). This example also demonstrates the change of the triggering
characteristics for different substrate-to-source spacing. It is also important to
C2
2.6e-12
L0 R1 R4
Charge Discharge
12.6e-6 1e8*(1-... 1.5K

C1
R3 ESDP
C0 1e-11
1e8*(1+... 1e-10
HBM Pulse Circuit
M3*
V0 $Vpulse...
Snapback NMOS D0*

Clamp W=50
W=400
R2*
1e4
ESDM
Fig. E4.1 Mixed-mode simulation circuit with a cross-section of the snapback NMOS device and
2 kV HBM pulse response for different gate-source resistor R2 values and sizes of the Snapback
NMOS device
compare the operation of the Snapback NMOS with that of other snapback
devices presented in this section, for example the LVTSCR devices presented in
Example 4.2.
Example 4.2 LVTSCR ESD Clamps

Project Name: E4.2a_NLVTSCR_HBM
Project Name: E4.2b_PLVTSCR_HBM
These mixed-mode simulation examples study a circuit with parameterized struc-

tures for N-type LVTSCR and P-type LVTSCR. The NLVTSCR example is
C2
2.6e-12
2.6e-12
R1 R3 L0 R1 Discharge R4
R3 L0 Discharge R4 Charge HBM_Pulse
Charge HBM_Pulse
12.6e-6 12.6e-6
1e8*(1+... 1e8*(1-... 1.5K C1 1e8*(1+... 1e8*(1-... 1.5K
C0 1e-11 C1
V0 $Vpulse...
1e-10 C0 1e-11
V0 $Vpulse...
1e-10
ESDM
ESDM
D2
LVTSCR0* W=100
R5*
10M
D0 R2* LVTSCR0*
W=100
W=100
R2* 1e4
1e4
W=100
Fig. E4.2 Mixed-mode simulation circuits with cross-sections of NLVTSCR and PLVTSCR
and mixed-mode simulation results of the HBM pulse response for different circuit and device
parameters
206 4 ESD Clamps
constructed with optional emitter biasing diodes to demonstrate the effect of the
holding voltage increase. The effect is used to eliminate the possibility of latch-up
by increasing the holding voltage of the SCR above the power supply level. The
effect of the diodes is introduced or eliminated by changing the value of the resistor
R5 in the circuit. A comparison of the HBM pulse waveforms for cases with and
without emitter biasing diodes is presented in Fig. E4.2.
Example 4.3 Two-Stage ESD Protection with Snapback NMOS

Project Name: E4.3_5 V_Two_stage_HNMOS_HBM
This example demonstrates the two-stage clamp ESD protection principle. The
example is composed of a grounded gate snapback NMOS clamp, second stage
C2
2.6e-12
L0 R1 R4
12.6e-6
1e8*(1-... 1.5K
C1
R3 C0 1e-11
1e8*(1+... 1e-10
R5*
INP_PAD
V0 $Vpulse... INP_INT
M0 1K
D1*
W=10
W=400
R2*
10K
Fig. E4.3 Mixed-mode simulation circuit and comparison of the 2 kV HBM waveforms at the
external and internal input nodes
resistor, and 7 V avalanche diode. Waveforms at different circuit nodes are

compared for different avalanche diode breakdown voltages and the second stage
resistor values (Fig. E4.3), demonstrating reduced overstress of the internal input
node that is connected to the avalanche diode. Voltage overshoot reduction is
important for CDM ESD pulse events.
Example 4.4 High-Voltage NLDMOS-SCR Clamp with High-Side

Avalanche Diode Reference

Project Name: E4.4_Snapback_HSCRZ_HBM
In this example, the effect of the high-side avalanche diode reference is demon-
strated using default components for 2 kV HBM pulse conditions. The connection
of the reference diode is enabled or disabled by changing the R5 resistor value.
C2
2.6e-12
L0 R1 R4
12.6e-6
1e8*(1-... 1.5K C1
R3 C0 1e-11
1e8*(1+... 1e-10
R5* ESDP
V0 $Vpulse...
D0*
W=50 M1
R2* W=200
1e4
ESDM
Fig. E4.4 Mixed-mode simulation circuit demonstrating the high-side reference effect in the
NLDMOS-SCR clamp and the change in the triggering and turn-off waveforms
208 4 ESD Clamps
Comparison of the HBM pulse waveforms demonstrates the effect of the diode on
both the triggering voltage and the residual voltage after turn-off in the case of
a high-impedance load (Fig. E4.4). Typical analysis includes optimization of the
clamp components for appropriate matching of the avalanche diode current, and
gate resistor and snapback device parameters to achieve low dV/dt triggering effect.
Example 4.5 PNP Clamp with Low Side Avalanche Diode

Reference

Project Name: E4.5_Snapback_PNP_HBM
In spite of a larger footprint, the PNP clamp is typically preferred for meeting high
holding voltage requirements. This example allows comparative analysis of the PNP
clamp with other high-voltage device clamps. The clamp is analyzed both with and
without the low-side avalanche diode reference under HBM ESD pulse conditions.
The low-side reference avalanche diode is enabled and disabled by varying the
2.6e-12
L0 R1 R4
12.6e-6
1e8*(1-... 1.5K C1
R3 C0 1e-11
1e8*(1+... 1e-10
V0 $Vpulse...
ESDP
R2* Q0*
R5* 1e4
1
W=1000
D0*
W=100
Fig. E4.5 Mixed-mode simulation of the low-side avalanche diode referenced clamp under 2 kV
HBM ESD pulse with different values of resistor R5. A 30 V compact model is used for the
avalanche diode in this example
value of resistor R5. Resistor R5 also adds on-state resistance to the ideal avalanche
breakdown characteristics produced by the compact diode model used for D0 in
this example. Comparison of the response to the HBM ESD pulse demonstrates the
impact of the low-side reference on both the triggering voltage and residual volt-
age after turn-off in case of a high-impedance load (Fig. E4.5). Further analysis
may include optimization of the clamp components for appropriate matching of the
avalanche diode and gate resistor to achieve a low dV/dt triggering effect.
Example 4.6 High-Voltage NPN Clamp

Project Name: E4.6_NPNZ_HBM
C2
2.6e-12
R3 L0 R1 Discharge R4
Charge HBM_Pulse
12.6e-6
1e8*(1+... 1e8*(1-... 1.5K
C1
C0 1e-11
V0 $Vpulse...
1e-10
R5* ESDM
1
D0*
W=50 Q0*
R2* W=400
1e4
Fig. E4.6 Mixed-mode simulation circuit for the NPN clamp with high-side avalanche reference
diode; voltage waveforms produced by the NPN BJT clamp under 2 kV HBM ESD pulse condi-
tions as a function of transistor width and base resistor values illustrating the effect of the high-side
diode reference
210 4 ESD Clamps
This simulation example of an ESD NPN BJT clamp studies clamp operation with
different high-voltage NPN parameters and clamp component values under a 2 kV
HBM ESD pulse. This clamp includes a high-side avalanche diode that can be
enabled or disabled by changing the value of resistor R5. The impact of circuit com-
ponent values on both the triggering voltage and the residual voltage after turn-off
in case of a high-impedance load is shown in Fig. E4.6. This example can be used
for optimization of clamp component values to achieve desired performance.
Example 4.7 Bipolar SCR ESD Clamp

Project Name: E4.7_BSCR_HBM
C2
2.6e-12
R3 L0 R1 Discharge R4
Charge HBM_Pulse
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 1e-11
V0 $Vpulse... 1e-10
ESDM
SCR0*
W=100
R2*
1e4
Fig. E4.7 Cross-sections of the bipolar SCR with different N-sinker peak doping values, HBM
mixed-mode simulation circuit for the bipolar SCR clamp, and simulated waveforms under the
2 kV HBM ESD pulse for different values of peak N-sinker doping level
This example provides a parameterized clamp circuit for the high-voltage bipolar
SCR clamp. The high-voltage bipolar SCR device is obtained from 20 V NPN BJT
device by embedding an additional P-emitter region isolated by an N-sinker. The
impact on the N-sinker peak doping level on the clamp holding voltage is shown in
Fig. E4.7. This example demonstrates the change of both the triggering and turn-
off characteristics of the device as a result of different doping levels of the N-type
isolation region around the P+-emitter. Alternatively, in practical cases, the N-sinker
implant location can be changed to reduce the level of P+-emitter isolation.
Example 4.8 Diode-Triggered SCR ESD Clamp

Project Name: E4.8_DTSCR_HBM
C2
2.6e-12
R3 L0 R1 R4
Charge Discharge HBM_Pulse
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
ESDP
C0 1e-11
V0 $Vpulse...
1e-10
R5*
R6* 0.1
D2 10M SCR0*
W=100
D0
W=100
W=100
R2*
D1 1e4
W=100
Fig. E4.8 Combined mixed-mode simulation circuit for the standard and diode-triggered SCR,
cross-section of the SCR implemented in the CMOS process, and voltage waveforms under 2 kV
HBM ESD pulse, demonstrating the capability of the DTSCR clamp to reduce voltage stress
212 4 ESD Clamps
A diode-triggered SCR is a key solution in the case of low-voltage pin protection.

It produces low-voltage waveforms to protect the gate oxide in CMOS circuits. The
example provides a parameterized SCR clamp circuit. The P+-emitter region of the
SCR is directly connected to the protected pin. Different options for connection of
the N-base electrode are realized by changing the values of resistors R5 and R6
to study both standard SCR and DTSCR modes of operation in a single example
circuit. The case of a high-side N-base connection is compared to the case of an
N-base connection to the ground via stacked diodes (Fig. E4.8).
Chapter 5
ESD Network Design Principles
This chapter covers material needed for understanding the next level of the ESD
design hierarchy – the protection network. The protection network or protection
circuit is usually composed of ESD protection clamps (cells) connected together in
a way that provides a high current path for all of the pin-to-pin combinations. This
network is engineered based on certain general principles and assumptions that are
discussed below.
In spite of a high variety of analog circuits, these major principles of network
design are rather universal. This fact is demonstrated in detail in this chapter, fol-
lowed by Chapters 6 and 7 that describe ESD protection network implementation in
analog signal path and power circuits with the help of practical examples.
The ESD protection network can further be understood as a virtual pulsed power
circuit composed of ESD clamps, appropriate high-current-capable metallization
routing, and components of the internal circuit of the product that interface with the
protected pins.
The main function of the network is to react to a high transient voltage and cre-
ate a discharge current path. The discharge current path should limit the voltage
level in the ESD time domain that suitably guarantees survival of the internal circuit
components.
Depending on internal circuit specifications, both transient-triggered and voltage-
reference clamps can be used to engineer the ESD protection network. Moreover,
in general, some internal circuit components that directly interact with protected
pins can also be used to provide a complete or partial discharge current path, thus
representing ESD protection network components.
The term ESD pad ring is often used in digital circuit design. The ESD pad ring
includes pads and a distributed ESD protection circuit created near the pads by a
combination of cells. ESD protection network components can be embedded in the
I/O cells. In most cases, the ESD pad ring can be decoupled from the internal circuit
and is capable of supporting different internal circuits.
For analog circuits, understanding ESD pad ring design is not always as straight-
forward as for digital circuits. The new level of complexity is the result of increased
interaction of the current path with the internal circuit.
In principle, an analog circuit pin can interface with many more connected active
devices. Often, deep inside the layout of the internal circuit, some components may

214 5 ESD Network Design Principles
see a significant ESD signal. Such a signal can be transmitted, for example, through
the parasitic capacitance of the integrated power devices. This situation may require
additional protection measures for the internal circuit nodes. Alternatively, in the
case of micro-SMD (surface-mount device) design, bump pads can be scattered on
top of the whole active layout area. Thus, the definition of the ESD pad ring for
analog design is rather “fuzzy.”
Nevertheless, although ESD network generally may not be decoupled from the
internal analog circuit, it can still be analyzed to a certain level using hierarchical
methods.
ESD pulse voltage limiting is targeted below the absolute maximum pulsed volt-
age rating of the devices interfacing with the pins. At the same time, during ESD
stress, the transient conditions of the control electrodes and internal circuit nodes
are generally unknown.
This conclusion is quite important. In Chapter 3, it has been demonstrated that
pulsed SOA of most standard active devices has a critical burnout voltage that is
dependent upon control electrode conditions. For example, in N-channel CMOS
devices, due to the multiplication of the channel current, the critical voltage for an
irreversible snapback event has a strong dependence upon the gate bias. This volt-
age is much higher below threshold, in comparison with open channel conditions.
Similar dependence is observed in NPN BJT, here as a function of the base current
(Chapter 3). Therefore, depending on the internal circuit design, the absolute maxi-
mum voltage of the internal components may vary significantly, impacting the ESD
protection window voltage range or even making it negative. Specifically, this aspect
is mainly responsible for why analog circuits often require custom ESD protection
networks or even custom clamp design.
Unfortunately, the conditions of the internal circuit are hard to predict or simu-
late. Accuracy or even the possibility of ESD pulse simulations is usually limited
by the absence of ESD snapback compact models, while an empirical analysis is
complicated due to the many floating points in the circuit.
The most robust solution for a simplified case is presented by a scenario in which
the voltage drop provided by the ESD protection network at the pin is below the
absolute maximum pulsed voltage for all the connected components at any allowed
control electrode conditions. This case is typical for high precision amplifiers,
for example, where the requirements for linearity of the components often auto-
matically provide a significant range between operational maximum and absolute
maximum voltages.
However, in the case of power-optimized analog circuits, the opposite situation
may arise. For example, in fast switching pins, due to the “aggressive” SOA of
switching power devices, the critical voltage in the on-state might be even lower
than absolute maximum voltage specified for the circuit.
As it will be demonstrated (this chapter and Chapter 6), the ESD protection net-
work is specific to its analog circuit. This implies a necessity of new experimental
evaluation of the product’s ESD performance each time the internal analog circuit is
modified. If verification results are negative and exhibit underpass issues, the next
5.1 Rail-Based ESD Protection Network 215
challenge is in detecting the reason for the issues and proposing and validating mea-
sures or solutions that would fix the problem. In this case, often several alternative
solutions can be generated, each with a different level of confidence.
The ability to make the right choice greatly depends upon the understanding of
ESD protection network design and the components used. This is a practical alter-
native to the comparison of new results with previous case studies that may not
be equivalent. Choosing the right corrective option from the several proposed ones
especially becomes a challenge in the case of product-dedicated tapeout. The mask
set may be rather expensive or a full change of the mask set may not fit the business
timeline. Often, running multiple die reticle (MDR) to validate different experimen-
tal ESD versions may not be a practical option. Under such circumstances, it is often
desirable to fix the problem using a minimal number of masks or by utilizing metal
options that can be applied to the wafer preserved before the backend process steps.
According to the authors’ experience in the field, ∼80% of all underpass issues
are related to mistakes that could have been avoided if simple principles and known
cases were studied and were taken into account.
The purpose of this and the following chapters is to provide the readers with
corresponding guidelines that will help them in practical design. One of the most
powerful tools in validating both ESD networks components and fixing options
is the mixed-mode numerical simulation. Several simulation examples for an ESD
protection network using DECIMMTM complement this chapter.
5.1 Rail-Based ESD Protection Network
5.1.1 Rail Based and Local ESD Protection

In order for an IC chip to pass the ESD qualification for the typical package level
specification (e.g., 2 kV HBM, 200 V MM, 750 V CDM), each individual pin must
be connected to the ESD protection network. In most cases, the type of application
determines the choice of ESD protection.
For a single voltage domain, the design of ESD protection networks can be
roughly subdivided into rail-based and local ESD protection types. The advantages
and drawbacks of both approaches are summarized in Table 5.1.
Local ESD protection is independent of bus resistance, as well as that of other
network components. It provides an easily predictable local voltage waveform at
every pin. However, especially in the case of high voltage protection or, alterna-
tively, in the case of high pin count, local clamps may not be the most optimal
space-saving solution. In addition, local clamps are generally more process-sensitive
or may greatly interfere with normal circuit operation.
A rail-based ESD protection network can be composed of a snapback power
clamp or a distributed network of active clamps and ESD diodes. Here, snap-
back clamps are contrasted with active clamps (Section 4.1) to emphasize the
Table 5.1 ESD protection strategies
Local snapback Diodes and active clamp
Main advantages Independent of bus resistance Area efficient

Local to every pin Highly process portable
SPICE-based design
Main drawbacks More area intensive Performance depends on bus
Highly process sensitive resistance
Difficult to simulate Distributed ESD network requires
complex chip level
implementation and verification
corresponding difference between the uncontrolled current in the conductivity mod-

ulation mode and the controlled current in the active operation mode of standard
devices.
In the case of distributed rail-based protection with active clamps and ESD
diodes, area efficiency depends on the pin count. For a product with a high pin
count area efficiency can be rather high, while for a product with fewer pins the
solution can be absolutely inadequate, even in the same voltage domain.
In principle, local protection can also be realized with active clamps, but this
solution is impractical.
The major advantage of the active clamp network is the possibility to use circuit
simulation to analyze performance under ESD pulse conditions. At the same time, at
a certain level of complexity, verification of the distributed ESD protection network
becomes a challenge.
Rail-based protection with a snapback power clamp is useful for high-voltage
analog circuits, for example, amplifiers. If the circuit is low voltage, then the rail-
based ESD protection network can be arranged using active clamps as described
in Section 4.1. The active NMOS clamp provides a significantly lower clamping
voltage due to its design principles.
ESD protection can be embedded in the input/output (I/O) cells. Under this
approach, a set of ESD and latch-up rules must be taken into account to provide
a non-conflicting operation of ESD elements and I/O cell components.
While the ESD compliance of each unique I/O cell is simply ensured by design,
ESD robustness of each product on the chip level can be achieved by the following
set of ESD-related rules during I/O ring integration. In the past, compliance of the
I/O ring with all the ESD rules had to be verified in a time-consuming and error-
prone manual review process.
To facilitate this process in high pin count digital domains, an automated tool is
usually developed that provides an automated assembly and verification of the I/O
ring for compliance with all ESD-related integration rules, in both wire-bond and
flip-chip designs.
In the case of rail-based network, there are several steps in the complex chip
verification. They usually include verification that every IC pin includes ESD
protection; analysis of the dedicated ESD discharge paths; extraction and verifica-
tion of interconnect robustness (contacts, metal, and vias) for ESD current density.
The verification can also be completed for ESD cells to confirm that the cells are
properly sized and their placement is correct.
There is a number of cases where active clamps cannot be used due to the
fast transient conditions at the power supply pin. Another conflicting instance that
limits active clamp use is in high-voltage circuits, where the size of the adequate
distributed network components is too spatially inefficient.
5.1.2 Rail-Based ESD Protection Using Snapback Clamps

Any rail-based ESD network can be identified by the presence of ESD busses (rails)
with a common clamp and the ESD diodes connecting the other pins to the rails
(Fig. 5.1). ESD diodes can connect input/output (I/O) pins or other control pins,
including power supply pins. The power supply pins connected to the rails typically
share the metallization routing. However, in principle, the network can be organized
with ESDPLUS and ESDMINUS busses that separate the power supply and ground
for the domain.
POWER (ESDPLUS) BUS
D1 D2
Snapback NMOS
INP OUT
Clamp
D4 D3
R
GND (ESDMINUS) BUS
Fig. 5.1 Rail-based ESD protection network
In rail-based ESD protection, the ESD diodes are designed to conduct pulsed
ESD current only in the forward bias mode. To avoid interference with the major
ESD current path, the voltage tolerance of the ESD diodes is an important parameter
that needs to be met. An appropriate breakdown voltage of the rail-based protected
domain – above the power supply level – should be provided to avoid ESD diode
burnout due to avalanche breakdown mode and to minimize direct impact on circuit
functionality in normal operation regime. For example, the p+ -Nwell diode with
a breakdown voltage of ∼11 V should not be used to protect a 12–10 V erase pin
EEPROM circuit block.
An example of the simplified four-pin rail-based circuit is illustrated in Fig. 5.1.

The network provides protection of the input (INP), output (OUT), POWER, and
ground (GND) pins. This network can represent an amplifier. Additional pins can be
added to the network by reusing similar diode connections to the I/O pins, as shown
in Fig. 5.1. However, the higher pin count will not change the analysis below.
For each pin-to-pin combination, this rail-based network forms an ESD current
path through the forward-biased ESD diodes, corresponding parts of the rails, and
the power clamp. The power clamp provides both the high-voltage current path and
the reverse path through the clamp diode. For example, in the case of positive zap
of INP vs. OUT, the current path is formed through the upper diode D1, the upper
part of ESDPLUS bus, then through the snapback clamp, the corresponding part of
ESDMINUS bus, and the forward-biased diode D4 (Fig. 5.1).
Rail resistance should also be taken into account at ESD current level; the total
voltage drop on the circuit and the voltage drop at each pin relative to the ground
can be estimated. Results of the calculation are summarized in Table 5.1. These
results are obtained under the assumption that the clamp voltage drop is equal to VC
and the voltage of the I/O and clamp forward-biased diodes is equal to VD , with the
contribution of bus resistances RESDPLUS and RESDMINUS into the current path.
The table below can be extended for a more complex case or modified for
additional accuracy. Higher accuracy can be achieved by taking into account the
distributed bus resistance extracted by CAD tolls from exact metal routing.
Similar reference tables are often used for ESD checkers implemented in the
CAD environment. The checkers also extract the parasitic resistance and capaci-
tance of the interconnects [104].
Several important conclusions can already be made through the analysis pre-
sented in Table 5.2.
Table 5.2 Rail-based peak voltage at peak ESD current
OUT voltage POWER

ESD zap Total voltage INP voltage vs. vs. voltage vs.
combination drop on zap pins ESDMINUS ESDPLUS ESDMINUS
INP to VD VD 0 0
ESDPLUS
PWR to VD ∼VD /2 ∼VD /2 VD
ESDMINUS
ESDMINUS to VD VD 0 0
INP
PWR to VC VC /2 VC /2 VC
ESDMINUS
INP to OUT 2VD +IH VD +IH VD ∼VD
(RE– +RE+ )+VC (RE– +RE+ )+VC
INP to VD +IH VD +IH (IH (RE– +RE+ ) IH (RE– +RE+ )
ESDMINUS (RE– +RE+ )+VC (RE– +RE+ )+VC +VC )/2 +VC
First of all, the biggest overstress occurs for the INP–OUT pin combination.
Therefore, the diode and power clamp components at that combination should pro-
vide a low on-state resistance to limit input voltage overstress at ESD pulse. In this
case, the major components are the I/O diodes, which introduce signal limitation
due to parasitic capacitance.
Another important consideration that cannot simply be neglected is the ESD bus
resistance. Usually, resistance of first metal layers is ∼30 m per square of the
metal area. For chip die dimensions of 4 mm × 4 mm, the total bus length at the
corner opposite to the clamp will be ∼8 mm. In the case of a closed ESD pad ring,
the bus is routed along both sides of the chip and should be accounted for twice. If
a single metal bus with a 100 μm width is used (or an equivalent double metal layer
bus of a 50 μm width), the total accumulated resistance in the current path will be
1.2 . Thus, at a 2 kV HBM 1.33 A ESD current, the protected pin opposite to the
power clamp in the chip will accumulate an additional voltage drop of ∼1.5 V. In
principle, this level is acceptable for a 5 V process, but might create a problem in
the 1 V process. If the bus width is only 10 μm, the corresponding 10-fold higher
voltage drop can result in immediate ESD failure of protected pins opposite to the
power clamp in the die.
One of the major disadvantages of rail-based protection is the cross-talk between
the input and output, mainly through the parasitic capacitance of ESD diodes.
Rail-based protection with a snapback power clamp has already been shown
in Fig. 5.1. In opposite to the circuit with an active clamp component, this con-
figuration relies on a parasitic device in high-current conductivity modulation
mode.
As was discussed in Chapter 3, in a 0.5 μm CMOS process, for example, the
lateral parasitic n–p–n structure of the NMOS can provide a current of 3–5 mA/μm.
Thus, the total width of NMOS array required for a 2 A ESD current is ∼400 μm.
Unlike the active clamp, the snapback device is rather sensitive to layout and pro-
cess variation. At the same time, the voltage of the clamp cannot be reduced below
the holding voltage of the NPN, thus providing a limiting factor of the network
clamping voltage.
Rail-based protection with a snapback power clamp is especially efficient in
bipolar processes, high-voltage products, transient power supply, or voltage domain
operation, and in the case of low pin count. It is advantageous in case of critical
requirements for the silicon space, when active clamp rail-based protection becomes
no longer optimal.
5.1.3 Rail-Based ESD Protection Using Active Clamps

Active clamps rely on the device working in normal operation mode [60–70]. The
original and most typical implementation (Section 4.1) of this clamp is in an over-
sized low-voltage NMOS array capable, in the on-state, of safely sinking the current
generated by HBM, MM, or CDM pulses for typical precharge voltages [60].
In the case of a 5 V 0.5 μm process, the typical ratio of the total NMOS width
and length, W/L, is 5000/0.5. This ratio is required to provide a saturation current
of ∼2 A. Bus resistance plays an important role and must also be included in sim-
ulations. The RC-timer time constant typically selected is ∼6 μs, which keeps the
active clamp in the off-state during normal operation and, at the same time, avoids
significant overload during voltage increase at the powering sequence.
Both the distributed NMOS array and RC timer are space-consuming clamp com-
ponents (Chapter 4). The clamping voltage depends on the on-state resistance of the
clamp and directly on NMOS width. For the 5 V process, active clamps are usually
designed with an NMOS voltage drop below ∼2 V. This provides an appropriate
design margin for the voltage drop on the I/O and clamp ESD diodes and metal
busses. For extremely low parasitic capacitance at the I/O pin the rail clamp can
be oversized in order to provide a much lower clamping voltage, while the size of
the ESD diode can be significantly reduced. In this case the additional voltage drop
on the diode will be compensated by reduction of the clamp voltage, providing an
advantage in lower equivalent capacitance at the protected pin.
The most typical applications for rail-based ESD protection with an active clamp
are low-voltage circuits. These circuits provide CMOS components with an appro-
priate voltage tolerance to the power supply level, a conventional (relatively slow)
powering sequence for the power supply and input pins, and a sufficient pin count.
In this case, the advantageous low on-state resistance allows the designer to build
a distributed network of RC-controlled power NMOS devices with a total width of
∼3000–5000 μm.
A simplified schematic and layout of the rail-based active clamp ESD protection
network is presented in Fig. 5.2. A more detailed description of active clamp design
principles and operation is given in Section 4.1. It is practical to compose the active
clamp library of standard cells that include “slave” active clamp cells and an RC-
timer cell. In the case of high pin count, the RC-timer cell can be placed in the corner
Filler cell Slave active clamp cells
RC-timer
cell
Pad ESDP
…
ESDM
Trigger-off line
P G P G
M0 M1 M1 M1 M1 M0 M0 M1 M1 M1 M1 M0
Fig. 5.2 Example of layout and schematic views for 3 V ESD domain obtained with redesign
according to new design rules in active clamp ESD library [104]
of the chip. Compatible filler cells can provide the connection between the “slave”
active clamp cells and the RC-timer cell. The filler cell continues the bus connection
for ESDPLUS, ESDMINUS, and the trigger line (Chapter 4) and can accommodate
an additional NMOS device.
Due to the large size of the NMOS device, the active clamp is usually designed
as a distributed network across the pad ring area, occupying ∼100 × 100 μm at
each pad. At the typical pad dimension of 70–120 μm, each “slave” active clamp
cell usually accommodates a 500–1000 μm wide NMOS with a turn-on capacitor of
∼500 fF and a pair of ESD diodes. An optional trigger line capacitor can be added
to compensate for the voltage drop across the trigger line. In this case, at least 8–10
clamps are necessary to provide appropriate voltage clamping in a single voltage
domain across the ESD ring.
An example of part of the pad ring schematic is shown in Fig. 5.3. Each I/O PAD
has a diode connected to ESDPLUS and ESDMINUS. An RC trigger cell acts as a
low-pass filter and turns off the active clamp during normal operation. The RC time
constant is set to 6 μs. When the chip is powered up, with a power supply voltage-
ramping time constant above 6 μs, ESDPLUS gains a higher potential relative to
ESDMINUS and the VTRIG node also remains at high potential.
ESDPLUS
ESD Bus ESD Bus ESD Bus ESD Bus ESD Bus ESD Bus
resistance resistance resistance resistance resistance resistance
Rmc ESD ESD ESD ESD ESD ESD

PLUS PLUS PLUS PLUS PLUS PLUS
Vtrig Vtrig Vtrig Vtrig Vtrig Vtrig
ESD Bus ESD Bus ESD Bus ESD Bus ESD Bus ESD Bus
Cmc ESD
-
resistance ESD
-
resistance ESD
-
resistance ESD
-
resistance ESD
-
resistance ESD
-
resistance
Fig. 5.3 Distributed Merrill clamp scheme suitable for 5 V ESD protection
The active clamp protection scheme is advantageous because the clamp voltage
is proportional to the number of distributed cells used in the design. If the clamp
voltage is below VDD, then there is no requirement for current-limiting resistors at
the input or output, or indeed any other precautions. The disadvantage of the scheme
is the large area required for the protection network.
Clamp usage is limited if chip specifications require a “hot plug-in” compati-
bility. The possible fast rise time during hot plug-in events may activate the ESD
protection, resulting in voltage overstress of the clamp. In this case, the snapback
solution is required.
I/O ESD compliance of individual I/O cells is achieved by the following ESD
integration rules.
In practical application, the active ESD clamp solution is often designed in con-
junction with I/O library cells, as this is an efficient use of chip space (Fig. 5.4).
I/O circuits in analog IC typically contain many different I/O cell types that are
optimized for the many different types of pins used in the circuit, for example,
high-speed, power, high-voltage, EEPROM, digital, LVDS, and USB pins.
Fig. 5.4 Example of I/O and active clamp layout with mark-up of the area used for distributed
ESD protection [104]
The complex design of the distributed pad ring requires verification of many sets
of ESD integration rules. Due to a high variety of I/O multiple verifications of the
whole pad ring are required. This is often achieved through automated extraction
and checker software [104] implemented in the CAD environment.
In a properly built active clamp protection with appropriate bus resistance,
the ESD-current-conducting NMOS operates in normal operation mode, far from
avalanche breakdown conditions. Therefore, at overload, there are usually no fail-
ures observed in this component. Perhaps one of the most vulnerable active clamp
components is the speed-up capacitor that is responsible for NMOS turn-on during
the fast rise time of the ESD signal. Therefore, a robust operation is required from
this component. For example, the most process-sensitive parts of the capacitors are
corner regions. An example of ESD damage location is presented in Fig. 5.5.
Other process-related issues are associated with the large NMOS itself. Often, in
small analog chips, the Active Clamp NMOS is the device with the greatest total
width, while the internal circuit may be either mainly BJT-based or include a negli-
gible amount of the minimum-width input NMOS devices. In this case, the yield of
the product may be significantly impacted by the choice of ESD solution, due to the
different yield of the large NMOS array in comparison with that of relatively small
snapback NMOS clamp.
ESDP
20u
VTRIG
3ux1u 5u
Fig. 5.5 One of the “weak spot” failures in the active clamp as a result of irreversible speed-up
capacitor breakdown. Circle indicates the capacitor which had a leakage between the P+ polytop
plate and P-sinker bottom plate
5.1.4 Specific of Active Clamp Design in BiCMOS Processes

5.1.4.1 Verification by Circuit Simulation
A circuit simulation with extracted parasitic metallization resistance and capaci-
tance is required to ensure that the clamping voltage is sufficiently low. In addition
to the simulation verifying the signal in normal operation regime, three types of
ESD simulations are usually required.
The first HBM source simulation targets the single pulse waveform to verify
the low clamping voltage level between ESDPLUS and ESDMINUS. The sec-
ond simulation is done for the sequence of ESD pulses between I and O to verify
that the clamp does not suffer from a voltage accumulation effect. This effect can
be observed irrespectively of whether the tester provides a short circuit condi-
tion between zaps, since the I/O diodes separate the clamp from the tester. This
simulation is important because it ensures that the clamp keeps the voltage low
for chips with a high pin count when multiple ESD zaps are done during ESD
testing.
Finally, a simulation should be performed for the power supply sequence. This
simulation is done to ensure that the clamp does not drain an excessive current dur-
ing the expected ramp of the power supply and provides low leakage after achieving
the power supply voltage level.
Most popular design solutions are based on the NMOS acting as a clamping
device. The circuit for this solution has been discussed in Section 4.1.
To illustrate circuit simulation aspects in this section, more examples for active
clamps are analyzed based upon NPN BJT and PMOS devices for a 3 V BiCMOS
process.
Slave NMOS Clamp Schematic

ESDP ESDP
I/O MP
R1 DN C2
INT R
M 80/2 20/0.4
600K 0.7pF
1
VTRIG
C1 I/O PAD VTRIG MN QN
DP 500/0.25
10p 5/0.4
80/2
ESDM ESDM
a) b)
7 12 10–2
6 10–4
Leakage Current (A)

10
Clamp Voltage (V)
Clamp Voltage (V)
1 clamp 1 clamp
5
8 10–6
4
2 clamps 6 10–8
3
10 clamps
3 clamps 4 10 clamps 10–10
2
5 clamps 10–12 1 clamp
1 2
10 clamps
0 0 10–14
10p 1n 100n 10u 1m 0 5m 10m 0 500u 1m
Time (s) Time (s) Time (s)
c) d) e)
Fig. 5.6 Circuit diagram for the RC timer (a) and slave clamp (b) cell based upon NPN BJT and
the results of a circuit simulation for the single pulse waveform (c), multiple zap sequence (d), and
power supply sequence (e)
The NPN BJT clamp has a 3 V CMOS driver with an NPN base directly con-
nected to the inverter output (Fig. 5.6a, b). Each clamp has an NPN with a total
emitter width of 540 μm. The RC-timer time constant is ∼6 μs.
The clamp provides an appropriate low leakage in the powering sequence
(Fig. 5.6e) and the desired single pulse waveform (Fig. 5.6c) with a low clamp-
ing voltage, if the number of slave clamps is sufficiently high (Fig. 5.6c). The peak
amplitude of the waveform produced by the circuit is below 1 V and practically does
not change starting with ∼10 clamps (Fig. 5.6c). However, the voltage accumula-
tion effect is observed in the case of a multiple zap sequence through the I/O diodes
(Fig. 5.6d).
A similar phenomenon is observed in the 3 V PMOS-base active clamp (Fig. 5.7),
unless the clamp count is significantly increased. The clamp uses the same RC
timer (Fig. 5.6a), and the slave clamp is built upon the PMOS with a total width of
1368 μm (Fig. 5.7a). With the increase of the clamp count, the problem of the accu-
mulation effect is solved (Fig. 5.7c). The accumulation effect physically represents
the effect of incomplete discharge between the rails. The effect can be significantly
different, depending on the power consumption and leakage of all the different cir-
cuit blocks that are not included in this section. If the circuit can tolerate the leakage
of a simple “bleed” resistor connected between the rails, the accumulation effect is
eliminated. Alternatively, a more sophisticated discharge, at low current levels and
with a larger time constant, can be proposed.
I/O ESDP
DN MP1 MP2 MP3
INT R
M 80/2 28/0.24
12/0.24 2K/0.24
1
I/O PAD VTRIG MN1 MN1

C1 C2
DP 3/0.24 7/0.24
DA 0.3p 1.4p
80/2
ESDM
a)
1.8 10
1.6 9
3x Width 30 Clamps
(4104um- 8
1.4 10clamps)
7
1.2
Clamp Voltage (V)
Clamp Voltage (V)

30 Clamps 6
1.0 5 3x With
(4104um)
0.8 60 Clamps 4
3
0.6
2 60 Clamps
0.4
1
0.2 0
0.0 –1
10p 100p 1n 10n 100n 1u 10u 0 4m 8m 12m
Time (s) Time (s)
b) c)
Fig. 5.7 (a) Circuit diagram for a slave clamp based upon 3 V PMOS and the results of circuit
simulation for the single pulse waveform (b) and multiple zap sequence (c), depending on the
number of clamps
5.1.4.2 Experimental Comparison

In the experimental setup, tests similar to the above simulation analysis are rather
difficult to reproduce. The adequate comparison requires access to a waveform cap-
ture tester. An alternative way to compare at least high-current performance is by
using TLP characterization.
Using standard TLP pulse tools, the ESDP to ESDM combination is compared
below for several alternative clamp designs in the BiCMOS process. With appropri-
ate clamp design, the corresponding IO-to-IO or other zap combinations are usually
different from the ESDP–ESDM with an additional voltage drop on the two corre-
sponding ESD diodes. A comparison of these characteristics for the case of 10 slave
clamps with a total width W = 540 μm in a 2.5 V NMOS device is presented in
Fig. 5.8.
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
8
7
ESDP
F9_10 vs ESDM
IO vs IO
F2_1
6 Gnd (Vleakage 2.5V)
Pulsed Current (A)
Gnd (Vleakage 2.5V)

5
0
0 1 2 3 4 5 6 7 8
Pulsed Voltage (V)
Fig. 5.8 Comparison of the 100 ns TLP characteristics for a 2.5 NMOS-based active clamp in the
case of ESDM–ESDP vs. IO–IO zap combinations
An additional important conclusion drawn from the data (Fig. 5.8) is that the lim-
itation of the clamp is not related to the level of the critical current IT2 . A network
assembled from 10 clamps can easily provide a current level that is significantly
higher than the one dictated by the corresponding standard package-level specifica-
tions requirements, for example, in the 2 kV HBM pulse. This level corresponds to
a peak current of ∼1.33 A, which creates a voltage drop of ∼4.5 V in the example
of IO to IO (Fig. 5.8). On the contrary, the limitation of the clamp is related to the
voltage waveform produced by the clamp. An appropriate low voltage is achieved
by increasing the number of clamps in the network.
The effect of the linear width scaling for the clamp design previously analyzed
by the numerical simulation (Fig. 5.6) is demonstrated in Fig. 5.9.
As expected, a better high-current capability is observed for the NMOS-based
active clamp vs. the PMOS-based active clamp, due to the better high-current char-
acteristics of the n-channel device (Fig. 5.10). Another major conclusion supported
by the experimental data is the comparably enhanced performance of NMOS-
driven BJT-based slave clamps with respect to pure CMOS active clamp design,
as demonstrated in Fig. 5.10.
In the case of non-stacked low-voltage solutions, the BJT clamps provide an
advantage mainly due to the smaller area of the BJT device as compared with
the same degree of performance. However, in the case of higher voltage toler-
ance protection, the BJT design provides a significantly superior performance when
stacked NMOS clamp drivers are used. This case is described in the following
section.
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
8
10 clamps
7 A9_10
5 clamps
B4_5
1 clamp
C2_3
6 3 clamps
J3_4
Pulsed Current (A)
2 clamps
N3_4
Gnd (Vleakage 3.3V)
Gnd (Vleakage 3.3V)
Gnd (Vleakage 3.3V)
0
0 1 2 3 4 5 6 7 8
Fig. 5.9 Comparison of the ESDP–ESDM TLP characteristics for an ESD protection network
assembled with 10, 5, 3, 2, and 1 slave clamps
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
8
6
Pulsed Current (A)
3 3V NPN BJT
A9_10
2.5V NMOS
F9_10
2 2.5V PMOS
D9_10
Gnd (Vleakage 3.3V)
Gnd (Vleakage 2.5V)
0
0 1 2 3 4 5 6 7 8
Pulsed Voltage (V)
Fig. 5.10 Comparison of the ESDP–ESDM TLP characteristics for the 10-clamp network
assembled with 3 V BJT NPN, 2.5 V NMOS, and 2.5 V PMOS slave clamps
5.1.4.3 Active Clamp Protection in Complementary BiCMOS

with Low-Voltage CMOS Components
One of the most complex active clamp solutions is realized in the complementary
BiCMOS process with mixed voltage tolerance of the CMOS and BJT components.
Similarly, in the case of modular process approach, the designer may select cost-
minimizing process options such that the mask set supports high-voltage bipolar
devices, but only low-voltage CMOS. In this case, a desirable solution needs to be
engineered based upon a relatively large footprint-stacked CMOS RC-timer block
and driver, while the active high-current clamp component can be implemented
using a BJT device.
A particular example of the solution is presented below for the complementary
BiCMOS process with 5 V high-gain SiGe NPN and PNP devices, but with only
2.5 or 3 V CMOS components available. The target requirement for the active
clamp solution is to provide an active power clamp implementation for the 5 V
power supply.
The starting point for design comparison is the alternative fully CMOS configura-
tion. Such a clamp can be engineered using stacked CMOS devices (Fig. 5.11a). The
clamp contains an RC timer based upon a stacked poly-capacitor. The RC timer gen-
erates two levels of triggering signals VTH and VTL to control the slave clamp driver.
These signals are connected to the two distributed trigger lines across the slave
clamp network. Each slave clamp is composed from stacked inverter and stacked
power 3 V NMOS devices MNL and NMH (Fig. 5.11a). The operation principle of
the stacked clamp is similar to the non-stacked CMOS clamp, discussed above. The
speed-up capacitors provide for a rapid turn-on of the slave clamp NMOS devices
MNL and NMH when the fast rise time of the ESD pulse voltage is applied between
the ESDP and ESDM busses. After the delay time generated by RC timer, the two
trigger line signals provide for MNH and MNL turn-off into the high impedance
state. At the slow rise time power supply ramp, the RC timer keeps the stacked
power NMOS device in the off-state.
The aggressive area-efficient solution for the power part of the clamp is presented
in Fig. 5.11. It can be realized using either NPN or PNP devices. A version of the
clamp with NPN BJT (Fig. 5.11b) is compared to the clamp version containing PNP
BJT (Fig. 5.11c). For practicality, the stacked NMOS clamp (Fig. 5.11a) is reused
to provide the low-side driver for the common emitter NPN device (Fig. 5.11b) or
the high-side driver for the PNP device (Fig. 5.11c).
The results of the simulations performed according to the methodology described
in the previous section and the layout implementation are summarized in Figs. 5.12
and 5.13, respectively. For the given timer and driver designs, both clamp versions
demonstrate appropriate voltage waveforms for the 5 V voltage domain (Figs. 5.12c
and 5.13c), though better clamping characteristics are realized in the PNP-based
solution. A voltage accumulation effect is observed in the network with a small
NPN-clamp count (Fig. 5.12d).
The conclusion of the superior efficiency of the BJT-based overvoltage clamp is
summarized in Fig. 5.14. A network with 10 stacked NMOS slave clamps is com-
pared to a network with a single 1080 μm PNP BJT clamp designed according to
the schematic of Fig. 5.13.
A more detailed comparison of BJT-based clamps is presented in Fig. 5.15
for solution footprint optimization. Six different versions of the clamp, “G,” “H,”
“I,” “K,” “L,” and “M,” are cross-compared. In this particular circuit design, the
Multiple Slave Clamps
ESDP VTH
MNH
RC
Timer
I/O MNL
VTL
ESDM
a)
ESDP VTH
MNH
RC
Timer
I/O MNL
VTL
NPN
ESDM
b)
ESDP VTH
PNP
RC
Timer MNH
I/O MNL
VTL
ESDM
c)
Fig. 5.11 Examples of 5 V active clamp implementation in 5 V complementary BiCMOS process
with 2.5 or 3 V CMOS module. Cascaded NMOS active clamp (a), NPN based (b), and PNP
based (c)
PNP-based clamps (“K,” “L,” and “M”) provide a better high-current performance
in comparison with NPN devices (G,” “H,” and “I”) of the same size, as expected
from the above simulation analysis. The lower performance of NPN-based slave
clamps in comparison with PNP-based clamps is most likely related to the driver’s
capability to provide a side base current, since the gain of the NPN devices is higher.
Non-linear width scaling is observed in the analysis. A twofold size reduction of the
BJT results in only a 30% loss in the on-state current (compare the “H” and “L”
clamp versions with the original-sized clamps “G” and “K,” respectively). Another
way to reduce the footprint of the power clamp is by reducing the size of the driver.
However, this measure results in a significant drop of high-current clamp perfor-
mance (Fig. 5.15, clamps “I” and “M”) for the NPN and PNP versions, respectively.
ESDP
MP2 C2
230
5V DA1 20/0.4 0.9p

Driver- MN2
RC- VTH
NMOS 640/0.4
timer I/O
DN C0 MP1
INT R 0.3p 20/0.4 DA3
M 80/2 C3
0.1 MN3
VTL 0.9p
640/0.4
I/O I/O PAD
ESDP ESDM C1 MN1 QN1
DA4
NPN Pad DP 0.3p 5/0.4 RD 500/0.25
80/2 DA2 10K
ESDM
a) b)
5 12
10–5
2NMOS-1BJT 10 2NMOS-1BJT
4 10–6
6NMOS-1BJT
8 10–7
3 6NMOS-2BJT
6NMOS-2BJT 6 10–8
2
6NMOS-1BJT
4 10–9
1 –10
Clamp Voltage (V)

Clamp Voltage (V)
2 10
Leakage Current (A)
0 0 10–14
10p 1n 100n 10u 1m 0 5m 10m 0 500u 1m
c) d) e)
5 ESD Network Design Principles
Fig. 5.12 Layout view (a) and circuit diagram (b) for the 5 V NPN-based slave clamp and the results of circuit simulation analysis for the single pulse
waveform (c), multiple zap sequence (d), and power supply ramp (e) for different clamp component width scalings
ESDP
5V Driver-
RC- NMOS
timer
PNP I/O
ESDP ESDM
Pad
ESDM
a)
ESDP
C2 RD QP1
DA1 0.9p 10K 540/0.25
MP2
VTH 20/0.4
MN2
I/O MP1
DN C0 640/0.4
INT 0.3p 20/0.4 D A3
RM 80/2 C3
0.1 MN3
VTL 0.9p
640/0.4
I/O PAD C1 MN1
D A4
DP 0.3p 5/0.4
80/2 DA2
ESDM
b)
3 10–4 10–5
4
10–5
Leakage Current (A)
Clamp Voltage (V)
Clamp Voltage (V)
3 2NMOS-1BJT 10–6
2
6NMOS-1BJT 10–7
2
10–8
6NMOS-2BJT
1 6NMOS-2BJT
10–9
1
10–10
2NMOS-1BJT
0 0 10–11
10p 1n 100n 10u 1m 0 5m 10m 0 500u 1m
c) d) e)
Fig. 5.13 Layout view (a) and circuit diagram (b) for the 5 V PNP-based slave clamp and the
results of circuit simulation analysis for the single pulse waveform (c), multiple zap sequence (d),
and power supply ramp (e) for different clamp component width scalings
K01 K02 K03 D01 D02 D03 D04 D05 D06 D07 D08 D09
a) b)
Leakage Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
10
9
8
Pulsed Current (A)
7 D1_10
K2_3
6 Gnd (Vleakage 5V)
Gnd (Vleakage 5V)
5
d
4 a se
P -b ed
3 as
PN S-
b
2 5V MO
e dN
1 ck
sta
0
0 1 2 3 4 5 6 7 8
c)
ESDP-ESDM Pulsed Voltage (V)
Fig. 5.14 Comparison of the layout footprint for 5 V ESD protection solutions: 1-clamp PNP
BJT (a) and 10-clamp stacked NMOS (b). Comparison of the experimental TLP characteristics for
ESDP–ESDM pad combination for the cluster of 10 stacked 3 V NMOS clamps and PNP-based
active clamp (c)
5.1.5 Bipolar Differential Input Protection

Rail-based ESD protection is one of the most typical solutions used for power ampli-
fiers. In the case of bipolar differential input, the typical emitter–base breakdown
voltage with an open collector (BVEBO) is ∼3–6 V. Thus, the base–emitter junc-
tion is usually rather sensitive to voltage overstress and requires appropriate voltage
limiting to avoid damage.
Since the absolute maximum voltage at the input pins is relatively low, diode
clamps are used to protect the differential BJT input pins. The voltage limiting
across the base–emitter junction can usually be implemented in two ways, using
either large differential diode clamp or the two-stage network approach with a small
differential diode clamp. The final choice depends on the circuit specifications.
If a larger amplitude of differential voltage is required (Fig. 5.16), up to three
diodes can be stacked in the clamp. When such a diode clamp is used as a primary
ESD path between the inputs during an ESD event, a 1.5 A current through each
diode results in a voltage drop. The diodes should be appropriately oversized to
Leakage Current (A)

1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
8
6
Pulsed Current (A)
4
G2_3
H2_3
I2_3
3 K2_3
L2_3
M2_3
2 Gnd (Vleakage 5V)
Gnd (Vleakage 5V)
Gnd (Vleakage 5V)
1 Gnd (Vleakage 5V)
Gnd (Vleakage 5V)
Gnd (Vleakage 5V)
0
0 1 2 3 4 5 6 7 8
Pulsed Voltage (V)
Fig. 5.15 Comparison of the experimental TLP characteristics for the ESDP–ESDM pin combi-
nation for the different versions for NPN and PNP BJT-based clamps. (G) 1080 × 0.24 μm2 NPN
BJT-based clamp; (H) 540 × 0.24 μm 2 NPN BJT; (I) clamp “H” with a three times reduction of
the output-stacked NMOS device width in the driver; (K) 1080 × 0.24 μm2 PNP clamp; (L) 540 ×
0.24 μm2 PNP clamp; (M) clamps “L” with a three times reduction of the output-stacked NMOS
device width in the driver
Power
Clamp
a)
Current
limiting
Current limiting R Power
R Clamp
b)
Fig. 5.16 Bipolar differential input protection without a current-limiting resistor at the pads, a
large ESD diode clamp (a); with current-limiting resistors at the pads, a small differential diode
clamp (b)
provide a high-current voltage drop of less than 1.5 V, which will keep the total
clamp voltage drop within 4.5 V. This approach is mainly used when the inputs
cannot tolerate additional resistors due to the degradation of circuit performance.
A more sofisticated approach can be implemented if a current-limiting resistor
can be included in the circuit (Fig. 5.16b). In this case, much smaller diodes can be
used in the back-to-back stacked diode clamp to provide a similar clamping voltage
at the inputs. The resistor value must be sufficient to limit the current when the
voltage between the pads is above the power supply level. A typical resistor value
is 150 and can be precisely calculated using BJT compact models and circuit
simulation.
The number of diodes in the differential diode clamp depends on the desired input
signal amplitude. However, even if the signal amplitude is small, in most practical
cases it is desirable to have at least two stacked diodes with shallow diffusion con-
nected to the pins to reduce the parasitic effect of the substrate junction and balance
the leakage current. For example, stacked diodes can be constructed by stacking the
p+ -Nwell and n+ -Pwell diodes, connecting the well tap diffusions in the floating
node.
5.1.6 Bipolar Output Protection

In the rail-based ESD network, bipolar output driver protection can be implemented
using a standard NPN BJT clamp and one of the voltage reference techniques in
the collector–base circuit (Chapter 4). This measure is effective if the output BJT
devices remain under low base current conditions during ESD stress and thereby
provide a high critical voltage. If the critical voltage is above one that corresponds
to an open base voltage, then an open base reference can be used in the power clamp
(BVCEO) (Fig. 5.17).
However, if the base coupling is high, then the NPN collector–emitter current
path should be protected in a different way. This statement is true unless the NPN
emitter area is sufficient in providing self-protection of the output devices. Usually,
self-protection is possible at an NPN emitter size of ∼500 μm2 .
The worst-case scenario is realized when there is an additional capacitor con-
nected between the base and collector circuit nodes. When the output pin connected
ESDP ESDP
Power
Clamp
+
I/O PAD I/O PAD
ESDM ESDM
Fig. 5.17 Two ESD paths through the circuit illustrating the possible destruction of the output
driver if the ESD current through the device is not limited
to the collector is stressed, the capacitor provides a positive base current with a
corresponding reduction of the triggering voltage, at a small multiplication fac-
tor (Chapter 3). As a result, all of the ESD current is directed through the NPN
(Fig. 5.17).
A countermeasure to suppress this effect is to add another current-limiting resis-
tor. To protect the output pad in this case, it is necessary to ballast the collector
with a sufficient resistance or to ballast the emitter with a sufficient polyresistance
to limit the current below a safe level.
The choice depends on the size of the bipolar device. These options can both
be implemented by having only a single resistor for each bipolar device or by
implementing separate local resistors at each segment (finger).
Due to a high current provided by the BJT devices, in most practical cases, the
total ballasting resistance should be within ∼1 . This value often insignificantly
impacts output performance, especially in the case of high-current output. The most
practical design in the case of power amplifiers takes into account heat generation
balance and mutual heating. In this case, the output drivers NPN and PNP are drawn
as multiple transistors, with polyresistors ballasting at each individual emitter. The
polyresistance per emitter value could be rather high, for example, 100 per device
finger.
In an isolated process, another efficient way to protect the output stage is by
adding a Pwell resistor between the pad and driver output. In this case, current
limiting is provided for both the NPN and PNP output drivers. The Pwell resistor
is preferred over the polyresistor, as it might benefit from some current saturation
under ESD stress – resulting in a higher effective resistance.
5.1.7 CMOS Input and Output Protection

In the case of BiCMOS circuits, both bipolar and CMOS outputs and inputs should
be protected. For the CMOS domain, rail-based ESD protection can be realized
using automatically voltage-tolerant CMOS-based power clamps. This can be done
by implementation of either the active CMOS clamp or one of the CMOS-based
snapback clamps, for example, the snapback NMOS clamp.
The snapback NMOS clamp is usually preferable, since FOX devices produce a
too high triggering voltage to be suitable for rail-based protection. At the same time,
SCR devices generally produce a relatively low holding voltage below the power
supply level, thus creating potential conditions for the transient latch-up scenario.
Therefore, snapback NMOS protection is usually preferrable as the self-aligned
option that is the least sensitive to process variation. However, since the NMOS
structures in the snapback and output devices are rather similar, there are several
important considerations that should be taken into account.
The first problem is in the grounded-gate power clamp, which provides relatively
high triggering and holding voltages. While the triggering voltage can be reduced
by the dynamic coupling of the gate, the high holding voltage should be accounted
for in the total voltage drop.
Based upon the SOA of NMOS and PMOS, the most critical factor is the snap-
back of the NMOS output driver, while the PMOS device can withstand a much
higher voltage level before burnout.
If rail-based protection is realized with the snapback NMOS power clamp for the
PAD–GND ESD zap, the current may be directed through the output NMOS rather
than power clamp, due to the additional voltage drop on the diode and metallization.
The first step in improvement is to add a local snapback NMOS clamp between
the PAD and GND. This solution will at least eliminate the additional voltage drop
on the upper diode and metallization (Fig. 5.18a). However, even during ESD,
the transient triggering voltage of this additional voltage clamp will compete with
the triggering voltage of the output NMOS, which might have rather large gate
coupling.
If the open drain protection is required, then a resistor at the PAD can be used to
limit the current into the output NMOS device. The value of the resistor can be cal-
culated using the measured value of the critical avalanche current that is supported
by the NMOS device reversibly prior snapback.
In case of low gate potential of the output NMOS, the lowest avalanche
breakdown current density per micron gate width that can be expected is jCR ∼0.02–
0.05 mA/μm. However, in this case, the triggering voltage of the NMOS is high and
greater than that of the local clamp. If gate coupling in the NMOS driver is high,
then the device can be expected to operate above the threshold. In this case, the
triggering voltage will correspondingly decrease down to the range of the holding
voltage.
ESDPLUS
ROUT
Power
Snapback
Clamp
Local
Snapback
Clamp
ESDMINUS
Silicide Block
a)
ESDPLUS
Power
Snapback
ROUT
Clamp
Local
Snapback
Clamp
ESDMINUS
b)
Fig. 5.18 Simplified circuit diagram of an NMOS output pad with a current-limiting resistor and
an additional local clamp for ESD protection (a); self-protecting output based upon NMOS with a
drain ballasting region (b)
An open channel current jCR can be used to roughly estimate the required decou-
pling resistor between the output node and pad. The resistor should provide a voltage
drop above the difference of the triggering voltage VT1 for the local snapback NMOS
clamp and the holding voltage VH of the snapback NMOS. The last is considered
a worst-case scenario. Thus, for an NMOS driver with a total gate width WG , the
resistor value should be above the estimated value:
ROUT ∼ (VT1 − VH ) / (WG × jCR ) . (4.1)
In a 5 V CMOS process with VT1 = 10 V, VH = 6 V, WG = 500 μm, and jCR =

0.1 mA/μm, a safe value for ROUT is >80 . In the case of practical design, one of
the possible sources of mistake of such an estimate for relatively large NMOS arrays
is a non-uniform transient current distribution that may result in local triggering of
the array below the estimated total current level. The total current level can also be
impacted by local process-related inhomogeneities formed in the array.
An alternative solution is illustrated in Fig. 5.18b, where the output driver NMOS
is modified according to the same rules that are used to design a snapback NMOS
cell: a ballasting region is added into the device to produce reversible snapback
operation. For the same purpose, the design of the device also includes maximum
well tap to source spacing.
As a result of such changes, the output NMOS driver can reversibly operate in
snapback mode with drain current saturation. In this case, the design can support two
variations, depending on the size of the output driver. If the device width exceeds
∼200 μm, then if properly designed, it can conduct an ESD current level and thus a
local clamp at the pad is not required. In the opposite case, the clamp can conduct the
excess ESD current. The only rule in implementing such an operation is to provide
a sufficient saturation region of the snapback characteristics to make sure that the
triggering voltage level of the local clamp is achieved before snapback and burnout
of the output NMOS.
The self-protecting NMOS solution is very efficient. In most cases, the silicide
blocking region or drain-contact-to-gate region for non-silicided processes of only
a 2–6 μm long is usually required. Thus, the total accumulated resistance in the
drain path is rather low. For example, in the case of sheet resistance of ∼100
/square for an n+ -diffusion region, even a small NMOS device with a 20 μm
width accumulates additional drain resistance of only 10 for 2 μm long drain
ballasting region. At the same time, such a saturation region is capable of providing
a sufficient saturation to engage the local ESD clamp before critical conditions for
burnout.
5.1.8 Array-Level Consideration
Certain changes can be made to the NMOS output device layout in order to support
the chosen strategy of ESD protection.
a) b) c)
Fig. 5.19 Versions of output NMOS array design with an arbitrarily placed non-butted Pwell
contact diffusion region (a), with butted diffusion region (b), and with interdigitated PLUS
region (c)
If the output NMOS is expected to work in snapback, then, in addition to the

drain ballasting region, an appropriate maximum spacing between the Pwell contact
and the n+ -source diffusion should be provided in order to improve the snapback
operation mode with the higher internal resistance of the parasitic NPN. Device
topology for this case is presented in Fig. 5.19a.
However, if the protection relies on the local snapback clamp, then increase of the
triggering current of the NMOS driver before snapback is desirable. In some cases,
a greater gate length can be selected, which will also provide a higher triggering
voltage.
The purpose of modification is to provide the lowest possible internal substrate
resistance to reduce the base resistance of the parasitic NPN. The simplest solution
is to provide each array finger with individual P-WELL contacts butted to the source
(Fig. 5.19b).
One of the topologies that provide the lowest substrate resistance with small
footprint penalties for an NMOS array can be implemented by the introduction of
interdigitated p+ -regions, as shown in Fig. 5.19c. In this case, contacts to the p+
-regions are shared with the source contacts, and the internal base resistance of the
parasitic NPN is minimal.
A similar measure also improves the parameters of NMOS arrays used in output
circuits that work on the reactive load in switching voltage regulators. In this case,
during the load mismatch short-term voltage overstress of the drain, the high voltage
may result in an accidental triggering and burnout of the output NMOS device.
LP
(a) (b)
HBM 2kV,W = 500um

12
LPG [μm]
11
DRAIN VOLTAGE (V)
–0.1
10
9
0.2
8
7
0.5
6
0.8
5
1.3
4
10–10 10 –9
10–8 10–7
TIME (s)
LAT. TEMP.(K)
800
600
400
200
c)
Fig. 5.20 Three-dimensional segment of the power array that represents the physical effect of the
p+ -contact as a function of p+ -to-polyspacing (LP ) (a), the example of net doping profiles for
LP = –0.1 μm (b), and the results of transient mixed-mode 3D analysis for ESD pulse operation
of this segment, with area scaling factor 500 at 2 kV HBM. The dependance of waveforms of the
drain voltage upon LP (c) demonstrate the high holding voltage of the array
Summary of a 3D numerical simulation of this physical effect is presented in

Fig. 5.20. As a physical representation of a periodic array, a 1 μm wide 3D seg-
ment of an NMOS device was selected (Fig. 5.20a). The segment was numerically
simulated with mirror boundary conditions. Variable spacing between the embed-
ded P-region and polygate LP has been used to demonstrate the effect. An extreme
case is obtained when the p+ implant mask overlaps with the polygate region LP =
–0.1 μm in the device (Fig. 5.20b).
The calculated transient waveforms for a 2 kV HBM operation of the ESD HBM
in mixed mode are presented in Fig. 5.20c for different values of the LP parameter.
At relatively large p+ -to-polygate spacing, the device exhibits typical snapback

operation with the holding voltage equal to ∼4.5 V. With LP reduction, the hold-
ing voltage of the device increases and, for the case of p+ -implant to polyoverlap,
the device provides a practically non-snapback operation with a bell-shaped wave-
form. This effect is the result of high efficiency in the collection of the holes from
the source–drain spacing, which produces a corresponding misbalance between the
electron and hole currents.
5.1.9 Concept of Two-Stage Protection

Practically, the two-stage approach has already been illustrated above each time
separation resistors are introduced between the protected pin and active circuitry.
This section just provides a formal summary of the two-stage approach.
5.1.9.1 CMOS Input

Two-stage ESD protection is a main approach for the protection of CMOS inputs.
The CMOS input is very sensitive to the ESD discharge, due to the relatively low
voltage and current of gate dielectric breakdown. The gate dielectric breakdown
is the major limiting factor for the passing level of charged device model (CDM)
testing for high pin count products implemented in scaled CMOS processes with
gate lengths 0.18 μm and below. At the same time, in the case of 5 V processes, no
special consideration is required because the pulse breakdown voltage of the gate
oxide is above the triggering voltage of the snapback clamps. The rail-based solution
is exceptionally robust in the case of an active power clamp.
Additional input resistors can be used to separate the inputs and realize two-
stage ESD protection using a set of diodes. The principle of the two-stage network
is widely used in system-level ESD protection (Chapter 8). In this case, the first
stage is formed by the discrete suppressor component.
In principle, the two-stage network consists of a combination of two ESD clamps
separated by a resistive current path or delay component. The resistor-based two-
stage network is presented in Fig. 5.21. In this case, the first clamp limits the
operational voltage at VP , while the much smaller second-stage clamp limits the
voltage in the internal node at VI . The second-stage resistor RS is selected depending
on the second-stage clamp current IC2 : RS = (VP –VI )/IC2 .
The second-stage clamp is very efficient in filtering short pulse overstress typical
for CDM events due to the RS CI time constant; CI is the equivalent capacitance of
the internal node.
The first clamp directly interfaces with the input PAD. It limits the voltage and
conducts the majority of ESD current, but is not designed to precisely target the
ESD protection window.
The advantage of the first-stage clamp is that usually the clamp can reversibly
conduct above approximately three times larger current at an additional voltage drop
above the desirable voltage limit within ESD protection window. For example, n+
-Pwell and p+ -Nwell diodes implemented within the minimum process design rules
VP RS VI INTERNAL
PAD
NODE
IC1 IC2
First Second
Stage Stage
Clamp Clamp
ESDMINUS
Fig. 5.21 Two-stage ESD protection for CMOS input
of a 0.5 μm process can deliver several times more current at an ∼5 V drop than at
a 1 V voltage drop.
Often, implementation of a much bigger diode cell is a rather poor alternative
due to its accompanied increase in parasitic capacitance, leakage, and noise.
Thus, the first-stage clamp is used to transform the ESD pulse into a waveform
with a much lower voltage amplitude of ∼5–7 V. Then, much smaller second-stage
diodes limit the voltage to ∼1 V at the current level limited by the stage resistor RS ,
which provides the remaining 4 V voltage drop.
Often, it is advantageous to implement RS as a saturation resistor. However, due
to the resulting resistor parasitics in high-speed I/O, a polyresistor provides a more
predictable result.
5.1.9.2 Diode-Based Compact Two-Stage ESD Protection Circuit

Principles similar to the component-level ESD protection network can be imple-
mented at the device level. A compact solution is presented as a three-terminal diode
structure (Fig. 5.22). It incorporates two-sided diodes that form a three-terminal
structure. One side of the diode is used for power ESD operation (a power diode
connected to the I/O pad) and the other side just holds the lower potential in the
internal circuitry (a potential diode connected to the internal circuitry). Both inter-
nal diodes share the same well, thus creating a two-stage ESD protection circuit
with a corresponding “built-in” internal well resistor. A key important feature of the
diode structure is the internal potential sharing during ESD operation.
Due to the realized internal potential sharing, a potential limiting is possible on
the internal circuitry. The ESD protection of the I/O pin can be realized using a
corresponding circuit presented in Fig. 5.23.
This approach is advantageous for a small footprint, integrated high-speed I/O
compatibility due to the reduced equivalent capacitance.
Operation of the two-stage components can be analyzed using numerical device
simulation of the cross sections generated by physical process simulation tool
(Fig. 5.24a, b). During the high-current ESD event and the high-voltage drop on
the pad side contact, the internal node contact potential remains rather low – 1–2 V
only.
C1 ANODE C2 A1 CATHODE A2
N+ STI P+ STI N+ P+ STI N+ STI P+
Nwell Pwell
P-Substrate P-Substrate
a) b)
C
C1 C2
A1 A2
A
c) d)
Fig. 5.22 Three-terminal diode structures for N-well (a) and P-well (b) replacement CMOS
diodes used in the suggested ESD protection circuit for I/O cells. The simplified equivalent circuits
of the three-terminal diode structures (c, d)
The next level of numerical validation can be achieved by the thermal-coupled

mixed-mode simulation of different ESD HBM pulse levels (Fig. 5.24e, f).
According to these results, ESD operation of the diode structures demonstrates that a
maximum lattice temperature near critical values for the structure melting is reached
under the power diode voltage level of ∼6 V. At the same time, the internal terminal
holds the voltage level below 2 V in this range (Fig. 5.24e, f).
VDD/ESDPLUS BUS
ESD PULSE C
A1 A2
ESD
Clamp
PAD
C1 C2
A
VSS/ESDMINUS BUS/GND
Fig. 5.23 I/O circuit ESD protection concept using the three-terminal diode structures for rail-
based protection
a) b)
PAD and INPUT VOLTAGE (V)

0.00
PAD and INPUT VOLTAGE (V)
Pwell DIODE STRUCTURE Nwell DIODE STRUCTURE

–1.00
6.00
–2.00
4.00 –3.00
–4.00
2.00
–5.00
0.00 –6.00
0.0 0.1 0.2 0.3 0.4 0.5 –0.4 –0.2 0.0
c) PAD CURRENT (A/um)
d)
PAD CURRENT (A/um)
Pwell STI diode W = 25um

"O" - PAD and "Δ"- INPUT VOLTAGE (V)
dot line - LAT. TEMPERATURE (K)

1800
"O" - PAD and "Δ"- INPUT VOLTAGE (V)
dot line - LAT. TEMPERATURE (K)
Nwell STI diode W = 25um 6

–7 1800
1600
–6 1600 5 1400
–5 1400
4 1200
–4 1200
3 1000
1000
–3
2 800
800
–2
600 600
1
–1
400 400
0 0
10–9 10–8
10 –7
10–9 10–8
10 –7
TIME (S) TIME (S)
e) f)
Fig. 5.24 Cross section of the N-well (a) and P-well (b) diode structures obtained from process
simulation and corresponding I–V dependencies with the “INPUT” terminal potential (c, d) and
results of numerical experiments for different amplitude HBM ESD operations: The pad and input
voltages and maximum temperature for the conventional and proposed triggering clamps on the
basis N-well (e) and P-well (f) two-stage diodes, respectively
5.1.9.3 Two-Stage ESD Protection Circuit for BJT Base

A similar protection approach can be implemented for the sensitive-to-voltage
overstress BJT input. The cross section for the BJT device and input circuit is pre-
sented in Fig. 5.25a, d. This solution is an alternative to the diode cell protection
(Fig. 5.25b).
EMITTER
COLLECT BASE N+ BASE

ESD PULSE
P-BASE
N-SINKER
STI STI
N-COLLECTOR
I/O
N-EPI PAD
NBL
P-SUBSTRATE GND
a) b)
ESD PULSE
B1 B2 B2
B1
I/O
PAD
C E
C E
GND
c) d)
Fig. 5.25 Two-stage protection of the base–emitter junction: input NPN BJT cross section (a);
typical protection circuit with diode (b); equivalent circuit for double base configuration (c); and
proposed operation circuit (d)
To realize two-stage protection, two base contact regions from the opposite side
of the emitter can be used to implement two-stage component (Fig. 5.25c) and the
corresponding network (Fig. 5.25d). In this case, the base sheet resistance under the
emitter region plays the role of the second-stage resistor clamp component. This
internal base resistance is relatively small.
However, according to numerical simulation results, the effect is substantial. The
numerical simulation has been performed using process-generated cross sections for
a SiGe NPN device (Fig. 5.26a). According to the numerical analysis, the input side
of the base remains at a potential of ∼1 V in the wide range of the input ESD current
even up to temperatures above the critical level for local device melting.
5.1.9.4 Two-Stage Network with Snapback NMOS

Similar principles can be applied to the snapback NMOS protection structure. The
solution requires an additional potential contact that will sense the voltage from
some location in the ballasting region.
The ballasting region of the NMOS already represents a diffusion resistor that
operates in deep saturation mode in the case of snapback. Thus, if the internal node
potential is sensed closer to the gate of the NMOS device, the voltage limiting in the
internal node might be achieved at a significantly lower level.
"O"- BASE and "Δ"- B2 VOLTAGE (V)

3
2kV HBM
2 80um
VB
1
VB2
0
10–9 10–8 10–7 10–6
TIME (S)
a) b)
Fig. 5.26 Process simulation cross section of SiGe NPN BJT (a) and mixed-mode simulation
results for voltage at pad side VB and internal node side base contacts VB2 for 2 kV HBM ESD
pulse for different structure width (b)
Thus, the separation of the “input” and “output” voltages is realized on the device
level in the snapback NMOS structure. In order to make this possible, an extra
“potential-sensing” terminal is introduced between the drain metal (silicide) and
the gate (shown in Fig. 5.27a, b). Figure 5.27c presents the suggested local clamp
for ESD protection.
During ESD operation with a high-current pulse, the voltage drop on the ballast-
ing region results in a corresponding electric potential drop in the drain ballasting
New Local Contact D2

GATE
BULK SOURCE DRAIN
P+ STI N+ N+
Pwell
P-SUBSTRATE
a)
ESD PULSE VDD
DRAIN
D1
D2 D2
PAD
BULK G
S
GATE SOURCE
VSS/GND
b) c)
Fig. 5.27 Suggested GGNMOS with local potential terminal (D2) (a), its simplified equivalent
circuit (b), and suggested local clamp implementation (c)
region. Thus, the voltage at the added contact depends on its position inside ballast-
ing region and, in general, is significantly lower than that at the drain contact and,
at least, does not include the voltage drop on the interconnects and contacts. A key
important feature of the structure is the internal potential distribution during power
(ESD) operation and is illustrated below using TCAD tools.
A side effect (or trade-off) of the implementation of this circuit is the resis-
tor between the pad and the internal circuitry. However, due to a rather low n+
-composite sheet resistance of <80 /square, the estimated resistor value is usually
negligible below 1–10 , depending on device dimensions.
At the same time, the “benefits” from the presented solution are rather significant.
Based upon the distribution of the electric potential inside the snapback NMOS
structure during high-current ESD pulse, low voltages can be sensed (Fig. 5.27)
closer to the gate region. Ultimately, the voltage level is comparable with the holding
voltage. If the internal circuitry is connected to such terminal, it remains protected at
a low 5–6 V level independently from the voltage drop on the clamp at high current,
thus eliminating the risk of gate oxide damage (Fig. 5.28).
a)
"O" - D RAIN "Δ"- D2 VOLTAGE (V)
dotted line - LAT. TEMPERATURE (K)
"O" - DRAIN "Δ"- D2 VOLTAGE (V)
HBM = 5kV 13 HBM = 5kV W=200um

15 1800
TOTAL WIDTH (um) 200um 12
1600 4.9um
11
1400 3.9um
10 400um 10 2.9um
1200
9
1000 1.4um
8 0.4um
5 800
7
600
6
400 POTENTIAL CONTACT-GATE SPACING
0 5
10–9 10–8 10–7 10–8 10–7
TIME (S) TIME (S)
b) c)
Fig. 5.28 Process simulation cross section of the snapback NMOS (a); results of the numerical
simulation for a 5 kV HBM ESD pulse operation with different structure widths (b); and the voltage
on internal circuit side contact “D2” depending on its position relative to the gate edge (c)
5.2 Local Clamp-Based ESD Protection Network 247
In practical design, in order to avoid interference of the sense contact metal with
the ballasting region, a separate local n+ -diffusion extension can be made at the side
of each finger to accommodate an additional electrode spaced away from the active
region.
5.2 Local Clamp-Based ESD Protection Network
5.2.1 Local ESD Protection

There are many analog circuit blocks that, in principle, cannot tolerate the rail-based
ESD protection.
One of the typical examples of when a rail-based ESD protection network with
diode clamps is not possible is the case of a back-drive compatible pin. The pin must
be independent from power and thus the diode-to-ESDPLUS rail configuration is
not possible.
A similar case is that of over-voltage input pins with thin film resistors. An
example of an automotive application amplifier realized in a 5 V process with a
65 V-tolerant differential input is presented in [42].
Another example includes circuit pins with high dual-direction voltage tolerance.
Protection solutions for such pins may not tolerate the lower diode-to-ground and
may require excluding the upper diode-to-power supply rail as well.
In such cases, local ESD protection is used as either the only protection approach
or combined with partial rail-based protection.
An example of a local ESD protection network is presented in Fig. 5.29. In prin-
ciple, every pin in the network is protected by a local ESD clamp that provides a
voltage tolerance suitable for the specified pin parameters.
The previously discussed two-stage ESD protection is similarly applicable for
local clamp protection. One of the most typical ways to implement two-stage ESD
INPUT OUTPUT POWER

Snapback Clamp
Snapback Clamp
Snapback Clamp
R R R
GND
Fig. 5.29 Example of the four-pin local-based protection network

PADRS INTERNAL
NODE
Snapback Clamp
Avalanche Diode
R
ESDMINUS
Fig. 5.30 Local-based network with the second-stage avalanche diode
protection is to include a snapback clamp separation resistor and a voltage-limiting

avalanche diode (Fig. 5.30).
The complexity of local ESD protection is related to the generally unknown con-
ditions of the control electrodes of the devices interfacing with the pad during ESD
pulse. As a result of this the triggering voltage of the internal devices can become
comparable to the triggering voltage characteristics of the ESD clamp.
5.2.2 Serial Data Line Pin Case Study
One of the cases for a CMOS pin protection can be illustrated with the serial data
line (SDA) pin for the I2 C-bus [105]. An open drain NMOS MDR with a pull-
up capacitor CU between the gate and drain is required for circuit functionality
(Fig. 5.31).
The open drain device MDR is laid out with minimum spacing between the source
and the well to increase the snapback triggering voltage. Nevertheless, the pull-up
capacitor CU ensures that the device remains in the on-state at the beginning of an
ESD pulse before the ESD MESD triggers into high-conductivity mode. As a result,
local burnout of the driver device MDR is observed (Fig. 5.31).
A possible ESD protection solution for this pin is presented in Fig. 5.32. An addi-
tional resistor can be added between the two NMOS devices to increase the voltage
level at the pad required for MDR snapback (Fig. 5.32a). Additional measures can
include engineering of the gate circuit in order to keep the MDR gate at low potential
during ESD stress until an ESD pulse sufficient to initiate snapback of the clamp,
thus preventing the snapback of MDR . Such a circuit can be realized by the addi-
tion of active components MK1 and MK2 , referenced to the power supply VDD node
(Fig. 5.32b).
The “keep-off” circuit relies on the additional capacitance CK during ESD stress,
while VDD node is assumed at low potential. This results in an off-state MK1 and on-
state MK2 during the transient rise time of the ESD pulse. After ESD NMOS clamp
MESD is engaged, the low holding voltage guarantees non-snapback operation of
MDR .
I/O PAD C1 To
1.5p Internal
MESD Circuit
500/0.5 MDR
RM SWS=10 100/0.5
ESDM 20K SWS=1
ESD Clamp
butted well tie source

source
drain
source
Fig. 5.31 Example of local protection for serial data line pin and the results of failure analysis
50 Ω
1.5pF
W = 480 μm W = 100 μm
Desired ESD current path
20 kΩ
ESD current path,
limited by 50 Ω
a)
Source to well > 5μm Source to well > 1.3μm
50 Ω VDD
CK
MESD MDR CU 0.5pF
1.5pF
W = 480 μm W = 100 μm
20 kΩ
Mk2
Mk1
b)
Source to well > 5μm Source to well > 1.3μm
Desired ESD current path
Fig. 5.32 Solution for SDA pin protection using a separation resistor (a) and additional “keep-off”
circuit (b)
At normal operation with VDD signal, the keep-off circuit is disabled by VDD
signal.
It is important that, unlike in active clamps, the required RC time constant pro-
vided by CK and the parasitic resistance of MK1 is rather low (∼10–20 ns). This
time is usually sufficient to trigger the snapback NMOS into avalanche–injection

conductivity modulation mode.
5.2.3 Erase Pin Protection in EEPROM

One of the typical symptoms of inadequate protection of the “Erase” pin in
EEPROM memory modules with an external high-voltage supply is functional fail-
ure. This failure is observed as a reprogramming of the memory cells by the ESD
pulse, rather than ESD damage. This event has been experimentally observed, even
though the programming normally happens only at some specific voltage combi-
nation on the “erase” and “enable” pins. However, during transient ESD events,
some pins and internal nodes can be floating, thus creating conditions for functional
failure due to reprogramming.
Thus, an ESD protection solution is required not only to provide an appropriate
voltage limiting during the initial part of the ESD stress but also to make sure there
is no high voltage left at the pin as a result of the stress.
For a 0.5 μm CMOS process, the typical erase pin voltage level is 12–20 V,
depending on gate oxide tunneling characteristics.
In the case of a small footprint solution, first-stage protection includes a cor-
responding high-voltage snapback device. However, a single snapback device
is insufficient, due to the relatively high residual voltage after snapback [62].
Therefore, a proper solution requires an additional high-voltage device. The first
version of the local ESD protection includes snapback SCR device (Fig. 5.33a). The
solution relies on the additional voltage clamping provided by the snapback NMOS,
in addition to the SCR.
However, the success of the application of this solution depends on the size of
the memory array. In a single cell, the residual voltage is still rather high (Fig. 5.34),
and this leads to reprogramming of the single cell array. The most important exper-
imental fact is that there is no discharge path in the pin after turn-off. The HBM
waveform measurements demonstrate long-term storing of the residual voltage after
ESD stress (Fig. 5.34).
The circuit has been modified to eliminate the high residual voltage on the erase
pin after the ESD event (Fig. 5.33b).
In order to add the discharge current path required after SCR turn-off, an addi-
tional large resistor RB is added between the pin and ground, and a large resistor
R2 is added between the pin and the memory. Finally, the second-stage clamp is
redesigned to provide a current path after the SCR turns off (Fig. 5.33c).
A critical parameter is the equivalent capacitance of the connected memory cell.
If only one cell is connected, the estimated capacitive load is ∼20 fF. For the
connected ∼50 cells, the total capacitive load is ∼1000 fF.
The capacitive load affects waveforms observed under ESD pulse conditions. In
the case of the low capacitive load provided by a single cell, the residual voltage is
above the erase level, while higher capacitive load of several cells provides a much
lower residual voltage (Fig. 5.34). The following measures can be taken to guarantee
ERASE ERASE
INTERNAL
HVSCR
EEPROM
ARRAY
ESDM a)
R2
ERASE ERASE
CD INTERNAL
HVSCR NLDMOS
EEPROM
NLDMOS ARRAY
RB
ESDM
b)
R1 R2
ERASE ERASE
INTERNAL
R3
NLDMOS
HVSCR EEPROM
NLDMOS ARRAY
ESDM
ENABLE
c)
VDD
Fig. 5.33 Original (a) and modified (b, c) EEPROM erase pin ESD protection clamp circuit
low sensitivity to overstress: provide a discharge path and use clamps with a lower
turn-on voltage.
An alternative, more compact circuit solution is presented in Fig. 5.35. This
circuit is based on an NLDMOS-SCR with implemented gate circuit control. The
principle of operation for this circuit can be explained as follows: the ESD device
provides discharge of the majority of ESD current (ISN ) in snapback mode similar
to a typical high-voltage NLDMOS SCR clamp with a grounded gate. However,
due to the gate biasing circuit, the self-turn-off at a corresponding minimum critical
current brings the device into a condition with channel current (ICH ) rather than into
the low leakage state.
15
HBM 2kV
PULSED VOLTAGE (V)

10
Single EEPROM cell
48 EEPROM cells
5
0 200 400 600 800 1000

TIME (ns)
PULSED CURRENT (A)
0 200 400 600 800 1000

1.5
1.0
0.5
0.0
Fig. 5.34 Voltage and current waveform for the EEPROM with single cell and 48 cell arrays
measured at 2 kV HBM
ERASE R0
PAD
to Internal
LDSCR
“ERASE” node
C1
R1
M1
to Internal VDD
GND for Shutdown
PAD
Fig. 5.35 Alternative small footprint EEPROM erase pin protection solution
The transient gate bias is supplied by the R1 C1 circuit connected after the sep-
aration resistor R0 (∼100 K). The R1 C1 time constant is less than ∼1 μs. During
this additional time, the gate of the ESD device is held above the threshold VT
level. In these conditions, the channel current ICH of the device provides the path
for the remaining current. The discharge continues until the internal node volt-
age drops below the threshold voltage VT or until the RC delay time constant is
reached. The critical minimum current in snapback mode can be as low as ISN ∼W
× 0.1 mA/μm∼20 mA. After the majority of the ESD pulse is discharged through
the open NLDMOS, the channel current is approximately constant for the rest of the
pulse.
The circuit provides an appropriate functionality for a low-voltage Electrical
Overstress (EOS) as well. In this case, at overstress below the snapback voltage
level, the ESD device remains in the off-state while the LDMOS channel current
path is supported.
At normal operation conditions, the VDD signal and small NMOS device M1
keeps the ESD device gate at low potential. The capacitor C1 is fully charged and
no parasitic leakage current path is present.
Thus, this alternative clamp solves the EEPROM protection problem with a
minimum set of components in a very space-efficient way. In this case, the
ESD device is used both as a snapback and as an active component, provid-
ing operation in conductivity modulation and monopolar current conduction,
respectively.
5.2.4 Local Protection of the Internal Pins

One of the most typical examples of the impact of ESD stress on the internal pins
in high-voltage circuits is the internal voltage regulator of low-side control circuits,
logic, and drivers for switching devices. Because of the relatively low impedance
of internal regulators, a second-stage resistor cannot be used in order to implement
two-stage protection. As a result, the regulators are usually rather sensitive to ESD
damage and must be designed to withstand certain ESD current levels. Usually, the
internally generated VDD voltage node requires an internal voltage clamp capable
of handling high ESD current. Such a clamp is used even if the VDD node is not
connected to the external pin at all.
The most typical locations of failure in internal voltage regulator design are the
high-side high-power components. In the high-side PMOS-based regulator, namely
the high-voltage PMOS fails (Fig. 5.36a). In this case, the gate clamping diode may
not provide a substantial improvement (Fig. 5.36b).
A similar ESD current induced failure can be observed in bipolar and stacked reg-
ulators. In this case, due to the domino effect, damage in both stacked components
might be observed (Fig. 5.36c).
ESD protection of the high-voltage PMOS used in VCC regulators (Fig. 5.36) is
based upon the expected ESD path formation between the VIN and BOOST pins.
In general, there are three current paths that could be realized for ESD pulse current
(Fig. 5.37).
(i) The first current path is provided by the large switching NLDMOS controlled
by the driver. The NLDMOS turns on for a short period of time due to the
VIN
MP
DZ
MN
a)
VIN
VIN
VCC
VCC
b) c)
Fig. 5.36 Examples of the FA results for PMOS (a, b) and NPN BJT-based VCC regulators (c)
[105]
drain–gate coupling that may be sufficient enough to discharge the ESD cur-
rent. This current is directed to the switch pin SW, followed by the reverse
current path through the BOOST-SW clamps. The clamp’s reverse path ESD
diode provides this remaining current path. This current path should not cause
a failure in properly designed ESD protection clamps.
(ii) The second alternative current path is realized through the high-voltage ESD
clamp to the power ground, followed by the ESD diode between the BOOST
and ground. This current path is also reversible.
(iii) Finally, the third, and generally irreversible, current path is provided by the
high-voltage PMOS itself and other stacked components of the regulator (for
example, the NPN diode shown in Fig. 5.37) directly to BOOST. If this current
path is dominant, the PMOS is exposed to failure. This current path explains
the damage demonstrated by Fig. 5.36a, b.
Thus, in the case of a low level of array coupling and a small ESD protection
window, ESD failures are related to the internal VCC regulator. The ESD protection
BOOST
MNH 3
VIN
PM
SW
EVIN
1
VCC
ESW
EVCC
Fig. 5.37 Simplified circuit of the DC–DC voltage regulators with three possible ESD current
path scenarios for VIN to BOOST ESD zap
network is very product-specific. There are several improvements that can be done
to the circuit. However, in general, the circuit is still rather sensitive to process
technology variation due to the following reasons: triggering voltage variation in the
high-voltage ESD protection VIN clamp, dependence of the NLDMOS array self-
protection capabilities upon layout, and internal circuit driver design. In addition,
variation of the pulsed SOA of high-voltage PMOS and NLDMOS devices may
provide an undesirable effect.
Design measures to improve the internal VCC regulator are based upon under-
standing of the current path alternatives listed above. One of the major assumptions
is that in the general case, the current path can be shared and thus gradual
improvement of all current path components may provide a desired effect.
The specific way to improve ESD protection depends on the dominant ESD
current path realized in the particular product circuit. Perhaps, the first, most logi-
cal solution is to improve ESD protection by reducing the VIN clamping voltage
to a lower level, thus limiting the voltage drop on the PMOS device. However,
this adjustment of the parameters of the high-voltage ESD protection clamp might
not be an option in case of relatively low SOA margins of the power-optimized
components. In this case, the most robust approach is to implement ESD-oriented
co-design of the VCC regulator, which provides a much higher pulsed absolute max-
imum voltage for the circuit pins. Design measures may include stacked or oversized
components or application of less power-efficient components with higher voltage
tolerance, if allowed by the process.
If the failed devices belong to the low side driver, then an additional internal
power clamp, for example, the snapback NMOS clamp, may provide the solution.
This clamp may replace the low-voltage avalanche diode between the VDD (VCC)
and ground (Fig. 5.38). Even though the VCC node is internal, the additional EVCC
clamp limits the critical voltage on the VCC node and provides a reverse current
path.
BOOST
VIN
PM
SW
EVIN
VCC
ESW
EVCC
Fig. 5.38 Increase of local ESD protection robustness by application of the addition internal VCC
to the power ground clamps
Another typical fix may not be related to the active area of the devices at all, but
instead requires improvement of the metal connection inside the regulator to support
the high current path. An example of the simple improvement effect of drain–source
metallization is demonstrated in Fig. 5.39. In the layout of Fig. 5.39a, the device fails
in the expected current crowding spot in the middle of the structure. A more uniform
current distribution with metallization (Fig. 5.39b) eliminates the problem.
a) b)
Fig. 5.39 Original metallization of the failed high-voltage PMOS with FA photo insert indicating
the damage (a) and the successfully improved (b) device
5.2.5 Local Protection of the High-Speed I/O pins

High-speed I/O pins are usually co-designed with a printed circuit board (PCB) to
work in a wide high-frequency band. In this case, the connection to the microchip
should be treated as a transmission line micro-strip network. In such a network,

matching of the load cannot be neglected [106].
From the ESD solution point of view, signal propagation and reflections must be
considered and matched to provide proper circuit operation with the implemented
ESD protection network.
The relationship between the incident Vinc and reflected voltage signal Vrefl in the
transmission line, with impedance Z0 and working on resistance Rl , is as follows:

Rl − Z0
Vrefl = Vinc = Vinc , (4.2)
Rl + Z 0
where is the reflection coefficient.

Three common cases of the transmission line on a matched load, open and short
circuits are presented in Fig. 5.40.
Fig. 5.40 Transmission line reflections [106]
ESD network design for high-speed voltage-mode drivers is generally compatible

with NMOS or diode protection. Driver stacks can be difficult to configure with low-
impedance drivers. Overall, driver resistance requirements limit the resistor value.
Additional device width and parallel R are required to achive the appropriate passing
level (Fig. 5.41).
VDD
Gate or
Substrate
Control
a) b)
Fig. 5.41 Local ESD protection for voltage (a) and current (b) mode drivers with matching
impedance resistors [106]
ESD for current mode drivers is also generally compatible with NMOS or diode
protection. The current source in this case is usually a transistor stack, which pro-
vides higher ESD immunity. Termination Rterm is usually calibrated, and transistor
and resistor combinations must meet resistance limits.
Receiver design for a CMOS receiver with a standard inverter requires the bal-
ance of the NMOS and PMOS to ensure proper switching characteristics over the
range of process variations, voltage, and temperature.
The major principle for successful high-speed circuit design is that the intercon-
nects, driver, receiver, clocking, and ESD protection components must all function
collectively. For greater speed, the interconnect channel and termination must be as
clean as possible and must not provide any reflections, overshoots, or undershoots.
Receiver capacitance provides a significant limitation on signal integrity. Thus,
the ESD protection solution requires a significant optimization of the ESD diodes
and clamps to lower parasitics. This is achieved by metallization optimization,
though even more sophisticated circuits involve auto-biasing of the ESD diodes
to eliminate the parasitic capacitance effect. One of the methods is illustrated in
Fig. 5.42.
Fig. 5.42 ESD protection for a high-speed receiver with a two-stage matching resistor
5.3 ESD Network for Multiple Voltage Domains
5.3.1 Multiple Voltage Domains
When the analog circuit has different voltage domains, the ESD protection network
can be constructed using several approaches that depend on circuit specifications.
The most typical example of a multiple voltage domain circuit is in the circuit
with the digital, analog, and power domains. Since the ESD protection of a chip is
required to support each pin-to-pin combination, all domains should be connected
into a network.
Consider an example of analog and power domains. The inter-domain connec-
tion in the network is usually realized between the power and analog grounds.
The simple reason for this arrangement is that the power supply voltages cannot
5.3 ESD Network for Multiple Voltage Domains 259
VDD1 VDD2
D1 D2 D1 D2
Power Clamp 2
Power Clamp 1
INP OUT INP OUT
D4 D3 D4 D3
GND1
a)
GND2
VDD1 VDD2
D1 D2 D1 D2
Power Clamp 12
Power Clamp 12
Power Clamp 1
Power Clamp 2
INP OUT INP OUT
D4 D3 D4 D3
b)
GND1 GND2
Fig. 5.43 ESD protection network for multiple voltage domains with a back-to-back diode clamp
(a) and a cross-coupled power clamp (b)
be shared. To avoid cross-talk due to the potential drop on the ground busses, the
grounds can be connected using a back-to-back diode clamp. This clamp decouples
the ground buses if the voltage drop is less than the voltage drop on the forward-
biased diodes and, at the same time, will provide a low voltage drop at ESD current
level (Fig. 5.43a). In this case, the stress on the pins from the different domains will
include the additional voltage drop on the diode.
An alternative solution may use a cross-coupled power clamp between the
domains (Fig. 5.43b).
A similar approach is used in the case of a two-voltage domain with a distributed
active clamp network (Fig. 5.44).
Fig. 5.44 ESD protection network for two-voltage domains with separate distributed active clamp
[104]
5.3.2 Protection of Multiple Voltage Domains with Single Active

Clamp Network
In small pin count, low-power analog circuits with multiple voltage domains, the
protection network often needs to be implemented using only one active clamp
domain. In this case, a common ground can be used for all the power sup-
ply domains. The principles for a successful implementation of this structure are
summarized below.
The active clamp ESDMINUS rail is connected to the ground of both power
supplies.
The ESDPLUS rail of the active clamp network is connected to the highest volt-
age power supply. If both supplies have the same voltage level, then the ESDPLUS
rail should be connected to the power supply whose domain creates less bus noise.
Bus noise is created due to a voltage drop on the power rails and bond wires induc-
tive load as a result of change in the current consumed by the corresponding circuit
blocks (Ldi/dt). For example, in the case of low-power analog and large digital
domains, the analog domain might create less bus voltage noise, while in the case of
power analog domain with a switching function the situation might be the opposite.
The bus voltage noise can be propagated back to the inputs of the circuit due
to finite parasitic capacitance of the ESD diodes and thus might impact the perfor-
mance of the analog circuit. The effect of bus voltage noise should be considered
even if both grounds or both power supplies are connected to the same package pin
in the final package. This major consideration is related to the bond wire inductance.
The main reason why one of the power supply domains needs to be connected to
ESDPLUS rail is related to the parasitic PNP structure. This structure is formed by
the p+ -region of the Nwell ESD IO diode acting as an emitter, the N-well region
of the same ESD diode acting as a base connected to the ESDPLUS, and the P-
substrate connected to ESDMINUS acting as a collector. When the ESDPLUS is
not connected to the power supply, this parasitic PNP can be turned on and thus
might produce a significant leakage path to the ground, interfering with the power
5.3 ESD Network for Multiple Voltage Domains 261
supply. As a compromise, the ESDPLUS can be connected through an additional

resistor to suppress the parasitic action of the parasitic p–n–p.
Another important consideration in such active clamp protection networks is
specification of power supply operation conditions. Usually, an active clamp solu-
tion is used when power supply ramping is rather slow (100 ms). Typically, starting
from a 50 μs rise time, a parasitic current through the active clamp is already
unacceptable for most designs. A major limitation of active clamp application is
in circuits with a “power-down” specification when IO pins remain connected to an
active data bus. Both fast multiple power-down events and the loading of the external
device can become undesirable and unacceptable. Thus, in the case of power-down
operation with an active data bus, preference should be given to snapback solutions.
When snapback solutions are used, one of the major considerations is transient
latch-up. Transient latch-up can be observed in certain cases when the holding volt-
age of the snapback is below the power supply voltage level. For example, in both
CMOS input and output as well as BJT input (Fig. 5.45), local ESD protection
with a holding voltage below the power supply level will not be able to provide a
latch-up current path. However, in the BJT output, for example, in the CML output
(Fig. 5.45), the latch-up consideration is rather critical. Due to termination of the 50
resistors (R), the snapback device with a holding voltage below power supply can
be turned on and then kept in high conductivity by the power supply VDD voltage,
thus creating a latch-up scenario.
VDD
R = 50 R = 50 OUTP
INPP OUTM
R = 100
INPM
Vbias
ESDM
Fig. 5.45 Illustration for BJT CML inputs and outputs
5.3.3 Local Bi-directional ESD Protection of Differential Input

The case of back-drive compatible pin is one of the examples when a rail-based ESD
protection network with diode clamps is not possible. The pin must be independent
from power and thus the diode-to-ESDPLUS rail is not suitable.
An example of multiple voltage domains with level shifts used in LCD applica-
tions is presented in Fig. 5.46. An example of a level-shifter ESD protection network
with dual-direction clamps for the display driver in presented in Fig. 5.46. This
V4 = 12V
V1 = 6V
AC
AC
AC
AC 3.3V VO =
VD = +/–6V
+12/–8V
AC AC SC
AC SC
V3 = –8V V2 = –6V a)
V4 = 12V
V1 = 6V
AC
AC
AC
AC 3.3V VO =
VD = +/–6V
+12/–8V
AC AC DIAC
AC
V3 = –8V V2 = –6V b)
Fig. 5.46 ESD protection for LCD driver using stacked NMOS (a) and DIAC protection for
output (b)
application is designed to operate at a substrate voltage of –8 V with power supply

pins of –8, –6, 0, 6, and 12 V.
The analog circuit is designed using 3.3 and 6 V isolated CMOS devices. The
ESD protection network combines five blocks with separate ESD clamps:
(i) Distributed active clamp (AC) blocks, with a clamp width of ∼500 μm,
Lg =0.8 μm, and an RC timer with RC = 4 μs
(ii) 6 V NMOS snapback clamp
(iii) Dual-direction ESD devices for 3.6 V inputs
(iv) Dual-direction ESD devices for 20 V inputs (level shifter)
The +6 V/–6 V-tolerant input pin VO is protected by the substrate-isolated stacked

snapback NMOS clamp. The isolation of the Nwell region is common for both
snapback clamps to enable smaller footprint for the stacked clamp. The Nwell is
connected to the +6 V power supply (Fig. 5.46a).
5.4 ESD Network Simulation with ESD Compact Models 263
Alternatively, this pin can be protected by the dual-direction DIAC ESD clamp
(Fig. 5.46b). The important parasitic resistance and parasitic diodes provided by the
internal circuit are illustrated in Fig. 5.46b by the dotted line. The additional voltage
drop and the alternative current path provided by these components of the network
should be accounted for.
5.4 ESD Network Simulation with ESD Compact Models

Significant progress in predictability of ESD design can be achieved by circuit
simulation using ESD compact models. The ESD compact models can be devel-
oped using standard device models and subcircuit components to model snapback
regimes.
Usually, the subcircuit components include parasitic BJT structures, an avalanche
current source, and saturation resistors. Development of the accurate snapback
model is generally not an easy task. Due to complicated extraction methods for
ESD current conditions, the construction of the scalable model requires advanced
expertise in the field. One of the most helpful ways of developing a snapback model
is to use the physical process and device simulation to generate the reference data
for model extraction. Such an approach provides a significant acceleration of the
extraction procedures while providing reasonable accuracy.
Conversely, the generated compact models can be used for both ESD and func-
tionality simulations of the large circuit, using standard circuit simulation tools such
as HSPICE and Spectre, as well as mixed-mode simulators, for example, using
DECIMMTM tool for this book examples.
There are several original research studies written in the field of snapback model
development [107–111]. For example, the snapback ESD compact model for an LV
voltage NMOS is presented in [111].
The model subcircuit uses the following components: a standard compact NMOS
device model, an avalanche current source, and a parasitic NPN structure. The snap-
back compact model for SCR-type devices requires an additional PNP structure.
With proper calibration, the model can be successfully used in simulation of ESD
protection circuits.
5.4.1 Compact Model for Snapback NMOS and PMOS Devices
The cross section and the equivalent circuit representing the low-voltage NMOS
device are presented in [111]. The basic cross section and equivalent circuit for the
snapback model are shown in Fig. 5.47a, b, respectively.
The model combines the following components:
The standard NMOS represented by a non-snapback N-MOSFET circuit model,

calibrated for a particular technology node.
SUB SRC GATE DRAIN
SWS L L SB Cl
p+ n+ n+
Pwell
a)
G
gate
NMOS
IS ICH ID
i1_out i1_in
S RE IC RC D
plus
E C
i2_out i2_in
BJT Iavc
Ibase
ISUB VRSUB
Bi
SUB minus b)
Fig. 5.47 Basic 2D cross section and ESD layout parameters of an NMOS device (a) and the
equivalent circuit for the snapback model (b)
The parasitic NPN BJT represented by a basic NPN transistor model, which
accounts for bipolar operation under ESD conditions with extracted parame-
ters for current gain, forward ideality factor, reverse ideality factor, equiv-
alent high-current collector resistor, and equivalent high-current emitter
resistor.
VRSUB is the current-controlled voltage source that represents the substrate

resistance modulation. It is modeled as:
ISUB (L + Ls + SWS)alfa_Rsub
VRSUB = , (4.4)
a_Rsub × W − b_Rsub × ISUB
where L, Ls , SWS, and W are geometrical parameters of the structure indicated in

Fig. 5.47a, and a_Rsub, b_Rsub, alfa_Rsub are extracted model parameters from the
experimental data [111].
The component Iavc represents the avalanche breakdown current source. It is

described using the following model:
Iavc = (Mb − 1)Ic + (Mch − 1)Ich , (4.5)
where the current components Ic and Ich are as shown in Fig. 5.47a, and Mb and Mch
are the avalanche multiplication factors due to the bulk and channel regions with a
high electric field in the vicinity of the drain–substrate junction depletion area.
The structure of the snapback PMOS model is essentially the same as that of
the NMOS model above, with the changes from NMOS to PMOS and NPN BJT to
PNP BJT.
5.4.2 Snapback LVTSCR Model

By using appropriate subcircuit components to provide positive feedback on the
device level, the LVTSCR model can also be created. Circuit components in this
model include additional devices (Fig. 5.48) to those already in the snapback NMOS
compact model.
G
C ID A
ICNPN
Re Rc
Iavc NPN Rbpnp

ISUB
B VRsub
Rcpnp ICPNP Repnp
Fig. 5.48 Equivalent circuit for an LVTSCR compact model
To reproduce the silicon-controlled rectifier, the major additional component

needed is the PNP bipolar transistor. Together with the parasitic NPN already
present in the circuit, this circuit creates a typical Ebers equivalent of an SCR.
Additional components include the resistors Rbpnp to control the base resistance,
Repnp to control the emitter resistance, and Rcpnp to control the collector resistance
of the effective PNP device.
5.4.3 Extended Voltage Snapback Compact Models
The most advanced models that have been developed support extended voltage and
high-voltage devices. Simplified cross-sections are presented below for extended
WELL SOURCE POLYGATE DRAIN WELL SOURCE POLYGATE P-EMITTER DRAIN
p+ n+ n+ p+ n+ p+ n+
Pwell Nwell Pwell Nwell
BLOCKING JUNCTON P-SUBSTRATE BLOCKING JUNCTON

P-SUBSTRATE
a) b)
0
10 100 GATE-SOURCE BIAS (V)
TLP CURRENT (A)
–1
CURRENT (A)
10 5
10–1 VGS=2.5
–2
10
10–2 VGS=0
10–3
10–3 10–4
0 5 10 15 20 25 30 0 5 10 15 20 25 30
TLP VOLTAGE (V) VOLTAGE (V)
c) d)
S D D
S
G G Oxide
Oxide N+
N+ N+ N+
Nwell
Nwell
Pwell
Pwell
In breakdown In snapback
e) f)
Fig. 5.49 Simplified cross section of 12 V NLDNMOS (a) and 12 V/20 V NLDMOS-SCR (b)
devices and corresponding measured TLP characteristics (τ P =100 ns, τ R =10 ns) at different gate
bias values for W = 200 μm (c, d). Simulation of the Kirk effect inside the NLDMOS device as a
change of impact ionization region localization for the condition of avalanche breakdown (e) and
snapback (f)
voltage NLDMOS (Fig. 5.49a) and NLDMOS-SCR (Fig. 5.49b) devices, with
corresponding TLP characteristics shown in Fig. 5.49c, d, respectively [112].
Equivalent circuits of the snapback compact models for the NLDMOS and
NLDMOS-SCR are presented in Fig. 5.50.
For the NDeMOS device at high drain–source bias, the breakdown is initiated in
the depletion zone of the Nwell–Pwell junction, as illustrated in the TCAD simu-
lation results shown in Fig. 5.49e, f. Increasing the drain bias causes a significant
voltage drop across the low-doped Nwell drain region, since it is depleted of car-
riers. This is seen in the I–V characteristics after junction breakdown. When the
G
S IS ID D
IC
VNw
c
VRsub BJT
ISUB Iavc Nw Iavc a)
c
B
G
C ID A
ICNPN
NPN
Re VRNw Iavc Rc
Rbpnp
ISUB Iavc Nw
B VRsub
b)
Rcpnp ICPNP Repnp
Fig. 5.50 Equivalent circuit of the snapback model circuits for NLDMOS (a) and NLDMOS-
SCR (b)
high-field depletion region expands with increasing drain bias, it touches the highly
doped N+ drain region. The conductivity of the Nwell region becomes completely
over-modulated by the generated electrons and holes. This results in a shift of the
maximum electric field (and impact ionization generation) toward the N+ /Nwell
interface, as shown in Fig. 5.49f. Often, this effect is referred to as the Kirk or
base-push out effect [14]. In this regime, the parasitic bipolar device is activated,
triggering the structure into snapback.
This behavior of the device has been used to develop circuit models for NDeMOS
and NDeMOS-SCR structures operating in ESD conditions [112].
Initially (Fig. 5.49a), at low drain bias, the equivalent bipolar device and
BJT , representing the impact ionization at the N+ /Nwell
avalanche current source Iavc
junction, are not active. The IavcNw avalanche current source describes the break-
down of the Nwell–Pwell junction and is expressed as Iavc Nw = M I . M is the

ch S ch
multiplication factor for the channel current and is given by the standard model
1
Mch = Nw n − 1. (4.3)
1 − VD VBR
VNw represents the voltage drop across the carrier-modulated Nwell region. It
accounts for the modulation of Nwell resistance with the increase of injected carriers
and is modeled as
I NW lW
V NW = avc
. (4.4)
I NW
Aeff qμn Nd + Aeffawc
qvsat
In (4.4), lw and Aeff represent the effective length and cross section of current
flow in the Nwell region and are treated as fitting (extracted) parameters; q, μn , and
vsat are the electron charge, mobility, and saturation velocity, respectively, while Nd
is doping. A similar equation is used for modeling the voltage source VRsub that
represents the increase of substrate potential in junction breakdown conditions.
The avalanche current source Iavc BJT represents the shift of the avalanche region
+
from the Nwell–Pwell to the N /Nwell region, which leads to the activation of
BJT is described as I BJT = M
the parasitic bipolar structure. Iavc avc BJT (kIS +IC ). MBJT is
described using a similar equation, as MCH (with different breakdown parameters)
and k are parameters used to control the gate-coupling effect in snapback operation.
Note that when Iavc BJT is activated, I Nw consistently self-deactivates due to the bias
avc
Nw
V . This voltage drop reduces the effective bias across the Nwell–Pwell junction
and, correspondingly, the avalanche multiplication that generates Iavc Nw . Such model
behavior is equivalent to the above-mentioned shift of the avalanche region inside

the Nwell.
A properly calibrated compact model can provide a rather good match between
the TCAD I–V characteristics calibrated to the experimental data and the circuit sim-
ulation results obtained using the model [112]. The model also accurately represents
the breakdown and triggering behavior for different structure dimensions.
Figure 5.51a, b shows the HBM response of the reference TCAD structure (a)
and the circuit model (b) for different HBM precharge levels. Initially, the drain
voltage increases with the transient increase of the amplitude of the incident HBM
pulse. The peak voltage indicates the snapback triggering condition, which causes
collapse of the voltage. When the discharge current becomes sufficiently low, the
device switches off and the voltage starts to increase again, due to the charging of
the drain capacitance by the small remaining HBM discharge current. It is seen that
at low HBM precharge voltage, the structure tends to oscillate due to the very low
discharge current that is not sufficient to support the bipolar snapback operation.
This is captured equally well by the circuit model.
Similar to LVTSCR, described above, the NDeMOS-SCR model is constructed
by attaching a PNP bipolar device to the NDeMOS snapback model, as shown in
Fig. 5.50b. The equivalent emitter and base resistance Repnp and Rbpnp are used
to control the activation and the high-current operation of the equivalent PNP
transistor. Rcpnp represents the Pwell resistance.
Simulated snapback triggering behavior of the NDeMOS-SCR device at different
gate bias conditions is presented in Fig. 5.51c and correlates with the experimental
results (Fig. 5.49d). The compact model reproduces the failing dependence of the
triggering voltage upon the gate bias very well.
To illustrate the usefulness of the compact models, an example of a circuit sim-
ulation reproducing the snapback of the internal circuit is presented in Fig. 5.52.
The original circuit failure analysis has indicated failed internal driver components.
VHBM = 8kV VHBM = 8kV
VHBM = 4kV VHBM = 4kV
VHBM = 500V VHBM = 500V
VD [V]
VD [V]
a) b)
10
VG = 5V
1m VG = 2V
VG = 1V
IC [A]
100n
VG = 0V
10p
0 10 20 30
VC [V]
c)
Fig. 5.51 HBM response of the reference-calibrated NDeMOS TCAD structure (a) and the
compact model (b) for different HBM precharge voltages. DC snapback simulation using the
NDeMOS-SCR model at different gate bias conditions (c)
The simulation schematic (Fig. 5.52a) is composed of the low voltage 5 V snapback
NMOS M1 , which represents the gate load of the large 20 V NLDMOS power array
protected by NLDMOS_SCR snapback ESD clamp. The ESD pulse for the analysis
is produced by the HBM voltage source.
The circuit can be used as a representation of the switching output stage of the
DC–DC converter.
According to simulation results for the waveforms of the ESD signal trans-
mitted through the parasitic gate-drain or gate-source capacitances, the internal
capacitor provides supply voltage of the driver circuit at the snapback voltage
level (Fig. 5.52b). Taking into account the possible non-uniform current distribu-
tion across the array, snapback of 5 V NMOS representing the driver circuit can be
expected.
r
a)
b)
Fig. 5.52 Simulation circuit (a) and waveforms (b) for the voltage and current at the highlighted
circuit nodes
5.4.4 High-Voltage Open Drain Circuit Analysis
Another example (Fig. 5.53) is presented by a simplified schematic of an open-

drain driver circuit, which combines NDeMOS and NDeMOS-SCR structures. The
NDeMOS device has an attached 30 k gate resistor to mimic the dynamic coupling
Output
nDEMOS nDEMOS-SCR
Output driver Protection
RLOAD RGATE
30kΩ
a)
Output
Output
nDEMOS-SCR nDEMOS
nDEMOS nDEMOS-SCR
b) c)
Fig. 5.53 Simplified schematic of ESD protection for open-drain output driver (a) voltage and
current waveforms for two values of RGATE – 10 k (b) and 1 k (c) representing two possible
circuit operation modes
load from the driver circuit. In ESD conditions, depending on the circuit and struc-
ture parameters, the ESD current can discharge either through the NDeMOS-SCR
ESD protection clamp or through the NDeMOS device.
As can be seen from the comparative analysis of the waveforms, two different
scenarios are observed. When R0 = 10 k (Fig. 5.53b), under given circuit param-
eters for a 2 kV HBM, the gate coupling on the NLDMOS-SCR gate is sufficient
to provide the early turn-on. As a result, the SCR clamp takes over the ESD cur-
rent. On the contrary, in the case of insufficient gate coupling where R0 = 1 k
(Fig. 5.53c), the clamp does not turn on and the current path is formed through the
5 mm NLDMOS device.
Thus, depending on the circuit parameters and the HBM pulse amplitude, the
critical regime can be determined as a condition of where the current changes direc-
tion in the turning-on circuit from the SCR clamp path to the NLDMOS path.
If the current through NMOS is uniform, the high-current operation will be non-
destructive. However, in a real 3D situation, the local snapback turn-on might result
in irreversible failure [113].
A similar scenario of competition between the snapback clamp and array turn-on
can be a result of the substrate potential effect. The substrate coupling technique is
also one of the most useful methods in controlling ESD devices [78].
5.5 Summary
The ESD protection network essentially represents a pulsed power circuit that pro-
vides different current paths for each pin-to-pin combination. In spite of large variety
of analog circuits, the network can be constructed based on common and rather sim-
ple principles. These include appropriate selection of ESD clamps, metallization
routing adequate for high ESD current, and accounting for the alternative current
path(s) throughout the internal circuit components.
The last is a very important design factor that cannot be neglected, but is used
instead to provide appropriate performance of the circuits. In most practical cases,
multiple possible scenarios of internal current paths should be analyzed, followed
by experimental verification of the adequate operation.
The choice of the local or rail-based network in different power domains is
application-specific, as well as the choice between the active and snapback clamps.
Since an analog circuit pin can interface with many active devices connected
to the pin, both mixed-mode simulations and circuit ESD simulations with compact
modeling tools can provide significant help in understanding the coupling of internal
circuit components and current path formation. The same approach can be used to
validate or estimate the effectiveness of the final measures.
The ability to make the right choice greatly depends upon the understanding of
the ESD protection network design and the components used in it, while simply
reusing existing solutions does not always prove adequate.

Example 5.1 Active 5 V NMOS Clamp

Library Name: Examples5A_ESD_Active_Clamps
Project Name: E5.1a_5 V_AClamp_HNMOS
Project Name: E5.1b_5 V_AClamp_PowerUp
Project Name: E5.1c_MM_AClamp_HNMOS
Project Name: E5.1d_5 V_AClamp_Consolidated
R1*
C2
1
2.6e-12
V0*$ Vpulse...
R3 Precharge L0 R1 Discharge R4
HBM_Pulse
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0$ Vpulse... 1e-10
M2*
R2
ESDPLUS
100000 C3*
M2* 500f
C3*
R2 500f M3*
100000
W=1
M3* M1*
W = 400
M1*
C4
60e-12 W=1
C4
60e-12 W = 10000
W=1
W = 200
ESDMINUS
Fig. E5.1 Mixed-mode simulation circuits of the active NMOS clamp for HBM ESD, MM ESD,
and power supply voltage ramp simulation and simulated circuit responses for 2 kV HBM, 200 V
MM ESD pulses
This set of examples is prepared for exploration of the most popular active clamp
solution based upon the NMOS device. The first example demonstrates waveforms
produced by the clamp under an HBM ESD pulse and can be used to study the
effect and optimization of the RC timer and driver parameters as well as the size of
the NMOS (Fig. E5.1). The second example shows the dependence of the parasitic
current of the clamp for the power supply ramp. The following example provides
the mixed-mode circuit for clamp operation under the machine model (MM) ESD
pulse. The last example has all three circuits consolidated in a single project.
Example 5.2 Active 5 V PMOS Clamp

Project Name: E5.2_5 V_AClamp_HPMOS
C2
2.6e-12
Precharge Discharge HBM_Pulse

R3 L0 R1 R4
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0$ Vpulse... 1e-10
ESDPLUS
M4* M2*
R2
100000
M3*
W = 10 W = 400
M5* M1*
C4 W = 10000
60e-12
C3*
500f
W=5 W = 200
ESDMINUS
Fig. E5.2 Mixed-mode simulation circuit for the active PMOS clamp and simulated waveforms
under HBM ESD pulse stress
This example presents the 5 V PMOS-based active clamp (Fig. E5.2). In some cases,
this type of clamp is preferable over the NMOS clamp due to the high holding volt-
age of the PMOS and corresponding elimination of the transient latch-up conditions.
Example 5.3 EEPROM Erase pin Protection

Project Name: E5.3a_EEPROM_Erase_Clamp_ver1
Project Name: E5.3a_EEPROM_Erase_Clamp_ver2
In this example, two alternative versions of the high-voltage erase pin protec-
tion design are compared. Both clamps rely on the high-voltage NLDMOS-SCR
snapback device (Fig. E5.3). High residual voltage is built up in the first clamp, pro-
viding undesired overstress of the pin that may erase information in the memory.
The second version of the clamp utilizes additional NLDMOS-based RC-controlled
circuitry to eliminate the high voltage peak after SCR turn-off.
Fig. E5.3 Mixed-mode simulation circuits for the two versions of the erase pin protection clamp
design with the simulation results for HBM ESD pulse conditions
Example 5.4 BJT-Based Active Clamps

Project Name: E5.4a_5 V AClamp_NPN
Project Name: E5.4b_5 V AClamp_NPN_casc
Project Name: E5.4b_5 V AClamp_PNP
This set of examples presents active clamps with the BJT device as an active compo-
nent. BJT-based active clamps provide a significant advantage in clamp size. Three
alternative versions of the BJT-based active clamp designs cover NPN, cascaded
C2
C2
2.6e-12
2.6e-12
R3 L0 R1 R4
R3 L0 R1 R4
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K C1 1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C0 2e-11 C1
V0$ Vpulse... 1e-10 C0 2e-11
V0$ Vpulse... 1e-10
ESDPLUS ESDPLUS
M2 M1*
C3*
R2 C3* R2 500f
100000 500f 100000
Q1*
M3
TRIG TRIG
W = 40 W = 40
Q2*
M1 M0*
Q0*
W = 10
C4 W = 1000 C4
60e-12 60e-12
R5* R5*
10000 10000
W = 10 W = 10 W = 1000
W = 1000
ESDMINUS ESDMINUS
C2
2.6e-12
R3 L0 R1 R4
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0$Vpulse... 1e-10
ESDPLUS
M2*
R5*
R2 C3* 10000 Q0*
100000 500f
M3*
TRIG W=40
M1* W=1000
C4 W=1000
60e-12
W=20
ESDMINUS
Fig. E5.4 Simulation circuits for three versions of the BJT-based active clamps and the results of
the 2 kV HBM ESD pulse mixed-mode analysis for different clamp parameters
NPN and PNP versions. The simulated waveforms for these clamps can be com-
pared with the CMOS versions (Fig. E5.4). These examples can be used as a starting
point for design and optimization of the CMOS driver and BJT clamp device.
Example 5.5 Stacked Active Clamps for High Voltage Tolerance

Library Name: Examples5B_Stacked_Active_Clamps
Project Name: E5.5_Stacked_AClamp_CMOS_Consolidated
This set of examples presents an active clamp solution that achieves a voltage toler-
ance two times higher than that of non-stacked active clamps. Two examples present
both ESD operation of the clamp and power supply ramp transient analysis. The
examples are built using compact models for a 1.5 V CMOS process with only
1.5 V poly-capacitors available (Fig. E5.5). The circuit is tolerant of 3 V interface
voltage. A simulation with a power supply voltage source demonstrates the current
level consumed by the clamp at rise times of 10 and 100 μs.
C9
2.6e-12 Power_S...
HBM_pulse R5
R2 L0 R1 R3
1
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K
C8
C7 2e-11 V1*
V0$ Vpulse... 1e-10 $Vdd*pu.
ESDP
VTRIG1
M3 M9
R0 R4
400K M5 400K M7
C1 C10
C3 W=40.0 900f C12 W = 40.0 900f
100f 100f
C5 W = 2000.0 C14 W = 2000.0
20p M2 20p M10
M4 M6
VTRIG2 C2 C11
900f 900f
W=40.0 W = 40.0
M1 W = 2000.0 M8 W = 2000.0
C6 C15
20p 20p
C4 W=20.0 ESDM C13 W = 20.0

100f 100f
Fig. E5.5 Simulation results for the stacked CMOS active clamp with 2 kV HBM ESD pulse and
power supply ramp sources
Example 5.6 Stacked Active Clamps with NPN

Project Name: E5.6a_Stacked_AClamp_NPN_HBM
Project Name: E5.6b_Stacked_AClamp_NPN_PowerUp
This and the following examples demonstrate the non-snapback ESD solution
implementation toward a smaller footprint by taking advantage of BJT devices. The
circuit represents the clamp assembled using a stacked driver and built with compact
models for 1.5 V CMOS components combined with a FEM NPN BJT device. The
circuit is tolerant of a 3 V voltage. Using the simulation, optimal driver parameters
can be found for matching the driver circuit with the FEM BJT (Fig. E5.6).
C9
2.6e-12
HBM_pulse Power_S...
R2 L0 R1 R3 R5
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K 1

C8
C7 2e-11
V0$ Vpulse... 1e-10 V1$ Vdd*pu...
ESDP
ESD
VTRIG1
M3 VTRIG1
M9
R0
400K M5 R4
400K M7
C1
W = 20 500f C10
C3 W = 20 500f
100f C12
C5 W = 500.0 100f
20p M2 C14 W = 500.0
20p M10
M4
M6
VTRIG2 C2
500f C11
W = 10 Q1* VTRIG2
500f
W = 20.0 Q2*
M1 W = 500.0
M8 W = 500.0
C6
20p W = 1000 C15
R6 20p W = 1000
10K R10
W=5 ESDM
C4 10K
100f W=5 ESD
C13
100f
Fig. E5.6 Mixed-mode simulation circuits for the NPN-based active clamp with a stacked CMOS
active clamp driver and examples of 2 kV HBM ESD pulse and power supply ramp waveforms
observed in each circuit
Example 5.7 Stacked Active Clamps with PNP

Project Name: E5.7a_Stacked_AClamp_PNP_HBM
Project Name: E5.7b_Stacked_AClamp_PNP_PowerUp
This example demonstrates ESD solution implementation using PNP BJT devices.
This approach is useful in a number of BiCMOS process technologies with low-
voltage CMOS. A mixed-mode simulation circuit represents the clamp, which is
assembled using a stacked driver built using compact models for 1.5 V CMOS com-
ponents combined with a FEM NPN BJT device. The circuit is tolerant of 3 V
voltage. Using this simulation, driver parameters optimal for matching the driver
circuit with the FEM BJT can be found (Fig. E5.7).
C9 Power_S...
R5
2.6e-12
HBM_pulse 1
R2 L0 R1 R3
V1* $Vdd*pu...
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K C8
C7 2e-11 ESDP
V0 $Vpulse... 1e-10
ESDP
VTRIG1
M8
VTRIG1 R7 Q1*
M3 R4 10K
R6 Q0* 400K
R0 10K
400K
M6
M5
W = 20
W = 20 C12 C10 W = 2000
C3 C1 W = 2000 100f 500f
100f 500f C14
C5 20p M9
20p M2 W = 500.0
W = 500.0
M4 M10
VTRIG2 VTRIG2
C2 C11
W = 10 500f W = 20.0 500f
M1 M7
W = 500.0
C6 W = 500.0
20p C15
20p
C4 W = 5 ESDM
100f W=5 ESDM
C13
100f
Fig. E5.7 Mixed-mode simulation circuits for PNP-based active clamp with stacked CMOS active
clamp driver and examples of 2 kV HBM ESD pulse and power supply ramp waveforms observed
in each circuit
Chapter 6
ESD Design for Signal Path Analog
The purpose of this and the following chapters is to demonstrate the implementa-
tion of different ESD protection approaches specific to analog products. Two major
categories of analog products are used as examples demonstrating ESD protection
challenges and solutions: the signal path and the power management products. The
first category is addressed in this chapter. Chapter 7 is focused on the specifics of
power management products. These products include low-voltage and high-voltage
integrated DC–DC converters and controllers, LED and display drivers, and other
power products.
In this chapter, ESD protection implementation is presented for signal path prod-
ucts, represented by high speed, precision and audio amplifiers, digital-to-analog
converters, and interface circuit blocks.
The subdivision of analog products into signal path and power management ana-
log circuits emphasizes a corresponding difference (based on pin specifications) in
the implemented ESD protection for the two cases. While ESD protection solutions
for control, digital, and low-power analog input pins can often be based on common
principles, a number of representative, product-specific pins may require significant
changes in both the clamp selection and ESD network design.
In the case of low-voltage signal path circuits, the major challenges are in the
minimization of the parasitic influence of ESD protection components on the signal.
The majority of digital–analog converters and interface circuits are based on low-
voltage CMOS devices with an operating voltage that corresponds to the supported
DGO CMOS processes. Thus, I/O pin protection is designed for a voltage range
below 1.2–5 V. The most typical exceptions are, perhaps, the external EEPROM
program pin (∼6 V) and the erase pins (8–18 V). Thus, corresponding solutions
mainly focus on overcoming signal integrity challenges.
Unlike in interface and digital-to-analog converters, the voltage range of ampli-
fiers is rather broad and can cover a 3–100 V operational voltage range. At the
same time, even high-voltage ESD protection of amplifiers is significantly differ-
ent from that of power management circuits (Chapter 7). Usually, the protection
of high-voltage amplifiers can be implemented using a conventional rail-based
approach with a voltage power clamp. This is possible because the requirements
of high quality and linearity of the signal path output device parameters almost

282 6 ESD Design for Signal Path Analog
automatically provide relatively large SOA margins for high-voltage output devices.
In other words, a relatively wide ESD protection window is usually expected.
In opposite, in the case of power management circuits, the major challenges are
related to the protection of the high-speed and high-voltage transient pins connected
to the internal power devices. In the case of power devices, SOA limits can be very
close to the absolute maximum voltage of power analog products. This creates a
very narrow and sometimes even negative ESD protection window which demands
rather aggressive ESD protection solutions.
Another specific of signal path analog products is related to high pin count pack-
ages. This specific is due to the fact that the charge device model (CDM) ESD pulse
provides rather high ESD current conditions for the large form factor of high pin
count packages, unlike in small pin count power circuits. This specific is similar to
the one of high pin count digital products.
The overall ultimate goal of the remaining chapters is to demonstrate how the
accumulated knowledge of Chapters 2, 3, 4, and 5, based on the physical understand-
ing of conductivity modulation in semiconductor structures, ESD device and clamps
design principles, and pulsed SOA and ESD protection network design principles,
can be applied in practical analog design.
This goal is attained by means of a high-level review of the most common fea-
tures of the representative analog products and ESD protection examples selected
for each chapter. Each section contains a discussion of the selected, most typical
ESD protection network for the given product class, and the related problems and
solutions for the specific product pins based on corresponding case studies.
6.1 Amplifiers
6.1.1 Amplifier Product Families and Specifications

For ESD protection implementation, it is convenient to subdivide amplifiers into
three major groups:
(i) High-speed amplifiers

(ii) Precision amplifiers
(iii) Audio amplifiers
The first group of amplifiers deals with the amplification of high-speed, low-
power signals of up to 10 GHz, with corresponding figures of merit such as signal
quality, noise, and the amplification factor.
The second group usually deals with a much slower signal that requires much
higher accuracy and matching and stability of characteristics.
Finally, audio amplifiers handle amplification of the signal in the audio frequency
range.
6.1 Amplifiers 283
The application specifics imply certain peculiarities of the ESD protection

design. High-speed amplifiers are mainly designed using corresponding high-speed,
low-voltage processes. In this case, additional parasitic capacitance from the input
ESD clamp may significantly impact the product performance.
Precision amplifiers, on the other hand, may tolerate regular ESD protection
components but may be rather sensitive to the matching of the differential input
component parameters. The same is true for power audio amplifiers, which can
be implemented in a rather high-voltage process to achieve the appropriate output
power.
Due to requirements of precision and linearity of the high-voltage output circuit
components, the operation regimes for power amplifiers are located far from the
physical pulsed SOA limits. Thus, high-voltage amplifiers are designed using com-
ponents that usually provide a suitable ESD protection window for implementation
of a rail-based network with a high-voltage ESD power clamp.
High-speed amplifiers are often used as parts of systems where a significant
rate of information is transmitted by the signal. For example, high-speed ampli-
fiers are designed to support the following applications: broadcast of video; clock
generators and sync separators; security in a closed-circuit television cross-point;
medical imaging in ultrasound analog front-end (AFE), base station receiver AFE’s;
digital-variable-gain amplifier (DVGA); mixers; low-noise amplifiers (LNA); driver
receivers; pro-video cross-points; low-noise analog clock buffers; comparators;
short haul optical links; power amplifier loop control; base station transmitters for
RF-DVGA; and many others.
Modern trends in audio amplifier design include the increase of the power supply
voltage from +/– 5 V to well over +/– 20 V. This high voltage is required for an
improved performance, for example, in the signal-to-noise ratio (SNR). Another
trend is the migration from Class AB to Class D.
The important factors in amplifier design are system performance and simpli-
fication of system design including reduction of the number of required external
components.
To address these needs, new integrated semiconductor process technologies tar-
get several challenging process development areas. Table 6.1 demonstrates the large
variety of BiCMOS semiconductor process technologies involved in supporting
audio amplifier products. High-voltage processes are designed to provide a high
gain at high voltage. Wide bandwidth is targeted by the development of a fully com-
plementary process technology for a high fidelity audio performance with transistor
matching and minimized thermal effects (in order to reduce or eliminate the required
heat sink).
Another focus is the reduction of customer design time; the driver stage is usually
the most time intensive for design. To achieve this high output power, multiple bipo-
lar transistors can be paralleled, targeting up to ∼300 W output power per channel
with NPN/PNP output pairs.
Typical packaging options include a high variety of small pin count package form
factors (Fig. 6.1).
Table 6.1 The large variety of BiCMOS semiconductor process technologies in the audio power
amplifier products
Product family BiCMOS process features
Class AB power amps 2 × 20/40 W Thin EPI 50 V

Class AB power amps 1 × 80 W Thin EPI 120 V
Class AB power amps 2 × 50/100 W 100 V DMOS
Class D power amps 6 × 50/250 W; Class D 1.5 μm DMOS
output stages 80–250 W H Bridges; Class
D Controllers; eight channels + DSP
±15 V analog uPot buffered; 128 step, 0.5 dB 36 V CMOS/BIP/Prec R
±5 V analog uPot buffered; 128 step, 0.5 dB; 12 V Bip/ Prec R
+5 V analog potentiometers – CMOS7;
128 step, 0.5 dB
Low noise preamplifiers; digitally controlled 36 V BJT/Precision R; 12 V
preamps BJT/ Prec R
Class AB power amps audio op amps; Thin EPI
low-noise audio op amps <2.5 nV/rt Hz
Micro SMD 0.5mm Pitch Micro SMD 0.4mm Pitch

LLP
New Release
PSOP
Possible Slug-Up
Option
11 & 15ld TO220 27ld TO220 15 & 19ld TO247
Fig. 6.1 Typical packaging options for audio amplifiers
A simplified block diagram for an audio power amplifier is presented in Fig. 6.2.
The applications include high fidelity power amplifiers; high fidelity multime-
dia; high-performance professional audio; high fidelity equalization and crossover
networks; high-performance line drivers and receivers; high fidelity active filters;
very high-voltage operation; scalable output power; and feature minimum external
components; external compensation; thermal shutdown; and mute.
One of the growing process-specific applications is the complementary bipo-
lar (CB) design [114]. For broadband and narrowband, for example, wireless, the
NPN/PNP symmetry is important and special attention should be paid to the thermal
self-heating issues, especially in SOI process technologies (Table 6.2).
An example of the fully complementary BJT amplifier design is presented in
Fig. 6.3 [114].
Examples of the specific requirements for a communication variable gain ampli-
fier (Fig. 6.4) may include a wide band above 150 MHz, which is required from the
6.1 Amplifiers 285
Fig. 6.2 Simplified block diagram for 2-channel audio power amplifier
process components; high fτ and fmax parameters; a wide dynamic range above 80
dB provided by the logarithmic characteristics and gain matching; low voltage and
current noise En and In ; a high output voltage swing and drive provided by device’s
DC symmetry; linearity achieved by the AC symmetry.
In the case of a high-resolution video application (Fig. 6.5a), the requirements
may include exceptional linearity at ∼4 MHz; level of 2nd/3rd harmonics below
100 dB; 700 MHz bandwidth and 3000 V/μs slew rate. This requires high fτ and
fmax ; exceptional gain and phase stability over the input common mode (CM) and
output voltage range; 0.02% differential gain error; 0.005% differential phase error
which requires AC/DC symmetry from process components; low noise of 158 dBm
(1 Hz).
Fig. 6.3 Complementary BJT amplifier design [114]
VIN
a) b)
Fig. 6.4 Variable gain amp (VGA) (a) and dependence of the gain on gate bias voltage VG for
various temperature levels (b) [114]
Fig. 6.5 Composite wide band amplifier and output signal at 4.43 MHz
6.1 Amplifiers 287
Table 6.2 Circuit need vs. device requirements for amplifier products
High-speed amplification High fτ /fmax for NPN/PNP
Wide bandwidth at low Minimize fτ /fmax rolloff; minimize

Isupply capacitive parasitics; oxide isolation and
thin film or Poly2 resistors. Availability
of MIM or Poly2-metal caps
High linearity (low Complementary NPN/PNP BJT for fτ /fmax ,
distortion) Cbc , Cbe , Ccs , re , rb , β, VA
Output drive (sink/source) Complementary NPN/PNP high current β
Low noise Low NPN/PNP base resistance,
minimize 1/f
Fig. 6.6 An example of an integrated switching regulator similar to those built into power
amplifier IC based upon BCD process
For portable applications, the amplifier circuit can be rather complex. To take
advantage of NLDMOS devices in a high-voltage BCD process, the amplifier
product may include an integrated switching voltage regulator (Fig. 6.6).
Another level of complexity can arise due to the integrated power blocks for
example EEPROM charge pumps. An example of the functional block diagram for
the high-speed amplifier is presented in Fig. 6.7.
The parameters of the amplifier imply a corresponding limitation on the parasitics
introduced by ESD clamps and the ESD protection network. The problem cannot
always be solved in a simple way. For example, the LNA for RF applications usually
have a reduced ESD protection level for input pins. Due to the absence of the ESD
Fig. 6.7 Example of the simplified phase lock loop high-speed amplifier block diagram: phase
lock loop, 5 V charge pump, programmable output buffer type, and new counter architectures with
synchronization, programmable delay, and duty cycle
protection clamps, these devices rely on the internal circuit to provide at least a
minimal level of ESD protection.
6.1.2 ESD Solutions for Amplifiers
In spite of a high variety of amplifier products with the complex integrated ana-
log blocks described above (Figs. 6.6 and 6.7), in most cases, the ESD protection
network can be successfully implemented based upon either a rail-based protection
network (Fig. 6.8a) or a local clamp network (Fig. 6.8b). Such networks create an
ESD current path between the amplifier pins. These pins can usually be classified as
follows.
(i) CMOS or bipolar inputs, completed using a buried channel NMOS, JFET,
super beta BJT devices. As an option, a differential input can be implemented.
(ii) The bipolar or CMOS outputs
(iii) A high-voltage or low-voltage power supply pins in cases of single or dual
power sources. In the case of separate voltage domains, the power supply
may combine both low-voltage domains for pre-amplification and high-voltage
domains for power output.
6.1 Amplifiers 289
VCC bus
HV ESD D
~
VIN- ~
-
~ VOUT
~ Amp
~
VIN+
~ +
~
~ ~
~
VEE bus
a)
Vcc
-
amp1 +
+ amp2 VOUT ~
-
Disable
Disable
circuit ~
~ ~ sense
GND
b)
Fig. 6.8 Block diagram for the rail-based (a) and local clamp (b) ESD protection for amplifier
circuits
(iv) The additional group of pins includes low-voltage control, service, and digital
interface pins.
Apparently, the amplified signal quality, as well as the performance of the whole
amplifier, is the most impacted by the ESD network components connected to the
output and, especially, to the input pins.
The ESD protection network for an amplifier with a single voltage domain can
be realized based on a rail-based or local ESD protection network according to the
major principles presented in Chapter 5. Usually, to achieve an appropriate perfor-

mance, specifically the rail-based ESD protection network is used, with I/O diodes
to ESDPLUS and ESDMINUS rails. The power clamp is used to provide a current
path between the ESD busses and the I/O pins.
In most practical cases, the ESDPLUS bus is not separate, but shares a rail with
the power supply bus of VCC, VDD, or similar. The ESDMINUS bus is represented
by a corresponding VEE, GND, or similar bus. In the case of dual voltage supply,
VCC+ and VCC– can provide functions for ESD busses with substrate-isolated ESD
power clamps referenced to the lower power supply node.
Typically, the snapback clamp has much bigger footprint in comparison with the
ESD diodes. Therefore, a more space-saving solution is based on the rail approach
(Fig. 6.8a). An additional advantage of signal integrity is provided by the lower
parasitic capacitance and leakage of the diodes.
It is important to remember that although the diodes do not conduct any ESD
current in the reverse-bias avalanche breakdown mode, they should still be tolerant
to the corresponding high-voltage level of the given amplifier power supply domain.
The same is critical for other components, for example, diffusion resistors.
Because of design constrains, it is not always preferable to provide wide metal
busses. For example, in processes with fewer supported metal layers, the design of
a low-resistive current path can be rather complicated and may be impacted by the
signal to bus coupling limitations. Therefore, high-voltage bipolar output protection
may become marginal even if the amplifier has high linearity components.
For low-cost analog process technologies with fewer layers of metal, the creation
of low-impedance ESD rails across the whole chip may be sacrificed in favor of a
local protection approach (Fig. 6.8b).
The product specifics of BJT or CMOS input and output designs should be con-
sidered separately and dictate the choice of the ESD protection strategy applied to
the pin.
In case of hot plug-in, system-level or automotive applications additional system-
level design requirements should be met for certain pins.
6.1.3 Bipolar Output High-Voltage Audio Amplifiers

Usually, it is a relatively easy task to implement a high-voltage power clamp for
protection of the power supply pin only. In relatively slow high-voltage amplifiers,
the design parameters for the output signal usually automatically provide a sufficient
ESD protection window.
In the case of a high-voltage audio amplifier with multiple power domains
(Fig. 6.9), for example, an audio amplifier with digital volume control (Fig. 6.10),
diode performance becomes critical. In this case, the voltage drop on the diode
should be accurately accounted for in each ESD current path scenario to ensure
the absence of overstress at the I/O pins above the absolute maximum limits. For
6.1 Amplifiers 291
HV ESD D VCC bus

~
VIN TFR -
~ VOUT ~
~ Amp CSP ~
MUX
~ +
~
~
~ ~
~
~ TFR
Power
ESD
DGND
Clamp
LV ESD D
AGND
VEE bus
Fig. 6.9 Block diagram for two-domain ESD protection implementation in an audio operational
amplifier with volume control; differential HV input, HV digital I/Os and high-voltage NPN and
PNP outputs
VCC1 bus VCC2 bus
rgnd rvo rvi ss sda scl pwr5v

~ ~ pwrn
Power
ESD
Clamp
~~
VEE1 bus VEE2 bus
Fig. 6.10 Alternative ESD current paths in a two power domain amplifier network realized
between different domain I/O pins
example, the total voltage for an ESD zap combination between the low and high-
voltage inputs will accumulate the voltage drop on three diodes, in addition to the
voltage drop on the busses and power clamps.
Thus, to provide a discharge current path for different input and output zap com-
binations, the diodes should be optimized both for low on-state resistance and low
parasitic capacitance and leakage.
The total voltage drop at the output depends on the pin-to-pin combination
(Fig. 6.10). Even for two voltage domains, the ESD current path may include cur-
rent path scenarios with the clamp only, the clamp plus one diode voltage drop, +2
diodes, or +3 diodes.
If the ESD protection window is insufficient and separation resistors cannot be
placed at the pins, then an additional local clamp, for example, one similar to the
power clamp, can be applied at the critical pins.
In the case of high-voltage circuits, it is advantageous to use low-voltage ESD

diodes in the back-to-back diode clamp between the domain grounds and at the
differential inputs. The last requires an appropriate isolation of the diodes from
the substrate in case of non-SOI processes. An additional benefit of this separate
diode selection is the smaller footprint of the low-voltage back-to-back diode clamp
(Fig. 6.10).
6.1.4 Bipolar Output Protection in Low-Voltage Amplifiers

An example of a low-voltage, high-speed amplifier with a BJT output and pro-
tection components is presented in Fig. 6.11a. This ESD protection solution has
been implemented for the high-performance, current-feedback differential amplifier
[105].
VCC
0.25u/10u
m:10
VOUT 5V
CLAMP
0.25u/10u
m:10
VEE
a)
Leakage Current (A) Leakage Current (A)
1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02 1.E-12 1.E-10 1.E-08 1.E-06 1.E-04 1.E-02
1.8 1.8
1.6 VCC_GND VCC_VOUT
Vleak=0.7V 1.6
1.4 Vleak=1.5V
Pulsed Current
1.4
Pulsed Current
1.2 1.2
1 1
0.8 0.8
0.6 0.6
0.4 0.4
0.2 0.2
0 0
0 2 4 6 8 10 12 0 4 10 15
Pulsed Voltage (V) Pulsed Voltage (V)
b) c)
Fig. 6.11 An example of the BJT output in a high-speed amplifier (a); TLP characteristics for the
power supply domain (b) and output to the ground (c) zap combinations
6.1 Amplifiers 293
In this case, the total width of the 5 V output devices is W = 200 μm, which
is sufficient to provide self-protection. With the particular design parameters, the
selected FOX snapback power clamp has a rather high triggering voltage.
For the VCC to VEE zap combination, the clamp, with a triggering voltage of
∼11 V and a holding voltage of ∼6 V, conducts the ESD current (Fig. 6.11b).
However, in the case of the output VOUT to power supply VCC ESD stress
combination, the TLP characteristics are different. The comparative analysis shows
that in this case, the output NPN BJT transistor already provides an ESD current
path to the ground while the power clamp remains passive. This conclusion about
the current path can be easily confirmed by the observed different triggering and
holding voltage levels (12.5 and 8 V, respectively) (Fig. 6.11c).
Thus, the circuit (Fig. 6.11a) relies on the self-protection of the BJT device of
appropriate size for at least one current path.
This measure is often advantageous in practical amplifier designs. Even if NPN
operation is unstable in high-current mode, its self-protection capability can often be
improved by additional polyemitter ballasting resistors [115] that ensure the current
balance in the BJT array (Fig. 6.12). A similar approach can be used for small
devices to increase the ESD protection window. In this case, an extension of the
pulsed SOA for small output BJT devices is used to ensure the turn-on of the ESD
clamp.
VCC
0.25u/10u
m:50
RE = 24ohm
m:50 5V
CLAMP
RE = 24ohm
m:50
0.25u/10u
m:50
VEE
Fig. 6.12 Using additional polyemitter ballasting resistors in the 5 V bipolar output stage for the
ultra-low-noise voltage-feedback operational amplifier
6.1.5 Input Protection

The BJT differential input is usually protected by the back-to-back stacked diode
clamp (Fig. 6.13). To avoid pin mismatch, the clamp is composed of substrate-
isolated p+ -Nwell and non-isolated n+ -Pwell diodes. The shallow junctions of the
diodes are connected to the input pins in order to reduce the parasitic capacitance
introduced by the diodes, while the well diffusion is connected to the floating nodes
of the clamp.
ESDP (VDD) BUS
BSCR Clamp
INP+ INP–
OUT1 R
ESDM (GND) BUS
Fig. 6.13 Differential input protection
For the CMOS digital input, the most sensitive device components are the MOS
devices, due to gate oxide overstress. In principle, rail-based protection can still
be used successfully in this case, especially if the circuit’s performance level can
tolerate an additional decoupling resistor for the second stage (Chapter 5).
One of the most important aspects of successful ESD design is avoiding
alternative damaging ESD current paths through the internal circuit.
This can be illustrated by an example of antenna diode placement. Usually, an
additional minimum dimension region in the Pwell region of the NMOS device
forms the antenna diode. The placement of the diode is a design choice. If this n+ -
region is drawn too close to the source diffusion, a parasitic NPN with rather high
base resistance may be formed (Fig. 6.14b). Depending on the process technology
and the Pwell connection, if minimum process rules are applied, the triggering volt-
age of such an NPN device may become even smaller than the gate oxide breakdown
voltage, thus providing a new severe limiting factor to the ESD protection window.
To avoid such an issue, the antenna diode should be either spaced apart from the
input NMOS source or an additional p+ diffusion region should be placed closer to
the antenna diode.
Pwell
R=200 M1 p+
INP M1
PAD
n+-Source
Polygate
DA
R=50k
n+-Drain
3V
SNAPBACK n+
CLAMP
DA
GND
a) b)
Fig. 6.14 Local protection of the CMOS input (a) and an example of the layout with an incorrect
antenna diode placement (b) that enables a parasitic NPN BJT device with high base coupling
6.1 Amplifiers 295
Similarly, an ESD protection window limiting can be caused by the MIM

(metal–insulator–metal) capacitors used at the input and so should also be
considered.
In the case of the low-noise amplifier (LNA) bipolar input, it may not be possible
to implement a second-stage resistor due to design requirements. At the same time,
minimally sized BJT devices may be used to provide the required circuit perfor-
mance (Fig. 6.15). There are current paths in the circuit both from the input to the
power supply (VCC) and from the input to the ground (VEE) that can be observed
before the triggering of the local protection clamp, depending on the conditions of
M3 and M4.
Core Circuit ESD Protection Circuit

VCC
1.8k M4
W=1um
INP1 INP2
M1 M2
W=0.5um W=0.5um
INP1 INP2
W=1um
M3 Snapback Snapback
900 Clamp Clamp
VEE
Fig. 6.15 An example of the minimally sized BJT LNA input circuitry with ESD protection and a
possible additional current path scenario inside the internal circuit
A possible fix for this circuit protection issue consists of fine-tuning the trigger-
ing voltages of the ESD protection snapback clamps (Fig. 6.15), increasing the size
of the input devices, and by additional resistor to collector of the BJT M3.
6.1.6 CMOS Output

Low-voltage CMOS output protection in amplifiers is similar to the strategies dis-
cussed in Chapter 5. If the design can tolerate the active clamp approach, then
the voltage limiting of CMOS circuit are rather straightforward and similar to the
protection of digital circuits.
As an option for CMOS output pins with an open drain, the NMOS can be
redesigned with a drain ballasting region to match each drain finger, instead of the
lumped separation resistor.
6.2 Digital-to-Analog and Analog-to-Digital Converters

The analog-to-digital converter (ADC) transforms (converts) an input analog volt-
age or current into a digital output proportional to the magnitude of the original
voltage or current.
ADC’s are widely used for processing an analog signal in a digital form. For
example, ADC’s are found in XM radios, digital video cards, microcontrollers, and
digital oscilloscopes.
The serial digital data interface is used to reduce the cost of the products, by
minimizing the pin count in ADCs.
Respectively, the functionality of the digital-to-analog converter (DAC) is the
conversion of an arbitrary finite-precision number at a digital input into a cor-
responding analog output signal that represents a physical quantity, for example,
voltage or current.
In the time domain, DACs are often used to convert finite-precision time series
data into a continually varying physical signal.
A typical DAC conversion method is based on delta–sigma modulation for pro-
ducing a pulse-density modulated signal that can then be filtered to produce a
smoothly varying signal.
DAC applications cover a broad spectrum of systems: digital audio systems from
players to sound cards; digital video systems and displays; and other consumer,
industrial and medical applications.
To some extent, the basic ESD protection approach used for digital–analog
converter (DAC) and analog–digital converter (ADC) products can be developed
according to a general understanding of the corresponding functional blocks of the
converters. At an even higher level, both DAC and ADC combine features that allow
application of ESD design principles for digital blocks and the previously described
analog blocks.
In most typical cases, the digital part can be protected using the rail-based active
clamp approach with corresponding upper and lower ESD diodes from the I/O to
the rails. In this case, the most attention is focused on meeting the CDM protection
requirements and some of the ADC and DAC digital blocks.
The specific case of the serial data line pins (or I2C) has been already discussed
above and are related to open drain protection.
There is a corresponding special case of ADC protection that is related to the
low-voltage differential signal (LVDS).
However, to some extent, protection of the analog part in ADC and DAC can
be treated somewhat similarly to the protection of the amplifier input and output,
respectively. In a high-end, multibit ADC, the ESD/signal interference is the major
challenge in providing an appropriate analog input signal resolution.
The analog output of the DAC can be treated as a CMOS or BJT output, similarly
to the amplifier discussed in the previous section.
To illustrate the major principles of ESD protection for the DAC, the major func-
tional blocks and most typical external pins of the 14-bit DAC are presented in this
section.
6.2 Digital-to-Analog and Analog-to-Digital Converters 297
The selected example is addressed for a DAC designed for the 0.13 μm DGO
CMOS process with two voltage domains: the 1.2 V domain for the digital blocks
and the 3 V voltage domain for the analog blocks.
6.2.1 Functional Blocks for High-Speed DAC

A simplified block diagram of the major functional blocks of digital-to-analog
converter is presented in Fig. 6.16.
TESTIN <0:2> TESTOUT <0:6>
3 Copy of Dig 7
Encoder DAC
DIN <0:13>
2X Digital
1.6
GHz Differential Core
(800 MSPS) Digital Analog
Interpolation Switch 2
Iout
14 14 Encoder Core
Filter Drivers
(Vb1,3)
(Vb4) 3 Bias
LVDS CLK LVDS Nodes
(1.6 GHz) to Biasing (Ib1,4)
2 VREF Injection
2 CMOS & 4 Pts
Conv V-TO-I
RFS
T_Diode 1.2V 3V PMOS IINM

BG BG BG
CMOS Asym
upads #1 #2 #3 IOUTM
Curr Mirr
1.2V Block External VREF

Bandgaps
3.0V Block
Fig. 6.16 Simplified block diagram for the digital-to-analog converter
The purpose of the system is to convert digital code received as serial low-voltage
differential signal (LVDS) with 1.2 V amplitude into an output analog signal of up
to 3 V (Fig. 6.16).
The 14-bid DAC contains 11 major functional circuit blocks. From the ESD
protection point of view the most important information is about the devices and
circuits directly interfacing with external pins. Practically all the digital parts of
the converter are based on 1.2 V devices, while the analog parts are based on 3 V
devices.
The functional circuit blocks are designed with different parts of the DGO CMOS
process technology, using high-speed 1.2 V CMOS devices with thin gate oxides
(NMOS and PMOS) and 3 V-tolerant thick gate oxide CMOS devices (HNMOS
and HPMOS). The deep Nwell layer available in the process isolates the NMOS
structures from the p-substrate, thus generating the corresponding isolated devices
NMOSI and HNMOSI.
The biasing block is mainly based on the 3 V HPMOS and HNMOS. The block
sets bias currents for the DAC and the LVDS-to-CMOS converter, based on an
external bias resistor. This block requires the reference voltage VREF to function.
The other 3 V pins are represented by the bandgap reference circuits based on
HNMOS, HPMOS, lateral PNP devices, and Deep Nwell NPN devices. The PMOS
current mirror is designed based on the HVPMOS and the PMOS with 3 V devices
that interface with the external pins in the biasing circuitry block and analog core
output.
The PMOS current mirror (Fig. 6.17) shows a single cell corresponding to DAC’s
output pins.
VDDA
HPMOS
HPMOS
PMOS
PMOS
INM OUTM
Fig. 6.17 PMOS current mirror
One of the three band gaps can be selected to produce a stable reference voltage
VREF . VREF can be externally measured and used to compare the differences in the
three band gaps. The band gap block requires the mode control block to function.
The block is built on the 3 V devices (Table 6.3) with the external pins VREF, analog
VDD (VDDA), and analog ground (AGND).
Table 6.3 Waveform parameters for short-circuit load conditions
Stress Simulator Cable Ipeak Rise time Pulse width

level voltage (kV) length (m) ±10% (A) (nS) ± 10%(μS)
2 +1.0 100 4.5 1–10 0.8

4 +2.0 100 9 1–10 0.8
2 +1.0 200 4.5 1–10 1.6
4 +2.0 200 9 1–10 1.6
The digital encoder converts a 3-bit binary code into a 7-bit code. The code used
by the DAC for MSB and LSB segments. The digital encoder relies on the LVDS-to-
CMOS converter for the clock signal. The corresponding external pins are Test_in
and Test_out, digital VDD (VDDD), and digital VSS (VSSD).
6.2 Digital-to-Analog and Analog-to-Digital Converters 299
The mode control block selects one of the three band gap references to use as a
reference for the DAC. This reference also provides power down conditions for the
DAC. The mode control outputs are internal signals only and thus do not require
separate ESD protection, as they rely on the ESD protection already provided to
the band gaps and the DAC. The 1.2 V NMOS and PMOS devices are used in
the block.
The 1.2 V LVDS-to-CMOS converter converts a low-voltage differential input
signal (LVDS) into a single-ended output signal. Only internal circuit uses this
output signal. The block requires mode control and biasing blocks to function.
The 1.2 V digital interpolation filter circuit block doubles the input data rate by
linear interpolation. The outputs of this block are only connected to the DAC. The
interpolator requires the LVDS-to-CMOS converter to function.
The DAC core produces a programmable current that depends on the input code
of the DAC. This input code is generated by the interpolator. The core requires
external 50 Ohm resistors to the AGND on the IOUTP and IOUTN.
6.2.1.1 ESD Protection Network

A two-power domain ESD protection network is implemented with two digital
(DGND) and analog (AGND) decoupled ground busses. One of the most effective
ways to protect the DAC with two-voltage domain is the rail-based active clamp
ESD design.
In most practical cases, the high pin count of the chip and the multiple (over 7)
layers of thick bond metal or copper top metallization provide appropriate process
capabilities for the rail-based active clamp network.
Practically, the local snapback ESD protection is required for the DAC only in
the case of high voltage, for example, in a charge pump or for EEPROM pins (as
well as certain specified system-level pins).
In principle, the ESD protection network for the DAC can be composed from
either the two-domain active clamp clusters or use only HV devices provided by
DGO process.
In the case of the double voltage domain approach, the 1.2 V digital and 3 V
analog circuit blocks require two separate rail-based ESD network domains based
on the 1.2 V active clamp and the 3 V active clamp.
For high-performance, the low-voltage domain should be designed as a substrate-
isolated circuit. This is achieved using DeepNwell isolation. A corresponding
specific for active clamp design is also related to substrate isolation.
An example of the principal circuit design for a low-voltage active clamp in the
case of a 0.13 μm LV CMOS device is based on a 1.2 V isolated NMOS with W/LG
= 500 μm/0.13 μm (Fig. 6.18). This design can provide an on-state resistance as
low as RON =4 and a leakage current of IDSS =1 μA at the nominal VDD power
supply voltage level VDS ∼ 1.2 V.
This high leakage current is present due to the utilization of 1.2 V devices and
may be considered too high in some cases. The alternative is to use the 3 V active
clamp with a much lower leakage.
ESDP
Nepi
TRIG
Nepi
ESDM
Fig. 6.18 Schematic for isolated 1.2 V slave active clamp
The high-voltage 3 V active clamp design is similar to the already discussed

solution in Chapters 4 and 5. In the case of the 3 V CMOS components with
W/LG =500 μm/0.3 μm, the HVNMOS device can provide a on-state resistance
below 10 and a leakage current below 50 nA at a nominal VDD power supply
level VDS =3 V.
Based upon these parameters, an ESD protection network with only one type
of the 3 V active clamp is sufficient if there is an adequate number of pins. Such
a design provides a significant simplification of the chip layout, especially if the
sequence for the pins for HV and LV domains is mixed in the pad ring.
In this case, the power supply and the other low-voltage domain pins are simply
treated similarly to the I/O pins in the 3 V domain, with a corresponding connec-
tion of the upper and lower ESD diodes to the 3 V ESD plus (ESDP) and ESD
minus (ESDM) metal busses. This design imposes a less stringent requirement for
the low-voltage power supply and ground busses and relies on the rather low-voltage
clamping level provided by the HV Active clamps.
In addition to the “slave” active clamps connected to each pad, four RC-timer
corner cells are embedded in the circuit to complete the distributed ESD network
and control the trigger line potential. The RC-timer cell provides the trigger-off
signal with a delay time of ∼10 μs.
For successful implementation of the design, one of the most critical design items
is the ESD bus. Ideally, the ESD busses use all layers of metal, which reduces the
accumulated resistance across the whole chip, since the bond pads can only use
three or four upper metal layers.
The full test chip simulation result for the 140-pin corner-to-corner ESD bus
resistance is below 1 . At a 3 kV HBM ESD pulse, the device gives only a 1.2 V
peak voltage drop across the ESDP and ESDM buses.
The CMOS input protection circuit is similar to all I/O circuits. It combines a
two-stage network with a 100- polyresistor to provide sufficient CDM immunity.
Each input gate of the circuit is tied to the local VDD and VSS circuit nodes with
minimally sized additional diodes (Fig. 6.19).
6.3 High-Speed Interface IO pins 301
VDDIO
I/O Pad I/O internal
VSSIO
Fig. 6.19 ESD protection for I/O pin with gates tied directly to the pin
6.3 High-Speed Interface IO pins
6.3.1 Interface Analog Products

The main function of the interface circuit is to provide a datalink between different
systems or subsystems. There are a number of interface standards that cover dif-
ferent datalink specifications for transmitting signals through a cable, from several
inches within the same system block to several thousand meters between systems
within the same building or facility.
In comparison with DAC, one the major specifics of advanced interface products
is the high speed of the data transmission (up to 10 Gbit/s and above).
From an ESD specific point of view, a typical requirement may include on-chip
system-level pins. Typically, the cable discharge event (CDE) (Chapter 8) protection
level is specified. Another peculiarity of the interface circuit is the presence of high
common-mode voltages in the input signal.
Just like RF-amplifiers, the high-speed input interface pins require broadband
characteristics [106]. This eliminates the possibility of application of a number
of narrow band, LNA-specific input solutions that rely on ESD pulse bandwidth
filtering [116].
Over recent years, interface applications are heading toward 8–12 Gbps links.
Achieving these goals requires not only increase of speed but also corresponding
reduction in power consumption and noise.
Typically, interface parameters are related to the power moving in the network
and in and out of the integrated circuits. The interface characteristics are often
described in terms of the equivalent S-parameters.
The transducer gain parameter GT represents the ratio of the power delivered to
the load to the power available from the source.
At matching load and source impedance, ZL =ZS =50 and GT =|S21|.
The operating power gain parameter GP represents the ratio of the power
delivered to the load to the power delivered to the network.
The power delivered to the network is dependent on the S11 parameter. In this
case, GP > GT and GT is closer to GP with better input matching.
The available power gain parameter GA represents the ratio of the power
available from network to the power available from source.
Since the power available from the network is dependent on S22, GA > GT , while
GT approaches GA with better load matching.
A functional block diagram an interface application is presented in Fig. 6.20.
LVDS LVDS
Deserialization
Deserialization
LVDS LVDS
Serialization
Serialization
Equalization
Decoding
TX
Input /
Input
LVDS Coding / LVDS

Driver
Output
Alignment
LVDS Channel LVDS
De-emphasis
Differential / Single-ended
LVDS LVDS
LVDS PLL Control PLL Control LVDS
Transmitter Receiver
Fig. 6.20 Function block diagram, for example, of LVDS interface solution
In terms of the ESD protection challenge, the interface application requires the
CDE protection for the selected pins directly interfacing with cable. To address the
design specific, the next section presents the CDE test methodology.
6.3.2 Cable Discharge Event Test Procedure for Integrated Circuits
The cable discharge event (CDE) is a form of system-level stress. System-level pro-
tection requirements are discussed in Chapter 8. CDE is very specific to interface
products.
CDE can be a significant threat to Ethernet transceivers because of the long cables
involved. Until recently, there was no industry standards and limited literature avail-
able on CDE [117–120]. Measured CDE waveforms are obtained by charging short
USB cables and manually discharging them. The discharge waveforms showed a
fast initial overshoot followed by a rectangular pulse. The rise time of the pulses
was estimated to be a few hundred picoseconds. In practical cases, due to the inher-
ent properties of air, discharge repeatability cannot be expected in manual cable
plugging.
The test procedure for integrated circuits was originally developed by industry
leaders [121].
The CDE testing is categorized as either Common-Mode Charging (CMC) or
Differential Mode Charging (DMC). At CMC, all the wires in the cable are charged
simultaneously with respect to a reference. At DMC, the individual wires in the
cable are charged with respect to each other.
At pin sequencing, some of the pins of the cable are connected to the circuit
prior to the other pins, thus resulting in a differential mode signal at the Ethernet
transceiver.
A typical waveform for CDE is presented in Fig. 6.21. Rise time is defined as
the time required for the leading edge of the pulse to increase from 10 to 90% of its
CDE Discharge Currents at 2 kV

10
100 meter cable
200 meter cable
8
4
Current [A]
–2
–4
–6
–8
0 1 2 3 4 5 6 7 8 9 10
Time [uS]
Fig. 6.21 Typical discharge current waveforms measured for a short-circuit load [121]
maximum value. Pulse width is the time interval between the start and end points of
the first peak of the discharge current waveform. The peak current Ipeak in Fig. 6.21
is the maximum current.
Typically, the ESD simulator is equipped with a 200-m long CAT5 Ethernet
cable, a 100-m long CAT5 Ethernet cable, a high-voltage power supply, and a relay
box. The testing may be done with either one or both cables depending on the test
requirements.
The collected discharge current waveforms should comply with the current wave-
forms specified in Fig. 6.21 and by Table 6.3 CDE testing is conducted for voltage
levels with a 500 V step, with Level 1 of +/– 500 V up to Level 4 of +/– 2 kV.
The test setup is shown in Fig. 6.22. In this example, the CDE stress pulses are
applied to the DUT (device under test) soldered onto a test printed circuit board
(PCB). The DUT can also be mounted into a socket on the test board.
The device pins receiving CDE stress (four pins or eight pins depending on the
Ethernet standard used) are accessed through an RJ-45 connector on the board. To
simulate the worst-case stress on the IC, the IC pins should be directly connected to
the RJ-45 connector without any magnetic or discrete components.
First, the charging relay is momentarily closed and opened to charge the Ethernet
cable. Then, the discharge relays in the relay bank are closed all at once or randomly,
depending on the test requirements.
Depending on the Ethernet standard used by the DUT, four of the discharge relays
may not be switched. For example, in the case of a 10 Base T Ethernet standard, only
four pins of the DUT are stressed by the closing of four relays.
Prior to testing the device, waveform verification should be performed at 2 kV
level with the 200-m long cable. A minimum of three samples should be tested
High Voltage
Power Supply 10 MΩ
Resistors To Earth
Ground
Overhead Cable Rack
Charging Relay
CAT5 Ethernet Cable

Test Board Short
Ethernet Cable
DUT
Relay
Bank
To
Ground Relay
Reference Controller
(Cable Rack)
PC
Fig. 6.22 CDE simulation test setup [121]
at each required voltage level. Ten positive and 10 negative discharges should be
applied to all Ethernet ports of the device for each voltage level. A ground terminal
on the test board must be connected to the ground reference (the cable rack).
As a system-level test, CDE testing should be conducted in both the powered and
the unpowered states of the DUT. In the powered mode of testing, the vulnerability
of the DUT to the latch-up phenomenon should be monitored. For that purpose,
the DUT should be powered at the maximum nominal power supply level using
external power supplies. If multiple supplies exist, all supplies need to be turned on.
The supply current(s) must be monitored after each stress to determine whether the
DUT is latched up. If the latch-up phenomenon is observed, the device is considered
to have failed at that stress level.
Each device must pass the electrical, parametric, and functional tests specified in
the device data sheet after exposure to CDE pulses. Each device also must not show
any latch-up during exposure to CDE pulses.
In case the test setup for addressing the system-level on-chip requirements
described above is lacking, elevated package-level requirements are often specified.
For example, instead of a standard package-level HBM 2 kV, an 8 —15 kV HBM
can be specified for the selected interface circuit pins.
Alternatively, cable interface pins can also be tested for meeting the IEC 61000-
4-2 8 kV contact and 15 kV air-gap specifications. However, the charge cable test is
still the main verification tool.
6.3.3 ESD Protection of Interface Pins with CDE Requirements

Typically, for multiple-pin-count interface products, an elevated level of ESD pro-
tection can still be achieved with the active clamp rail-based approach by a simple
scaling of the number of clamps.
General guidelines for implementing system-level protection include connect-
ing the “heavy-metal”/ESD diodes to the external decoupling caps. To achieve
local clamp, high-current-level SCR protection for power-on ESD, the SCR holding
voltage must be higher than the power supply voltage.
This alternative design includes high holding voltage dual-direction clamps
described in detail in Chapters 3 and 4. Wide common-mode range standards
require application of the isolated NMOS and LVTSCR ESD clamps, as well as
the dual-direction SCR’s (DIAC) structures.
The dual-direction voltage tolerance is determined by the signal standards. For
multipoint LVDS (MLVDS), the common-mode voltage range is VCM = –1.4 to
+3.8 V, while the signal voltage Vsignal = –2.6 to +5 V on a 3.3 V supply. For the
eXtended LVDS (XLVDS) interface standard, VCM = –5 to +12 V and Vsignal =
–6.2 to +13.2 V on a 3.3 V supply. Other typical standards are the high-speed HS
(2.5 V) and a slow SMBUS.
Often, for a typical ESD library designed for compact mixed-signal applications
in the absence of the CDE tester, a correlation between the package-level test pulses
can be assumed followed by the agreement with the customer upon the alternative
test specifications. For example, an ESD library compact mixed-signal application
can be designed to meet >8 kV HBM, 300 V MM, and 1250 V CDM.
The most challenging problem for high-speed applications is the low capacitance
of <250 fF of analog IO pads. The pads are protected by ESD diodes that utilize
a micro-SMD bump metal or inductor metal to achieve low capacitance and low
resistance.
The cable interface pins are designed to withstand an ESD discharge with power-
on. Due to the high CDE current, LVTSCR devices with holding voltages Vhold >
Vsupply are used. The target performance is verified by the IEC61000-4-2 8 kV con-
tact/15 kV air-gap tests. Similarly, solid metal/diode paths to external decoupling
caps are required, while the low parasitic capacitance requirement can be below
600 fF.
At the same time, in opposite to the interface pins, the power supply does not
require system-level protection. Active clamp protection on power supply pins and
non-cable interface pins are tested at the package level in power off conditions.
Thus, to protect against CDE, specialized cells are needed to satisfy ESD
protection requirements for both power-off and power-on conditions.
Thus, interface pin protection requires rather challenging device-level ESD solu-
tions. Such solutions compose the system-level capability with a high holding volt-
age, thus incorporating an anti-latch-up feature on the device level; for MLVDS and
XLVDS, the solutions should be dual-direction and provide low parasitic capaci-
tances. At the same time, the clamp should be capable of withstanding high-speed
CDM. If the process supports only a 2.5 V device, the 3.3 V I/O and ESD protection
can be based on stacked ESD devices.
In practical design work, such solutions are usually achieved through extensive
experimentation cycles. Often, the solution demands additional space on the chip
and may be process-sensitive.
The size of the ESD diodes in high-speed I/O can be reduced by 40%, thus
enabling aggressive interface product design.
For the BiCMOS process, the use of the integrated back-to-back dual-direction
devices of superior performance and a small footprint is described in Chapter 4.
Often CDE level can be achieved using conventional active clamp protection
network. An example of the CDE protection for Serializer and Deserializer over a
single differential pair is presented in Fig. 6.23. The product example is specified
to work with cable up to 10 m and designed for automotive video displays with
LVDS standard compliant to ISO10605 “Test methods for road vehicles – electrical
disturbances from electrostatic discharge.”
In addition to standard package level ESD protection requirements system-level
ESD protection is built into the IC. High cable discharge level protection is verified
by ISO gun +/– 10 kV contact and +/– 30 kV air discharge between “pin under
test” and ground. The system-level pins are the LVDS input and output of the single
differential pair.
Video Delta +Control Video Delta+Control
DAC
a)
VDD
External
power supply
Active
Active
Active
Active
Active
Active
Active
Active
Active
Active
Active decoupling
clamp
clamp
clamp
clamp
clamp
clamp
clamp
clamp
clamp
clamp
clamp cap > 0.3uF
b)
Fig. 6.23 Example of interface application network (a) and ESD protection scheme with
application board bypass capacitor (b)
An important part of the protection system and the testing setup itself is applica-
tion board capacitors. In this particular example 0.5 μF bypass capacitor was part
of the system and thus it has been included for ESD evaluation too (Fig. 6.23b).
At the same time the ESD scheme rely on distributed active clamp approach
with specially designed ESD diodes capable of handling very high pulsed currents
of ∼30 A.
6.4 Summary
In this chapter, ESD protection challenges and solutions have been demonstrated
for signal path analog circuits. In the case of signal path products, the ESD protec-
tion network is based on the conventional rail-based or local clamp approaches; the
specifics of ESD protection are related to the product-specific pins.
When the active clamp rail-based approach cannot satisfy the requirements,
major help is expected from device-level ESD solutions for the dedicated pins. The
main challenges of ESD protection are in the minimization of the parasitic influence
of the ESD components on the input signal.
The high pin count signal path analog product packages create challenges for the
CDM ESD pulse protection. Since in high-speed circuits the second-stage resistor
cannot always be of a desired value, the design effort is focused on the appropriate
layout and schematic design of the I/O itself.
While the DAC and ADC can be protected in most cases by applying the same
principles as described for amplifiers, the interface circuits bring in the system-
level CDE protection requirements for pins interfacing with external cables and
transformers. In high data rate pins of 10 Gb/s and beyond, the device-level ESD
protection solutions are perhaps the most important methodology to understand and
implement.
The device-level positive and negative feedback for implementing such solu-
tions is discussed in Chapters 3 and 4. The practical design work of such complex
solutions is usually achieved through extensive experimentation cycles. To reduce
the time-to-market for such R&D work, physical device and mixed-mode simula-
tions are critically important tools. Some of the simulation examples related to this
chapter are presented below.

Example 6.1a–6.1c Rail-Based Protection with Active 5 V NMOS

Clamp and ESD Diodes
Library Name: Examples6_Signal_Path_Analog_Circuits
Project Name: E6.1a_Rail_Protection_5 V_AClamp_IOIO
Project Name: E6.1b_Rail_Protection_5 V_AClamp_ESDPIO
Project Name: E6.1c_Rail_Protection_5 V_AClamp_IOESDM
This set of examples (Fig. E6.1a) illustrates the current path and waveforms
observed at different pin-to-pin combinations in a simple rail-based ESD protec-
tion network with two IO pins and an active NMOS clamp. The HBM waveforms
(Fig. E6.1b) simulated for the example circuits can be compared to results obtained
in the following example for the rail-based network protected by a 5 V snapback
NMOS clamp. A separate analysis can be done for different designs of the Nwell
and Pwell ESD diodes.
C2
HBM Pulse Generator

2.6e-12
R3 Charge L0 R1 Discharge R4
HBM_PULSE
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0 $Vpulse... 1e-10
Active 5 NMOS Clamp Domain

ESDP
M2*
R2 D0* D2*
100000 C3 W=50 W=50
500f
M3*
IO_GND IO_ZAP
W=20
M1* D4*
W=50
C4 D1* D3*
60e-12 W=2000 W=50 W=50
W=5
ESDM
Fig. E6.1a Mixed-mode simulation circuits for different pin-to-pin combinations (IO–IO, IO–
ESP, and IO–ESDM, respectively) and cross-sections of the finite element devices (5 V PMOS,
5 V NMOS, and Nwell and Pwell diodes, respectively) used in the circuits
C2
HBM Pulse Generator

2.6e-12
R3 Charge L0 R1 R4
Discharge HBM_PULSE
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0 $Vpulse...
1e-10

ESDP
M2*
R2 D0* D2*
100000 C3 W=50 W=50
500f
M3*
IO_GND IO_ZAP
W=20
M1* D4*
W=50
C4 D1* D3*
60e-12 W=2000 W=50 W=50
W=5
ESDM
C2
HBM Pulse Generator

2.6e-12
R3 Charge L0 R1 R4
Discharge HBM_PULSE
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0 $Vpulse...
1e-10

ESDP
M2*
R2 D0* D2*
100000 C3 W=50 W=50
500f
IO_GND
M3*
IO_ZAP
W=20
M1* D4*
W=50
C4 D1* D3*
60e-12 W=2000 W=50 W=50
W=5
ESDM
Fig. E6.1a (continued)

Fig. E6.1a (continued)
Example 6.2 Rail-Based Protection with 5 V Snapback NMOS

Clamp and ESD Diodes
Project Name: E6.2_Rail_Protection_5 V_Snap_IOIO
This example (Fig. E6.2) illustrates the current path and waveforms at different
IO and power pin combinations in the case of rail-based protection with an active
clamp. The HBM waveforms simulated in this example can be compared to the
corresponding waveforms in the above example for the rail-based network protected
by the 5 V active NMOS clamp. This example can be used as a basis for analysis of
different designs of the Nwell and Pwell ESD diodes and different snapback clamps,
for example FOX and LVTSCR.
Fig. E6.1b Voltage waveforms at different circuit nodes for different zap combinations
C2
2.6e-12
R3 Charge L0 R1 R4
Discharge HBM_PULSE
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C1
C0 2e-11
V0 $Vpulse...
1e-10
PIN2
ESD_Clamp
D0* D1*
M0 W=50 W=50
PIN0 PIN1
D4*
W=50
IO_ESD_Diodes
W=400
D3* D2*
W=50 W=50
R2*
10K
PIN3
Fig. E6.2 Cross-sections of the finite element devices (5 V SNMOS, and N-well and P-well
diodes, respectively), mixed-mode simulation circuit with two IO pins and snapback NMOS device
and voltage waveforms at different circuit nodes under 2 kV HBM pulse
Example 6.3 Trans Impedance Amplifier

Project Name: E6.3_Trans_Impedane_Amplifier
This example (Fig. E6.3) is assembled using 1.5 V compact model (BSIM3) devices
and demonstrates amplification of the low input current. An ESD protection network
can be added to the circuit to study the voltage overstress at different amplifier
circuit nodes. Amplification of 500 nA input current in the nominal operation regime
is demonstrated in Fig. E6.3.
VDD
R3 R4
R5 R7 160K 60K
0.7K 1K R11*
M1 M3 M5 M7 M9 0.1
R1 R0
Current Input
8K 1.5K
D0
V2 $Vdd*pu...
Output
W=0.5 R2* W=2 W=1 W=100 W=500
W=1 D1
M0 M2 M4 M11 M10 M6 M8
R10 R6 R8 200K
1 R12 1.8K 1.5K
W=1
1 Input Feedback
R9* C1*
100p
1K C0 W=0.5 W=1 W=0.5 W=3.2 W=25 W=50 W=250

V3* $VIN*1....V1 $VIN*pu...
70f
GND
Fig. E6.3 Mixed-mode simulation circuit for compact model trans impedance amplifier and
voltage and current waveforms at different circuit nodes
Example 6.4 CMOS Output Stage ESD Case

Project Name: E6.4_CMOS_Output_Stage_Case
This example (Fig. E6.4) presents a typical case of the CMOS output stage. The
ESD current path can be studied as a function of the drain ballasting resistor values
R2 and R5. The ESD current through the output stage usually results in irreversible
C2
2.6e-12
R3 L0 R1 SW2 R4
SW1 HBM
1e8*(1+... 12.6e-6 1.5K

1e8*(1-...
C1
C0 1e-11
V0 $Vpulse...
1e-10
ESDP
M2*
PGATE
D0
W=40 W=50
C3* R2*
1p 1
OUTPUT
M3*
C4* R5*
1p 1
Snapback NMOS
M1*
Clamp
D1
W=50 W=400
R10
NGATE
10K
W=10
ESDM
Fig. E6.4 Mixed-mode simulation circuit for CMOS output case and the waveforms illustrating
the effect of the drain ballasting resistors on the current path
burnout of the output devices. The circuit can be used to study different pin-to-pin
combinations and the output PMOS breakdown case in particular.
Example 6.5 CMOS Open Drain Case

Project Name: E6.5_Open_Drain_Case
This simulation example (Fig. E6.5) illustrates a current path alternative to the clamp
(M2) through the internal circuit device (M1) as a function of the values of cir-
cuit elements C4, R5, and clamp resistor R10. Typically, the current path through
the internal NMOS device indicates irreversible failure and the current through the
clamp should guarantee reversible ESD operation. At the same time, application of
the snapback NMOS as M1 may provide self-protection of the entire circuit. Both
scenarios can be explored in this simulation example.
C3
2.6e-12
R2 L0 R1 R3
Charge Discharge HBM_PULSE
1e8*(1+... 12.6e-6 1e8*(1-... 1.5K

C2
C1 2e-11
V0 $Vpulse...
1e-10
OUTPUT M2*
C4* R5*
1p 1
M1* W=400
R10*
10000
W=0.5
ESDM
Fig. E6.5 Mixed-mode simulation circuit, snapback NMOS cross-section and current waveforms
through the open drain device M1 under 2 kV HBM pulse
Example 6.6 BJT Output Stage Case

Project Name: E6.6_BJT_Output_Stage_Case
This simulation example (Fig. E6.6) for the open drain case demonstrates the current
path through the internal circuit device Q1 as a function of the emitter ballasting
resistor and clamp Q0 parameters (Fig. E6.6). This case covers output to VCC zap
combination. The typical practical scenarios may include both irreversible failure
and self-protection operation of the Q1.
Chapter 7
Power Management Circuits’ ESD Protection
This chapter discusses the applications of ESD network and clamp principles, using
examples of different power management analog circuits. The trends in the applica-
tion of integrated power products are discussed in Section 7.1. Then, a more detailed
analysis of both the most common and the most “aggressive” ESD protection solu-
tions is presented in the subsequent sections with examples of cases concerning
integrated power products, controllers, and LED drivers.
The specifics of power circuits significantly impact the chosen ESD protec-
tion strategy. Fast-switching and high-voltage pins, with their very limited ESD
protection windows, create a significant challenge and often require a custom,
circuit-dependent approach. Self-protection aspects of integrated power devices
in complex, multiple power domain networks are hard to simulate. Substantial
product-level experimentation is usually required to match power-optimized out-
put devices with the proper protection solutions. These devices usually provide a
rather narrow and sometimes even negative ESD protection window (in its classical
understanding).
Similar to signal path analog circuits, the major principles of local and rail-based
ESD protection network design, as well as ESD clamps architecture, apply to power
analog products. However, a new specific is generated by the multiple current path
scenarios, which can be realized in the power-optimized analog circuit due to the
interaction of the internal components with the power pins.
In the case of fast transient pins connected to output devices with relatively low
absolute maximum voltages, a significant challenge is related to the design and
selection of the local clamps. In this case, a precise triggering voltage reference
is the most desirable feature, which helps to minimize the dV/dt triggering effect.
The possibility of transient latch-up on switching pins or with a hot plug-in
specification also requires countermeasures. Among them are the design and appli-
cation of rather challenging high holding voltage solutions and the utilization of
the self-protection capabilities of the internal circuits. These measures are espe-
cially important in high-voltage circuit cases. At the same time, realization of the
two-stage ESD protection network becomes rather challenging due to performance
losses with the additional resistance in the current path.
These and other aspects of power circuit ESD protection are addressed in the
following sections by means of a more detailed analysis and case studies of the

318 7 Power Management Circuits’ ESD Protection
selected specific analog circuits. This is done in order to create a substantial basis
of general understanding that can be subsequently applied to arbitrary power analog
circuits.
7.1 Power Management Products
7.1.1 Power Management Products and ESD Challenges
7.1.1.1 Market Trends

Understanding of ESD design for power products cannot be complete without con-
sideration of the process technology specifics. One of the major trends for power
management circuits today is their integration into practically all of the analog
circuits.
DC/DC converters are usually a part of the power modules used today in telecom-
munication and data communication, computer and office automation, and industrial
control and instrumentation. The fastest growing areas of use today is automotive
electronics and general automation.
Voltage regulators are used in consumer electronics, computer and office automa-
tion, and tele/data communication.
Pulse width modulation (PWM) and PFM controllers are used in merchant
DC/DC converters, computer and office automation, consumer electronics, general
automation, and industrial control and instrumentation.
One of the major trends in power analog products is the increase in the output
current of the integrated power regulator.
A current level of 4–8 A is used for point of load (POL) regulation; notebook
computers; portable instruments, distributed power systems, battery powered equip-
ment; high-end printers; DSP, memory supplies; automotive; distributed 2.5, 3.3,
5 V power systems; cellular base stations; xDSL modems. Even higher current (10 A
and above) is required for ASIC, DDR memory; fiberoptics power supplies; telecom
routers; switchers; and other network applications.
During recent years, another growing area of high-current applications includes
large LED displays and LED lighting, with focus on backlighting, automotive, and
general illumination applications.
The broad market portfolio usually includes building blocks for switching regu-
lators and switching controllers with an input voltage of ∼2 to ∼100 V and a current
of up to 10 A, LDOs of 0.2 to ∼10 A, and a high, medium, or low precision band
gap reference.
There are many other power analog and mixed signal applications. One of these
applications is in the power over Ethernet (PoE). PoE uses existing Ethernet infras-
tructure to deliver power at a voltage level of 48–52 V, with a safety limit of ∼60 V
based upon the IEEE 802.3 standard. The maximum power is 30 W or greater. PoE
combines power-sourcing equipment such as Ethernet routers, switches, and hubs
at one end of the cable. At the other end is a powered device, such as a VoIP or,
WLAN access point, network camera, or security system. In a typical application,
7.1 Power Management Products 319
the power can be sourced from the power-sourcing equipment through the data lines
via data transformer center taps.
One of the PoE challenges is the capability of the high-voltage mixed-signal inte-
grated circuit with a fast digital signal to operate at 60 V. Other challenges include
conversion efficiency, transient protection, ease-of-use/support, and the need for
galvanic isolation.
Overall, the modern trends in this power product combine the following features:
a high switching frequency, higher current levels, higher efficiency, and low power
consumption during standby.
Adding to the value of this solution are enabling features such as power good,
sync, sequence, tracking, and multiple outputs. Another advanced feature is digi-
tal control. Digital control allows for switchers with programmability and remote
monitoring capabilities. The overall target is to design smaller, cheaper and faster
switchers, as well as to provide high-voltage analog products (up to 100 V). The
trends in low-voltage analog products are a higher output current (up to 20 A) and
lower regulated output voltages.
Switching voltage regulators with a switching frequency of up to 50 MHz are
currently under research for possible reduction of the form factor, using integrated
inductors and capacitors or other integrated and co-packaged passives on the chip.
Unique performance characteristics can be provided by serial digital power con-
trol of analog and digital feedback loops, using data converters for digital feedback
loop and monitoring.
7.1.1.2 ESD Challenges

Power analog product demands create significant challenges for on-chip ESD design
(Table 7.1).
Table 7.1 ESD challenges generated by power market demands
Power market demands ESD challenges
Low-cost solutions; smaller Small footprint of ESD cells, under-the-bond pad ESD cells,
form factor solutions self-protection
Fewest external New ESD circuits to explore on the product level
components
Highest efficiency Low parasitic structures; ESD protection of very fast switching
voltage nodes; low substrate noise
More current capability Lower ESD protection window; power array protection, internal
circuit ESD damage through the power array ESD signal
transmission
Higher total output voltage Precise ESD clamp characteristics
accuracy
Dynamic control; faster Voltage reference ESD circuits with low dV/dt effect
response
Control and Multiple voltage domains; EEPROM protection, local snapback
programmability; smart protection for digital and low-voltage control pins
solutions
To address the demands of the power market, a significant evolution of the

integrated process technology is required.
The integrated components of the CMOS, BiCMOS, or BCD processes are
designed toward increasing the switching speed. One of the major changes in phys-
ical parameters is the decrease of the RDSON CGS product figure of merit, where
RDSON and CGS are on-state resistance and gate–source capacitance of the integrated
power component, respectively.
Reducing the form factor by integrating the typical external components on the
chip, mainly the inductor and capacitor, allows for the design of very high frequency
(>20 MHz) efficient switchers, presenting state-of-the-art solutions.
One of the major challenges of this solution is related to the power density of
the integrated transistor arrays. On the technology level, this problem is addressed
by the implementation of thick and thin copper-top metallization with low sheet
resistance. Thus, other major figures of merit for integrated power arrays are the
product of RDSON and the area consumed by the power device (R × A).
The trend toward control, programmability, and smart power circuits requires
small signal circuit blocks. These blocks require an improved density with integra-
tion of digital and enhanced analog contents.
Many critical (life support) system applications, for example, medical and auto-
motive, require a new level of ESD protection to guarantee a robust and reliable
system performance, even in the case of overstress and upset events.
For portable applications, the small form factor and low cost require a compact,
space-saving ESD design with under-bond pad solutions or solutions integrated in
the power structure. Practically, integration of the ESD robustness into the process
is required to avoid overdesign of the ESD structures. The usual expectation is that
ESD clamp size is kept below the pad size. Structures that can be created in the
process are expected to enable ESD clamps with dV/dt-triggered solutions (at least
for switch, bootstrap, and power supply nodes).
In these conditions, the solution strategy is refocused on the protection of the
individual internal circuit devices rather than the rails of the analog power cir-
cuit. This means, for example, that the high-side nodes are protected to the switch
node, rather than to the ground, and only the low-side nodes are protected to the
ground.
The new demands of analog power circuits also impact semiconductor pro-
cess technologies. The most power-optimized processes include superior copper
top metallization; high-density, low-voltage CMOS parts – down to 0.13 μm; an
SOI process to eliminate substrate noise coupling; a superior laser trim capabil-
ity (relative to Zener trim); power-optimized integrated high-current-density switch
arrays.
These semiconductor process capabilities target low-cost power management
units (PMU), digitally controlled switchers, and analog products with high voltage
capabilities. They also provide voltage scaling capabilities.
The process development trend across industry generates more and more
advanced and sophisticated BCD technologies under high cost-effectiveness.
Process technology is focused on the advanced drain–source on-state resistance

RDSON vs. the drain–source breakdown voltage, switching speed, as well as the
elimination of latch-up effects at higher current conditions by effective deep trench
isolation (10 μm and above), combined with high-energy buried layer and sinker
implants.
A significant upgrade of the simulation tools is usually required before simulating
the behavior of integrated switchers, taking into account the coupled substrate noise
and ESD capability.
At high power density, constraints of integration of another technology become
related to thermal packaging characteristics at higher power. Similar to discrete
components, a substantial reduction of the thermal resistance is required.
Another packaging characteristic concerns finding a better way to produce high
current out of the package with a small form factor. This involves advanced ball-
bond and wire-bond technologies.
7.1.2 Integrated DC–DC Converters and Controllers

In this chapter, many application examples are related to switching voltage regu-
lators. Therefore, this section provides some minimal introductory material on the
topic. Among the DC–DC convertor product family, some of the most common
products are the step-down buck and step-up boost switching blocks. However, other
convertors, like buck-boost, multiphase, or resonant [122–124], can be designed
with the same basic principles, even for producing a voltage of opposite polarity to
the input voltage.
The most simple implementations of buck and boost converters using integrated
CMOS power devices are shown in Fig. 7.1 for an asynchronous switching princi-
ple that involves the flyback diode and for a synchronous switching principle that
involves two switching devices.
A buck converter (Fig. 7.1a, b) operates in continuous mode by using the cur-
rent through the inductor supported by the repeatable commutation cycle. During
the first phase, the MOSFET switch provides initial current through the inductor.
Then, the energy stored in the inductor is transferred to the output of the converter
and integrated in the constant voltage by the capacitor. After turn-off of the switch
device, the current path is provided by the flyback diode. The total integrated output
voltage is determined by the duty cycle controlled by the driver and varies linearly
with the duty cycle for a given input voltage.
In an asynchronous buck converter, the freewheeling diode turns on after the
switch turns off, as a result of the rising voltage across the diode. This voltage drop
across the diode results into a corresponding power loss.
To overcome this loss, the synchronous buck converter is built with a modified
version of the basic buck converter circuit topology, in which the diode function
is performed by a second switching power array. In spite of the increased cost, the
improved efficiency of this design is often preferable in portable applications.
VIN VIN
G D G D
B B
S VOUT S VOUT
L D L
G
C R B C R
PGND S
PGND
a) b)
VIN VOUT
VOUT VOUT
VIN VIN
c)
Fig. 7.1 Typical circuit diagrams explaining the operation principles for the step-down (buck)
asynchronous (a) and synchronous (b) and step-up (boost) voltage regulator with diagrams
illustrating principles of operation (c)
Another advantage of the synchronous converter is its bi-directionality. When

power is transferred in the “reverse” direction, it provides a boost action that can be
used to provide a power supply for the high-side driver.
A typical example of buck converter application is different battery chargers for
portable electronic devices using a dc voltage supply, for example, an automotive
on-board 12 V power outlet.
Alternatively, batteries, solar panels, or rectifiers can provide the power. To con-
vert the original DC voltage level to a higher level, step-up or boost converters are
used (Fig. 7.1c). For high efficiency, the switch must turn on and off quickly and
have low losses. Therefore, MOSFET devices are usually chosen for highly efficient
circuits.
Examples of application for boost circuits are portable LED lighting and hybrid
cars, as well as cold cathode fluorescent tubes (CCFL) in devices such as LCD
backlights.
The key principle that drives the boost converter is the inherent tendency of an
inductor to resist changes in current. When being charged, the inductor acts as a
load and absorbs energy, and when being discharged, it acts as an energy source.
The voltage the inductor produces during the discharge phase is related to the rate
of change of the current and not to the original charging voltage, thus allowing
difference between input and output voltages.
In the on-state, the switch provides an increase in the inductor current. In the off-
state, the switch is open and the only path offered to the inductor current is through
the flyback diode D1 , the capacitor COUT , and the load RLOAD . This path results in
the transfer of the energy accumulated by the inductor during the on-state into the
capacitor.
The boost converter may be operated in either the continuous or discontinuous
conduction mode. Operation in the continuous mode is more efficient and provides
lower EMI characteristics than the discontinuous mode. In continuous conduction
mode (when the inductor current never reaches zero), the boost regulator operates
in two cycles. In the first cycle of operation, MOSFET Q is turned on and energy
is stored in the inductor. During this cycle, diode D1 is reverse-biased and the load
current is supplied by the output capacitor, COUT . In the second cycle, MOSFET Q is
off and the diode is forward-biased. The energy stored in the inductor is transferred
to the load and output capacitor. The ratio of these two cycles determines the output
voltage.
A more detailed analysis and analytical description of converter operation can be
found in [122–125].
Similar to buck converters, a synchronous version can provide better efficiency
with the additional cost of another integrated NMOS array.
The typical ESD protection solutions for converter pins are discussed in detail in
the following two sections, separately for low-voltage and high-voltage converters.
From the circuit diagrams (Fig. 7.1), it is already expected that switch node protec-
tion will be a challenge, due to the fast transient voltages at this pin (connected to
the inductor). Also challenging is the transient voltage at the input pin, due to the
inductance of the bond wire at high current.
In general, DC–DC converters can be designed with different integrated or
discrete components. Usually, high-quality factor passives are either external or
integrated into the package. An alternative converter design may include integrated
controller circuit driving the gates of discrete FET’s, IGBT or GaN power devices.
The ESD specifics of controllers will also be discussed in this chapter.
7.1.3 Integrated Power Arrays

7.1.3.1 Power Losses
One of the most intriguing, and complex topics related to analog power circuits is
the topic of integrated power devices. The design of integrated power components
greatly impacts the figures of merit of analog power products.
In the case of low-voltage power components, standard NMOS and PMOS are
used. In the case of high-voltage circuits, integrated lateral double-diffusion MOS
(LDMOS) (Fig. 7.2a) or vertical DMOS (VDMOS) (Fig. 7.2b) architecture is used
for the integrated power device. A more advanced solution can be based on trench
architecture, which are also used in discrete components (Fig. 7.2c). While the max-
imum operating voltage of the LDMOS devices is scalable by the lateral drift region,
a) b)
c)
Drain
Cu Top,
Metals &
Vias
Gate Bond Wire
Bulk
Lead frame
FET
Solder
Source
d) e)
Fig. 7.2 Cross sections of RESURF lateral (a), vertical (b), and trench DMOS (c) devices;
schematic representation of the integrated power component with parasitic components (d); and
illustration of the resistive components in the final package (e) [126]
the vertical devices usually provide a fixed voltage level determined by the thickness
and doping profile of the N-epi region.
The major goal in a switching circuits design is to increase the power efficiency of
the circuit. There are five major loss channels associated with the switching device:
(i) The drain–source on-state resistance loss PRDSON

(ii) The gate charge loss PGATE
(iii) The switching losses PSW
(iv) The deadtime losses PDT
(v) The reverse recovery losses PRR
The drain–source on-state resistance RDSON is an equivalent resistance from

drain to source. It simply leads to resistive-type losses in the switching circuit
according to PRDSON = Irms 2 R
DSON , where Irms is the random mean square current
in the circuit. In real applications, the resistive losses are accumulated across the
entire network within switching current path. The total resistance is accumulated by
resistive components from the print circuit board (PCB), solder, frame, bond wires,
metallization layers, transistor, and inductor Fig. 7.2e:
RTOTAL = Rboard + Rsolder + Rframe + Rwire + Rmetals + RFET + Rinductor

The gate charge loss Qg represents the amount of charge necessary to raise the
gate/source potential from the off state to a given voltage for on-state conditions
(Fig. 7.3). In this case, the loss is proportional to the switching frequency fsw and
the voltage level provided by the external power supply or the internally regulated
voltage of the gate driver VCC ; PGATE = Qg VCC fsw .
Turn-On: Turn-Off:
Vcc Vcc
Cgd
Cgd
Rg Rg
Cds Cds
Cgs Cgs
Fig. 7.3 Illustration of the gate charge losses in MOSFET during turn-on and turn-off [126]
Switching losses occur when the power FET is transitioning between the off-
and on-states. The loss is affected by most of the parasitic resistances and capac-
itances associated with the device. The switching loss can be estimated according
to the expression PSW = 12 fSW VDS ID (tr + tf ), where tr and tf are the correspond-
ing times for turn-on and turn-off processes, respectively. Essentially, the switching
loss occurs due to the finite time of overlap of the switch current and switch voltage
waveforms. Respectively, the larger the gate charge Qg , the longer the rise and fall
time and thus the larger the switching loss component (Fig. 7.4).
The deadtime loss is related to the conduction process in the body diode during
the deadtimes: PDT = fSW VDiode · ID (tDT1 + tDT2 ). During synchronous switcher
deadtime, the power FET is off and the low-side body diode does not conduct. In an
asynchronous switcher, the “deadtime” loss occurs during the entire duration of the
off-state.
The reverse recovery losses are related to the charge QRR accumulated across
the junction. This charge must be recovered when switching resumes (Fig. 7.5). In
order to reverse bias, it is necessary to remove the accumulated injected charge in the
depletion region of the diode. This is achieved by using current from the high-side
MOSFET and creates the reverse recovery loss. Reverse recovery (RR) time can
slow the turn-on transition and increase switching losses as well. Smaller deadtimes
lead to smaller RR losses: PRR = fSW · VIN · QRR .
The peak efficiency point is primarily controlled by the on-state resistance
RDSON and Qg . Thus, the integrated or discrete FETs inside (or off-chip) a switcher
determine the peak efficiency point (Fig. 7.6).
A major consideration is a trade-off between product performance and manufac-
turing cost. The cost increases with the number of smaller features in the design;
the number of mask and process layers; die size; Cu top usage; package type; multi
Vswitch
VSWITCH Overlap of Switch

ID Current and Voltage in
Shaded Area
tf
Fig. 7.4 Illustration diagrams for the MOSFET switching losses [126]
Vswitch
VPEAK
IF
Voltage
dIF/tRR
tRR time
QRR Current
IRR dIRR/dt
Fig. 7.5 MOSFET losses – deadtime and reverse recovery [126]
die vs. single die usage; die attachment steps; number of bond wires; lead frame,
uSMD, and material vs. process costs.
Thus, efforts are focused not only on reducing the resistance in the mean current
path to the lowest square count possible in the metal, but also on avoiding current
crowding and being able to bond to the package.
With increase in the total power array area, the relative contribution of the chan-
nel resistance is reduced and metallization resistance become dominant (Fig. 7.7a).
There are several layout techniques that are mainly empirically developed. For
example, the “interdigitated” or so-called “Christmas tree” metal pattern helps to
avoid current crowding (Fig. 7.7b). The main idea of this structure is to roughly
equalize the resistance to any point within the FET. Sometimes, current density may
RdsonLoss
Power Loss
Gate Charge Loss

Switching Loss
Deadtime Loss Efficiency
Reverse recovery
Increased Fsw
RDSON,
Qg, Iq, LDCR
LCORELOSS
Load current
Peak Load Current
Fig. 7.6 Losses over output current [126]
R
Metal R
Thicker metal
Lower aspect ratio
Area
a)
PS01 grey = bump metal
N+ alternating S&D
b) c)
Fig. 7.7 Dependence of the array resistance components upon area (a), example of the “inter-
digitated” copper top (b); and three-metal micro-SMD layout for power FET array (c) [126]
require parallel layers. This approach might be especially effective with a copper
top metal layers.
One of the major trade-offs for RDSON is the reduction of the breakdown voltages.
Data across industry analog processes are summarized by feature size in Fig. 7.8 for
published sub-50 V processes. One of the major problems is related to the ESD pro-
tection capability of the particular design of a power array. This problem is brought
into focus in the following section.
100
90
80
Rdson.sp (mohms.mm2)
70
0.7um
60
50 0.7um
0.5um
40
0.5um
30 0.5um Jazz 0.18um
Freescale
SMOS8
20
0.5um
0.5um TI LBC7
10
(@Vgs=7V)
0
10 20 30 40 50 60
Breakdown Voltage (V)
0.5-0.7um Nodes 0.25-0.35um Nodes
=<0.18um Nodes NSC Processes
Fig. 7.8 Dependence of the on-state resistance per unit area vs. breakdown voltage across industry
analog processes derived from published data
7.1.3.2 Self-Protection Capability (SPC) of Integrated Power Arrays

According to the previous section, power arrays are primarily optimized to minimize
losses. This, however, does not automatically guarantee a suitable performance of
the arrays in ESD conditions [127, 128]. The aspects of array self-protection capa-
bility (SPC) assessment by experimental methods are the focus of this section, with
the simulation methods described in the following section.
It is logical to assume that the self-protection capability of the power array is
determined by two coupled factors: device-level effects, governed by the device
design, and array-level effects, mainly determined by the distributed current con-
duction across the array surface from the pads. Device regions and doping profiles
determine the critical avalanche current per micron width, on-state parameters, and
gate coupling. The array layout design is responsible for the non-uniform current
conduction of the array due to both the distributed interconnect resistance and the
gate network.
In principle, in power analog ICs with integrated power arrays, a major challenge
is the protection of the IC pins connected to the power array. Most typical examples
of such pins include the switch pin in DC–DC converters, the gate pin in controllers,
and the output pins in LED drivers. Due to high transient voltages, an ESD clamp
cannot be used and the ESD protection network typically relies on the SPC of the
power array itself.
At the same time, for an array voltage above 20 V, the snapback event usually
results in immediate device failure [129], with exception of some rare cases [130].
This phenomenon is observed even in very large integrated power arrays.
An example of failure in a large integrated 100 V power array with 60 mm total
gate width is presented in Fig. 7.9. The array is a part of an asynchronous buck
switching voltage regulator circuit. The local burnout of the array is observed at
a 2 kV HBM stress. If calculated with assumption of uniform current distribution
across the array, the failure corresponds to an average current density of only 22 μA
per micron width.
10
Vgs = 0V
Vgs = 1V
Vgs = 2V
1 Vgs = 3V
Pulsed Current (A)
Vgs = 4V
Vgs = 5V
Vgs = 7V
Vgs = 10V
0.1 Vgs = 12V
Vgs = 14V
0.01
0.001
0.0001
0 20 40 60 80 100 120 140
Pulsed Voltage (V)
Fig. 7.9 Local W = 60 mm array damage in power IC case and 100 ns pulsed characteristics of
the output array for W = 1 mm
As it was described in the section above, integrated analog power products are
usually designed with significant trade-off between the on-state resistance and the
breakdown voltage. As a result of such aggressive optimization, the SPC of an array
in the ESD pulse regime is primarily determined by the amount of avalanche current
level that the array can withstand prior to entering the irreversible snapback mode.
This figure of merit is somewhat similar to the avalanche energy listed in the data
sheets of discrete MOSFET and IGBT components (Section 8.5).
Several components of successful power array design include (i) an experimen-

tal methodology for evaluation of the self-protection capability of integrated power
arrays in ESD pulse regimes and (ii) design methods to improve the SPC or the
NLDMOS arrays on the device and array levels.
It is important to emphasize that the focus of this section is on improving the
SPC of large integrated arrays by increasing the current in the critical regime before
snapback rather than attempting to achieve a reversible operation of the NLDMOS
device in the snapback mode [129], in opposite to the examples described in Section
4.5. In general cases, attempts to achieve reversible operation in snapback mode are
usually unsuccessful [130].
In particular, a significant increase in the SPC of the array in non-snapback
mode can be achieved by avalanche current balancing using a rather deep N-sinker
implant in the drain-diffusion region. A practical methodology for array comparison
is proposed and experimentally validated in [131].
A comparative analysis of the arrays with different designs is presented below.
All measurements presented have been carried out on wafer level. For compara-
tive analysis of the DC electrical parameters of the power arrays, an industrial HP
electrical tester has been used to obtain precise array on-state resistance (for drain–
source voltage of 0.1 V and 5 V gate bias) and breakdown voltage (at 1 μA drain
current).
The self-protection capability is characterized as the dependence of the critical
HBM failure level on constant gate bias (Fig. 7.10a, b). Additionally, transmission
line pulse (TLP) measurements have been used in a similar setup (Fig. 7.10a) to
obtain the 10 ns rise time pulsed SOA of the arrays and to establish the dependence
of the snapback triggering point (IT1 and VT1 ) on gate bias, as in Fig. 7.9.
The experiments presented below investigate the effects of drain contact region
changes and alternative source region design on the self-protection capability of
1 and 24 mm total gate width NLDMOS arrays.
The layout of the original design of the NLDMOS array is presented in
Fig. 7.10c. This design combines an n+ -drain region with an n-drift implant and
has a polygate region enclosing the source n+ -region and the p+ and Pbody regions.
Using conventional NLDMOS design practice, the second diffusion Pbody is
aligned by a polymask to form the body of the MOS structure.
The critical avalanche current of the NLDMOS is determined by the para-
sitic n–p–n structure. The degrees of freedom for improvement of this avalanche
current level are limited by two major constraints: (i) improve the array SPC with-
out process change, i.e., using only available mask layers (for example, varying
the Pbody implant dose to reduce the internal base resistance of the parasitic
n–p–n, which has been optimized for the operational regime of the NLDMOS,
is not allowed); (ii) deliver a solution that provides on-state resistance and other
relevant figures of merit that meet or exceed those of the reference design
(Fig. 7.10c).
Three distinct and appropriate methods of increasing SPC have been incorporated
into experiments based on related studies [129, 130, 132–134]:
y
8000
7000 60mm 30mm ze 2
arraysi
ARRAY
6000
HBM LEVEL (V)

Drain SIZE (mm)
5000
4000
100 V
Gate Power 3000
NLDMOS
Array 2000 7.5mm
array size3
DC Gate-
Power Source 1000
Supply Decoupling 0
Cap 0 1 2 3 4 5 6 7 8
Source
GATE BIAS (V)
a) b)
Drain Gate Source Pbody mask Nwell End-of Finger pplus
Interdigitated
Pbody contact
design
p+
n+
L 0.92xL
c) d)
Fig. 7.10 Experimental setup for SPC evaluation (a); measured dependence of the critical HBM
level upon the dc gate bias for different array sizes (b); snapshot layout view of the conventional
100 V NLDMOS array (c); and the array with modification of several design parameters including
8% reduction in size (d)
The first alternative device design implements interdigitated n+ -source/p+ -Pbody

diffusion regions (Fig. 7.10d). This approach has been previously used [4, 5] pri-
marily as a measure to achieve reversible snapback operation. Interdigitation ratios
range from every other contact to only six p+ -regions along the 100 μm device fin-
gers. This approach brings the added benefit of reduced source region layout area.
A chief concern with this approach is possible impact on RDSON .
The second device design measure includes blocking the n+ -implant at the ends
of the fingers (Fig. 7.10d). A drawback of this approach is the slight RDSON loss due
to reduction in the source layout area in arrays with short fingers.
The third measure includes additional Nwell or Nsinker implants in the drain
contact region. This measure has been discussed as an improvement that may
provide reversible snapback operation.
For each process technology it is important to understand and quantify the effects
of these measures on SPC improvement.
In analog circuit applications, different conditions can be realized at the gate of
a power NLDMOS array, so comparison of the self-protection capabilities of power
arrays requires gate control of the arrays. In this study the dependence of the critical
HBM level (the highest voltage HBM discharge that the array safely withstands)
upon gate bias is treated as the key SPC figure of merit.
According to the SPC dependence (Fig. 7.10b) the total HBM current through
the array is a superposition of two components, channel and avalanche current. The
channel current depends on the DC gate bias and the gate coupling during a transient
HBM event. As gate bias increases, an increasing fraction of the HBM current is
safely conducted through the channel, leading to the observed increase in HBM
passing level.
The critical avalanche current is limited by the electrical avalanche–injection
current instability, which is caused by the parasitic n–p–n BJT structure in the
NLDMOS device. In the avalanche–injection regime, conductivity modulation
under the critical current density is controlled by the parameters responsible for
current gain and avalanche multiplication in the parasitic n–p–n structure.
The major results of the 1 mm array studies at 0 and 5 V gate bias are summa-
rized in Fig. 7.11 and Table 7.2, in comparison of the original reference NLDMOS
device A01 with improved devices A16, A17, and A21. The best device, including
all three features described above, is compared to the reference device in Fig. 7.10c.
These measures have increased the critical avalanche current at zero gate bias by
two orders of magnitude, and the snapback voltage has been increased by 15–20 V
at both 0 and 5 V gate bias (Fig. 7.11a).
The following conclusions can be drawn based on the analysis of the experimen-
tal results (Table 7.2). The RDSON loss due to interdigitation of the source–Pbody is
insignificant. For the minimum n+ -source area, when n+ -contacts are only 50% of
the source region area (alternating with p+ -diffusions), RDSON increases only 2.6%,
which is compensated by a 7.9% RDSON improvement due to the smaller cell dimen-
sion. For the device with only a 1:9 p+ /Pbody to n+ -source area ratio, the RDSON
increase due to the reduced source area is only 1%, with a total RDSON improve-
ment of ∼7%. However, interdigitation of the source and Pbody does not by itself
provide any SPC advantage. The critical current of these arrays is similar to that of
the reference array.
Both the drain Nwell implant and the end of finger p+ -implant lead to at least one
order of magnitude improvement in the critical avalanche current.
For the array already upgraded with the drain Nwell implant, the end of finger
p+ -implant improves the avalanche current level by a factor of 2.
These various methods also lead to an increase in the snapback voltage of the
array at the gate bias of 5 V.
These results and conclusions for 1 mm arrays are confirmed by results from
24 mm arrays with the same design alterations.
The effects of the end of finger p+ -diffusion and the inclusion of Nwell or Nsinker
in the drain have been analyzed and supported by numerical simulation. An N-sinker
0.16
0.15
0.14
0.13
0.12
TLP Drain Current (A)

0.11
0.1
0.09
0.08 Q4 P5 5V
Q4 P5 0V
0.07
Q2 P8 5V
0.06
Q2 P8 0V
0.05
0.04
0.03
0.02
0.01
0
0 25 50 75 100 125 150
Drain-Source TLP Voltage (V)
a)
0.015 A17-Interdigitated, end of finger

0.014 p+; Nwell; IT1=11μA/1μm (>100x)
0.013
0.012
TLP Drain Current (A)
0.011 A21-interdigitated; no Nwell

0.01 IT1=6.5μA/μm (>50x)
0.009
0.008
0.007 A16-interdigitated; Nwell
0.006 IT1=3μA/μm (>30x)
Q5 P2 0V
0.005 Q5 P2 5V
0.004 Q4 P2
Q4 P2
0V
5V
0.003 Q4 P5 5V A04 (not shown), Nwell, no
0.002
Q4 P5
Q2 P8
0V
5V
interdigitation IBR=0.5μA/μm 5x
Q2 P8 0V
0.001
0 A01 standard POR device
115 120 125 130 135 140 145 IT1<0.1μA/μm (Ref)
Drain-Source TLP Voltage (V)
b)
Fig. 7.11 The effect of different array design features. Comparison of the experimental TLP
ID –VDS characteristics for the standard NLDMOS and the devices with different design changes
implant provides significant improvement of the critical avalanche current level,

but requires a substantial increase in the drain length of several microns to avoid
breakdown voltage reduction and is thus impractical.
As expected the major effects can be understood using avalanche–injection con-
ductivity modulation in n–p–n structure with corresponding critical conditions αM>
1, where α is the current gain and M is multiplication coefficient. The Nwell
implant is too light to have a significant effect on current ballasting when implanted
at the drain. According to 2D TCAD analysis (Fig. 7.12b), the benefit of drain
Nwell implant is reduction of the avalanche multiplication factor M, resulting in
achievement of the instability criterion at a higher critical voltage and current, thus
providing the SOA improvement.
Table 7.2 Pulsed and DC characteristics for reference and improved devices
Vgs = 0 Vgs = 5 V
1 mm VT1 IT1 VT1 IT1 Vbr RDSON

cell Features description (V) (mA) (V) (mA) (V) × area
A01 Reference 120 0.018 88.6 156 117.4 363

A16 Nwell, interdigitated 129 3.05 102 148 108.3 322
with nine p+
islands
A17 Nwell, interdigitated 142 11.1 105 146 108.7 339
with nine p+
islands and p+ end
of fingers
A21 No Nwell, 137 6.39 97.3 140 111.8 359
interdigitated with
nine p+ islands
and p+ end of
fingers
Understanding the second dominant effect requires taking into account possible
manufacturing effects resulting in the formation of spherical (as opposed to cylin-
drical) junctions in the corners of the poly ring due to the tilted implant shading.
As a result the base of the parasitic n–p–n structure in the corners becomes lightly
doped, thus providing elevated gain α and enabling the avalanche–injection current
instability in the device at low critical currents. The doping profile obtained from
3D process simulation shown in Fig. 7.12c demonstrates surface reduction of the
developed doping level in the corners of the polygate.
Thus a significant increase in the SPC of the array in non-snapback mode can
be achieved by both changing the avalanche multiplication coefficient at the drain
region using an Nwell implant and eliminating the cylindrical junction in the corners
of the source region. This improvement is achieved using “free” approaches that do
not require process alteration or result in degradation of the major DC and switching
characteristics of the power arrays.
7.1.3.3 Physical Simulation of DeMOS Power Arrays in ESD Regime

A simulation methodology [135] applied to study the physical effects in large high-
voltage power arrays demonstrates that the critical HBM pulse regime for snapback
in a power array is a non-linear function of the gate bias. This simulation has also
been used to visualize the scenario of local burnout of the power array as a result
of localized filament formation. Analysis of array operation under ESD stress is
complicated by the unpredictable effect of the gate driver circuit on gate coupling.
The physical processes in ESD conditions in such a large distributed object as a
power array are not well understood due to the grid size limitations of conventional
simulation tools.
1.0
GATE BIAS (V)
DRAIN CURRENT (mA/um)

0.8
0.6
0.4
7
Pwell 0.2 5
3
0.0 0
0 20 40 60 80 100 120 140
DRAIN-SOURCE VOLTAGE (V)
a)
b)
Polygate
SiO2 Spacer
Si
c)
Fig. 7.12 A 100 V NLDMOS TCAD cross section (a) and comparison of TCAD ID –VDS char-
acteristics for the standard NLDMOS and the device with additional HNWELL implant (b); (c)
3D structure of the NLDMOS gate–source segment and simulated doping distribution produced
by implants (P-body, NLDD, n+ ) after all diffusions with dose, energy, tilt and rotation from 2D
calibrated process flow
The methodology for analysis of snapback behavior and current distributions in

power cells and arrays under ESD stress has been first proposed and experimentally
validated in [135]. It includes extraction of the array netlist using the interconnect
clustering method, construction of the equivalent netlist using compact snapback
device models (Fig. 7.13), and circuit simulation and visualization of the current
distribution. This methodology enables mapping of the current distributions across
the surface of the array in order to visualize the current distribution on transient
timescales. The specific goal of the simulation is to understand the physical pro-
cesses responsible for distributed power array burnout under ESD stress. Successful
modeling of the non-uniform current distributions in large arrays may assist opti-
mization of the cell and array topologies and the safe operation area in the ESD
regime.
A 2.5D simulation of a 24 mm power NDeMOS array has been carried out across
a range of constant gate biases. The simulated waveforms produced by the array
under stepped increases of the HBM pulse have been monitored and recorded. The
SOURCE Drain DRAIN

Rcs Rcd cluster
Source
cluster
Pads
Source
connection Drain
connection
Fig. 7.13 Extracted distributed RC model of the power array
point of snapback is considered to correspond to the critical regime that leads to real
array burnout.
Examples of the simulation results are shown below. At zero gate bias, reversible
non-snapback operation is possible only up to the low HBM level of ∼80 V
due to a finite avalanche current and coupling of the non-silicided gate of the
array. At a critical voltage level near ∼90 V, however, a significant voltage drop
and current increase indicate snapback of the array (Fig. 7.14a for 90 V HBM).
For higher HBM amplitudes, the delay time for snapback is reduced (Fig. 7.14a
for 1 kV HBM).
One of the important simulation results [135] is that the current distribution
remains uniform both before and after snapback (Fig. 7.14c). Only after certain
delay is a highly non-uniform state formed in the array (Fig. 7.14d). This non-
uniform state corresponds to the current filament solution and is observed until the
end of the ESD pulse.
When a DC gate bias is applied, the passing level of the array increases due to
the addition of a channel current to the avalanche current. The dependence of the
passing level upon the gate bias is presented in Fig. 7.15. Depending on the gate
bias conditions, the filament can be spontaneously formed in two opposite corners
of the array.
A two-pulse scenario for physical failure of the integrated power array is pro-
posed based upon the experimental and simulation results presented above [135].
Reversible operation is possible until the critical pulse regime, as long as the snap-
back mode is not realized. At first critical pulse a uniform turn-on with negative
differential resistance is followed by a non-uniform turn-off with a local filament.
This results in local melting or accelerated electromigration until the end of the
first pulse and formation of the array with local nonuniformly distributed damage.
After zap leakage changes only at a voltage close to the breakdown level. At the
a) b)
c) d)
Fig. 7.14 Simulated voltage (a) and current (b) waveforms for 24 mm NDeMOS array at different
HBM pulse levels in zero gate bias conditions and mapped simulated array current distributions
for 90 V HBM pulse before breakdown, (c) and after snapback (d)
HBM Robustness Vs. Vgs

8000
Square Array
7000
HBM Robustness (V)
6000 Rectangular Array

5000 Square Array-Reduced
4000 Metal
3000
G
2000
D 1000
S 0
–1 0 0.5 1 1.5 2 2.5
B
Gate Voltage (V)
Fig. 7.15 Example of the layout view for a square 24 mm array and simulated critical HBM pulse
level as a function of the gate bias
second critical pulse a local turn-on array in the pre-damaged spot is followed by an
immediate local burnout due to current channeling and creation of the high leakage
path.
70
60
50 Gate bias: 0V
Voltage [V]
40
3.2kV HBM: before failure
30
20 3.3kV HBM: during failure
10
0
0 250 500 750 1000 1250 1500
Time [ns]
Fig. 7.16 HBM voltage waveforms before and during failure measured for 24 mm 20 V NDeMOS
array
This conclusion correlates with the experimental data presented in Fig. 7.16,
where the HBM waveforms for the array before and during failure are presented.
Near the critical regime at 3.2 kV HBM, the array fully passes the test. However,
at the critical pulse level, the change in the waveforms indicates the failure at the
beginning of the pulse. Since the current distribution is uniform according to the
simulation analysis at the beginning of the pulse, it is logical to assume that the
array already receives local irreversible changes after the first pulse due to local
current filament, as described above.
7.2 Low-Voltage Power Circuit ESD Cases
7.2.1 LV Power Switching Blocks

Similar to signal path analog circuits, an ESD protection network for the low-voltage
power switching blocks can be realized with both with local snapback and rail-based
active clamp protection.
In the case of rail-based active clamp protection, the design has a relatively low
risk of missing the ESD protection specification. In this case, protection of the open
drain pins is significantly more predictable, since the active clamp provides a much
lower voltage waveform during ESD stress. However, this approach is not always
compatible with hot plug-in requirements and conditions with fast power supply
ramp voltages.
An example of a 5 V power train with active clamp protection is illustrated in
Fig. 7.17a. In this low-risk design, the 5 V active clamp cells protect both the control
and power pins. Respectively, each protected pin includes isolated Nwell and Pwell
diodes connected to the ESDP and ESDM buses. The diodes provide an ESD current
path that depends on the pin-to-pin combination.
The ESD protection network for the whole power train domain uses a com-
mon RC-timer module to provide a corresponding turn-off signal for the distributed
7.2 Low-Voltage Power Circuit ESD Cases 339
ESDPLUS
ENP VIN
High MP
Side
DRIVER
VDD SW
Dead
Time
Control
Low MN
Side
DRIVER
ENN
ESDMINUS PGND
a)
ENP VIN
High MP E5V
Side
E5V
DRIVER
CLK
VDD SW
Dead
E5V Time
TD1 E5V
Control
TD2 MN
E5V
E5V Low
E5V Side
E5V DRIVER
EN
PGND
b)
Fig. 7.17 Low-voltage power train protection using the rail-based active clamp approach (a) (not
all control pins of the network are shown) and the same circuit protection using 5 V snapback
NMOS clamps (b)
active clamp cells. The created network includes the RC-timer cell, upper and lower
ESD diodes, and active clamps based upon a 5 V NMOS device (Chapter 4).
The design requires metal ESDP and ESDM buses to provide a low resistive path.
In sufficiently large switch devices (such as PMOS MP and NMOS MN devices),
the switch pin SW does not require explicit ESD diodes, since they are already
represented in the ESD network by the body diodes of the MOS devices. However,
an appropriate metallization for the body diodes should be implemented.
Depending on the circuit specifications, the active clamp circuit network may be
inapplicable, for example, in operation regimes with a fast variable input voltage
(VIN). Therefore, an alternative, more aggressive design could use 5 V snapback
NMOS solutions at every protected pin (Fig. 7.17b).
Typical control pins are enable, clock, and delay signal pins. They can usually
easily tolerate an additional second-stage resistor of ∼1 k and thus typically do
not create ESD protection challenges.
A low-voltage avalanche diode with a breakdown voltage of ∼7 V can be advan-
tageous if the process provides an opportunity for its inclusion. The diode can be
effectively used to realize a two-stage protection clamp (Chapter 4) for the control
pins.
Depending on the SOA of the standard device and the circuit design, the external
VDD power supply for the driver can also tolerate snapback ESD protection.
Thus, a major challenge in the case of LV power train is the local snapback
protection of the switch pin SW. In general, the challenge may include both the
protection of the output NMOS array MN and the driver circuit connected to the
NM gate.
If the snapback NMOS clamp “E5V” (Fig. 7.17b) is based on the same device,
the ESD current path will depend on the conditions of the MN gate during ESD
pulse, the well tap diffusion to source space, as well as the size and layout of the MN
device. Since switch pin SW operates on the inductive load, fast transient voltages
are expected at the pin. In this case, a simple grounded gate NMOS solution that
utilizes the dV/dt is inapplicable due to gate coupling.
For a process with an insufficient ESD protection window due to the low SOA
of NMOS and PMOS devices, there are no universal recipes for local protection of
the SW pin and the final solution usually depends on both the process and product
specifics.
If the process provides low-voltage avalanche diodes, the reference voltage in
the drain–gate circuit can provide low dV/dt clamps with a triggering voltage in the
ESD protection window (Section 4.3.1). In large arrays, a self-protection capability
can be realized at the switch pin with appropriate layout and driver design.
The process-specific approach relies first of all on the pulsed SOA of the devices
and the availability of the avalanche breakdown voltage reference. A simple demand
of the process technology would be a requirement to increase the SOA limits
for a 5 V NMOS. However, this conflicts with the power-optimized integrated
components required to achieve superior switching characteristics and efficiency.
Recently, alternative solutions that involve active ESD clamping by the output
array itself have been demonstrated through sophisticated driver circuit modifica-
tion. However, these solutions are very product-specific.
7.2.2 Step-Down DC–DC Converters

ESD protection for complete low-voltage DC–DC converter products is similar in
principle to the protection of the power switching blocks discussed above. However,
the particular analog product specific implies additional constraints due to the size
of the output arrays, pin signals, absolute maximum limits of the product, as well
as the additional requirement for the ESD protection network to include other
circuit modules. Digital and analog blocks with different voltage domains,
EEPROM memory, and other devices can represent these modules.
An example of a step-down 6 MHz DC–DC converter with a maximum load
current of ∼0.5 A is presented in Fig. 7.18.
Fig. 7.18 A typical application circuit for fixed output voltage
The final, small form factor product is implemented with copper top metalliza-
tion in a micro-SMD package. The last also enables ESD cell placement under the
micro-SMD bumps. The under-the-pad design requirement often limits ESD clamp
implementation to using only one of the metal layers.
The peculiarity of this particular example is the internal CBOOT pin with an
internal bootstrap capacitor (Fig. 7.19) that eliminates the need for package-level
protection in this pin. As a result, the ESD protection network is rather simple. The
noisy input voltage (VIN) node is protected by an active clamp. The control pins
MODE, EN are protected by local clamps with a two-stage isolation resistor to the
internal nodes.
The most sophisticated switch node SW cannot tolerate a snapback device
solution due to the fast transient voltages and either relies on the self-protection
capability of the array or uses the same active clamp network as the VIN pin.
In this particular case, required ESD performance is achieved using the following
approach. The ground pin GND is at zero voltage and connected to the global ESD
Internal Boot Cap

CBOOT
VIN
NMOSI
SW
VIN
5V
Active
Clamp NMOS
PGND
Fig. 7.19 Active clamp ESD protection of the VIN pin with simplified representation of the power
side with an internal CBOOT pin and integrated boot capacitor as an internal node DeepNwell is
tied to CBOOT
bus. The VIN is protected by the 5 V-tolerant active clamp. The 5 V switch pin
SW is left without an explicit ESD clamp, relying on the self-protection capability
of the relatively large output arrays. The remaining control pins are protected by a
two-stage snapback NMOS clamp with a 1 k two-stage resistor.
Such a mixed clamp approach is implemented to provide substantial noise iso-
lation between the analog and power domains. It combines a snapback clamp at
the control pins and an active clamp at the power pins. This solution also intro-
duces diode isolation for the internal analog and power grounds, thus reducing the
coupling between the domains.
The body diodes of the output NMOS and PMOS arrays, for a 0.5 A output
current used in the synchronous output stage, provide a sufficiently high current
path to the active clamp rails (Fig. 7.19).
Another essential part of the ESD protection approach is the local active clamp
design. For a small-pin-count product, such an active clamp needs to be designed as
a local clamp rather than a distributed network.
Measures should be taken to avoid false snapback during a fast transient at the
VIN pin. Fast transients are produced due to the finite bond wire inductance. For
example, the clamp can be designed with a fully butted NMOS source/well to pre-
vent this problem. Additional attention should be paid to the metallization routing.
At the same time, rise time control should also be implemented on the trigger line
to avoid unexpected turn-on. An oversized NMOS device in the active clamp with
a total width of W = 10 mm is comparable to the size the array of W = 15 mm
connected to SW pin.
Alternative ESD protection of the switch pin by the local clamp and the related
specific is discussed in the next section.
VDD~
MP LW = 2nH RL 5.5V
PGATE +
-
W/L = 40000/0.5 IL~700mA
NGATE MN LW = 2nH LF =10μH

CP=10μF
W/L = 20000/0.5 CF =10μF
PGND
LW = 2nH
VDS
6.5μs
FSWITCH =1MHz
VDD
PGATE 0 5V
tD = 50ns
tR = 30ns
VDD
NGATE 0
Time
Time
Fig. 7.20 Simplified output circuit block with external components, and waveforms for the driver
signals at the gate of the output NMOS and PMOS arrays
7.2.3 Local Snapback Protection of LV Switch Pin

7.2.3.1 Case Study
A possible parasitic effect of the snapback NMOS grounded gate clamp connected
to the switch pin is the detection of the transient current due to gate coupling above
the threshold voltage. This effect usually precedes a false triggering of the clamp in
the transient latch-up mode.
Thus, at some critical level of the input voltage, ESD clamp snapback may result
in damage of the output circuit and switch devices. This effect is illustrated below.
For this particular product case, local burnout of the switching power NMOS has
been observed for circuit operation at elevated input voltage levels (VDD).
A simplified output circuit block diagram with external filter components LF CF ,
decoupling power supply capacitor CP , bond wire inductance LW , and driver signals
at the gate of the output NMOS and PMOS arrays is presented in Fig. 7.20.
a)
10–1 SWS=0.8um
80 GATE BIAS (V)
SWS (um) 10–2
70 10–3
CURRENT (A)
CURRENT (A)
60 10–4
50 10–5
40 2.8 10–6
30 10–7
20 10–8
10 10–9
0 10–10
14.9 15.0 15.1 15.2 15.3 15.4 0 2 4 6 8 10 12 14 16
b) c)
Fig. 7.21 Two-dimensional cross section for a finite element NMOS device used in mixed-mode
TCAD analysis of a power array (a) and initially calculated quasistatic isothermal ID –VDS charac-
teristics for different spacings between the P-well and source contacts (SWS) (b) and for different
gate bias VGS (c)
The original 5 V circuit demonstrated irreversible failure during switching after

the elevation of the input power supply voltage VDD up to 6–6.5 V.
Analysis of the circuit uses the following assumptions for the particular circuit.
The load resistor RL is ∼3.6 for VDD = 5.5 V, and a duty cycle of 50% limits
the output load current at ∼(5.5/2)/3.6 = 0.763 A. The waveforms provided by
the driver (not shown) apply a controlling “pgate” signal to the high-side PMOS
MP, causing it to be in the low current state when the “ngate” signal is also low.
Conversely, the “ngate” signal is provided at the high level only when the “pgate”
GATE VOLTAGE
10
8 7V
6
4
2
0
0 50 100 150 200 250
TIME (nS)
14
12
10
VDD (V)
8
6
4
2
0
0 50 100 150 200 250
TIME (nS)
5
4
3
VD (V)
2
1
0
–1
–2
–3
0 50 100 150 200 250
TIME (nS)
80
70
L-current (mA)
60
50 VL
40
30
20
10
0
–10
0 50 100 150 200 250
TIME (nS)
Fig. 7.22 Waveforms of the input gate voltages (first graph), VDD node (second graph), out-
put node of the inverter (third graph), and the current through the inductance (last graph)
for VDD = 7 V
signal is also high. The non-overlap time is about 50 ns (Fig. 7.20b). In the initial
conditions, both the output NMOS and PMOS devices are in off-state. Then, the
VDD node is set to some voltage level, followed by a preset of the output capacitor
to the same voltage. This way, both terminals of the inductor are set to the same
voltage as the output capacitor.
This major case study is related to the switching NMOS array layout effect. In
the original design, the NMOS structure layout had random P-well contacts. The
P-well diffusion to the source spacing followed the requirement for a 10 μm max-
imum space given by the process design rules. Such a design has been originally
chosen as a space-saving measure, since it provides a high-density array placement.
7.2.3.2 Mixed-Mode Simulation

The mixed-mode numerical analysis has confirmed the experimentally observed
failure scenario and helped validate the appropriate improvement measures. These
measures involved the implementation of a fully butted NMOS array design with an
additional redesign of array topology for appropriate current balancing.
The original 2D cross section for the finite element NMOS device used in the
mixed-mode TCAD analysis of the power array is presented in Fig. 7.21a. The
initially calculated quasistatic isothermal I–V characteristics for different spacing
between the P-well contact (SWS) and for different gate bias (Fig. 7.21b) demon-
strate consistent (Chapter 3) dependence of the critical voltage upon both the
internal base resistance of the parasitic NPN structure and the channel current that
provides the current in the drain multiplication region.
14
12
0
VDD (V)
8
6
4 VDD=3V
2
0
0 50 100 150 200 250 300 350
TIME (nS)
16
14
2
VDD (V)
0
8
6 VDD=9V
4
2
0
0 50 100 150 200 250 300 350
TIME (nS)
Fig. 7.23 Increase of the transient voltage noise with the input VDD voltage increase
As demonstrated by the simulation results, due to the inductance of the bond

wires during circuit operation, the overstress of the NMOS array can significantly
exceed the VDD level (Fig. 7.22). In this case, the amplitude of the transient voltage
noise increases with the increase of the input voltage level (Fig. 7.23). This effect
is assumed responsible for the early damage of the product between the operational
and absolute maximum voltage range.
Using the mixed-mode simulation setup, it has been demonstrated that reduction
of the p+ to source diffusion spacing (SWS) to 0.1–0.8 μm results in an increase
of the critical current density for snapback mode during switching by ∼3–10 times
GATE VOLTAGE
10
8 7V
6
4
2
0
0 50 100 150 200 250
TIME (nS)
14
12
10
VDD (V)
8
6
4
2
0
0 50 100 150 200 250
TIME (nS)
8
6
VD (V)
4 VIN
2
0
0 50 100 150 200 250
TIME (nS)
80
70
L-current (mA)
60
50 VL
40
30
20
10
0
–10
0 50 100 150 200 250
TIME (nS)
Fig. 7.24 Waveforms of the input gate voltages, VDD node, output node VL, and the current
through the inductance for the case of eliminated bond wire inductance components
7.3 ESD Protection of Integrated High-Voltage Regulators 347
(or even more, considering an isolation effect by the intermediate fingers, which act
similar to guard rings).
In addition, it is logical to assume that the 3D effects of the array’s topology are
important. The critical impact of the bond wire inductance components is demon-
strated by the absence of transient voltage noise when bond wire inductances are
removed (Fig. 7.24).
A similar conclusion can be reached using a compact model simulation with a
snapback ESD compact model for the NMOS component. This approach allows for
the inclusion of more driver components into the circuit.
7.3 ESD Protection of Integrated High-Voltage Regulators
7.3.1 Asynchronous Integrated Buck Regulator Case

7.3.1.1 Functionality and ESD Protection
A block diagram for the HV-integrated 100 V buck regulator is presented in
Fig. 7.25. This asynchronous voltage regulator is based on a 100 V switching
power NLDMOS array driven by internal circuit. The array supplies switching cur-
rent pulses through the integrated LC filter to create a corresponding lower output
voltage level (corresponding to the switching duty cycle).
The list of pins for package-level ESD protection includes the following control
pins that can be assigned to the I/O group. The low-voltage 8 V pin RCL is required
to connect resistor for current limit control. The pins’ absolute maximum voltage
is 8 V and the 7 V snapback NMOS clamp provides an appropriate ESD protection
solution.
The RTN pin is simply a power ground. The low-voltage VCC pin provides a
regulated output at 7 V from the series pass regulator. This pin can be alternatively
used to draw power from an external voltage source, which reduces the power dis-
sipation in the circuit. Since the VCC pin is specified up to 14 V, it requires a 20 V
snapback clamp.
The first high-voltage pin is the feedback pin. This pin receives a fraction of the
voltage from the output divider. For a 100 V circuit, this voltage can be as high
as 20 V. However, the voltage at feedback is already regulated, thus eliminating
any fast transient conditions. Therefore, conventional middle voltage, 20 V-tolerant
snapback protection presents an adequate solution.
For the high-voltage power group of the pins the requirements are the most
challenging. The operational input voltage range VIN is from 9.5 to 95 V. Due
to the high current during switching and the parasitic bond wire inductance, the
voltage waveform at the input pad of the die might be significantly different from
the package terminal, even if an external capacitor is used. The VIN pin requires
high-voltage protection using a 100 V snapback clamp. Due to fast transient volt-
age, the clamp should have a very small dV/dt effect and use an avalanche diode
reference. ESD protection of the pin can be achieved using snapback components
VIN
VIN=10-95V VCC Regulator VCC
BST
RON VIN
SHUTDOWN
VOUT
FB
SW
RCL
GND
Fig. 7.25 Application circuit and functional block diagram of the 100 V buck dc–dc voltage
regulator
with a holding voltage below the maximum input voltage (95 V). However, a more
reliable approach involves the application of the high holding voltage lateral PNP
ESD clamps (Chapter 4).
Protection of the switch pin SW is dependent on the capability of the array. In
a 1 A power array with an appropriate design of the circuit layout, the switch pin
can be self-protected by the power array. In this, most favorable, case, no ESD pro-
tection is required. An important consideration for the ESD current path is that the
power MOSFET still contributes the body diode into the ESD current path, and thus
appropriate metallization routing should be designed.
Finally, the boost-strap capacitor input (BST) is required to connect the output
inductor, re-circulating diode, and bootstrap capacitor. The boost pin is an overvolt-
age pin. The voltage tolerance of the pin exceeds the switch node by the additional
7 V required for the adequate power supply for the driver. This is required to provide
a typical boost functionality of on-state gate bias for the power MOSFET, including
the conditions of low output voltage. If no higher voltage devices are available, then
the only way to supply boost pin protection within the 100 V process limitations is
to protect the pin relative to the switch pin. The corresponding level of high-voltage
VIN
HV
VCC
MV
Bias BOOST
MV
Driver
DN SWITCH
LV
CL
PGND
LV Controller
CL
LV
ND AGND
Fig. 7.26 ESD protection network for a 100 V asynchronous switching buck regulator
isolation that enables such network implementation is usually requested of process

developers.
The final ESD protection diagram composed of low-voltage (7 V), middle-
voltage (20 V), and high-voltage (100 V) clamps is shown in Fig. 7.26.
7.3.1.2 Case Study

For reference, the pulsed SOA area for the power-optimized 1 mm 100 V NLDMOS
device is presented in Fig. 7.27a. In agreement with Chapter 4, the critical voltage of
the device in pulsed conditions is a strong function of the gate bias. These character-
istics can be measured for a small array using conventional TLP tools (Fig. 7.27b).
However, the total current that array can withstand under ESD stress is a more non-
linear function of array topology and other factors, including parametric process
yield.
0.05
GATE BIAS (V)

0.04 W=1mm
12
0.03
7
0.02 5
4
0.01 3
2
01
0.00
0 10 20 30 40 50 60 70 80 90 100110120130
PULSEDDRAIN-SOURCE VOLTAGE(V)
a) b)
Fig. 7.27 Drain-source TLP characteristics measured at different constant gate bias (a) and plot-
ted dependence of the gate bias upon the critical drain-source voltage (b) for a 100 V NLDMOS
device
At the same time, the self-protection ESD current level of the array is a function
of both the initial gate bias and the gate coupling level that can be tolerated by the
gate driver.
Therefore, it is not a surprise that the original array design may not provide
sufficient ESD robustness.
In particular, an example of machine model (MM) damage can be originally seen
in the corner of the power array (Fig. 7.28a). This problem has been eliminated by
redesign of the array. Two current paths were located by partition testing, revealing
a VIN-SW and a VIN-BST current path (Fig. 7.28b). The redesign involved array
layout optimization toward higher self-protection levels [136].
VIN -SWITCH
ator VCC VIN -BOOST
BST
VIN
SW VOUT
a) b)
Fig. 7.28 Original array showing damage after MM ESD stress (a) and ESD protection network
for a 100 V asynchronous switching buck regulator (b)
Another challenge that typically causes practical problems is the protection of

the internal VCC regulator (series 7 V regulator, Fig. 7.29).
The first current path formation leading to failure (Fig. 7.29a) involves HBM
pulse coupling through a voltage divider that is composed of the parasitic drain–
gate and gate–source capacitors on NLDMOS (Fig. 7.29b). In this case, the stacked
NLDMOS devices are driven into a low-impedance state prior to HV ESD clamp
activation.
In case of the second failure, the failure mechanisms have been initiated by the
insufficiently high voltage of PMOS breakdown or leakage, leading to gate bias that
turns on the NLDMOS prior to HV ESD clamp activation (Fig. 7.29a). This effect
has been confirmed by electroluminescence of the PMOS device region using the
PHEMOS technique (Fig. 7.29c).
The reliability improvement measures for these types of failure involve redesign
of the VCC regulator using stacked devices, process-level fixes of the high-voltage
PMOS, and implementation of more precise triggering characteristics for the high-
voltage clamp.
60V NLDMOS
burnout
Initial
breakdown
(PMOSHV)
7V drop 60V NLDMOS
On-state
a)
b) c)
Fig. 7.29 VIN to VCC HBM failure current path (a), FA photo for NLDMOS failure under all pin
HBM testing (b), and photoemission from PMOSHV at 110 V between VIN and GND pins (c)
7.3.2 Synchronous Regulators

7.3.2.1 HV Power Train Block
In a synchronous buck regulator, two high-voltage devices are used to realize the
transient current path during the switching cycle. Instead of a diode to provide the
current path through the LC filter, both the high-side power HV PLDMOS and the
low-side HV NLDMOS devices are used in a synchronized turn-on and turn-off
sequence with appropriate delay times (Section 7.1). The rest of the operation prin-
ciple is similar to an asynchronous regulator and presents itself as an integration of
the switching pulse with the input voltage amplitude VIN on the external LC filter
in order to obtain the output voltage VOUT, regulated by changing on the switching
duty cycle.
PVDD
E5V BOOST E5V

VDD
E5V
E5V High VIN
side
E20V
EM driver
PWM SW
Low
side
E5V
E5V
driver
AGND
pgnd
+
E20V
Fig. 7.30 A 24 V power train with ESD protection
A simplified circuit diagram and local clamp-based ESD protection network for
a 24 V synchronous power train is presented in Fig. 7.30. The output low-side
NLDMOS and high-side PLDMOS are driven by the 5 V low-side and high-side
drivers, respectively. The low-side driver is powered by the PVDD external voltage.
The supply voltage of the high-side driver is obtained from the voltage formed on
the BOOST pin. The two domains, power and analog, have separate power (PGND)
and analog (AGND) grounds that are separated for noise isolation by back-to-back
diode clamps. The analog domain is powered by the VDD power supply.
The analog circuit block is represented by the corresponding power supply VDD
pin, enable (EN) control pins, and pulse width modulation pin PWM.
Implementation of local ESD protection for this simplified power train circuit
schematic practically represents ESD solutions for a wide variety of DC–DC prod-
ucts (Fig. 7.30). ESD protection of the low-side drivers is ground referenced and
realized using a low-voltage snapback solution.
The high-side driver BOOST pin is protected relatively to the SW pin by treating
it as a high-side ground. For this, isolated low-voltage snapback clamps are used.
The current path is completed using HV local clamps between the PVIN and PGND
and the PVIN and AGND with an optional clamp at the switch pin (Fig. 7.30). The
HV clamp at the switch pin faces the same challenges and limitations as already
described for LV power train local snapback protection.
Since control pins usually do not require low-impedance current path, two-stage
protection is used for the input gate of MOS devices relying on current dissipation
in the internal circuit with an additional current limiting polyresistor of ∼1 k.
Alternatively, a second-stage 7 V avalanche breakdown diode can be used, which
provides the corresponding current of ∼1–10 mA to dissipate the second-stage cur-
rent, thus dropping the pin voltage from the voltage level produced by the snapback
NMOS clamp (∼10 V) down to a safe level.
The power ground PGND is decoupled from the analog ground AGND, allowing
it to be tolerant of relatively high voltage swings. This is accomplished by a back-
to-back diode clamp.
The high-voltage switch (SW) and BOOST pins can be protected by a different
high-voltage 24 V snapback clamp. Since high-voltage clamps are isolated from
the P-substrate in a BCD process, the power input PVIN could be protected not
only by a high-voltage clamp to the power ground PGND, but by a similar clamp
to the analog ground AGND as well. Thus, Nepi isolation between the grounds is
automatically provided.
In addition to power array protection, the most critical circuit block is the drivers
where a match should be achieved between the 5 V snapback NMOS device wave-
forms and the midsize internal NMOS devices connected through the low-resistance
current path.
Additionally to high-voltage node protection, the most challenging solution is
required for the high-side driver. Since the boost pin follows the transient volt-
age at the switch pin, the low-voltage isolated clamp needs to have a low dV/dt
effect. At the same time, power circuit requirements do not allow implementation of
second-stage protection. As a result, targeting of the ESD protection window creates
challenges.
In this case, the most typical failure is in the low-voltage NMOS and PMOS
components circled in the fragment of the driver circuit in Fig. 7.31. Depending on
the size of the driver components, two different current paths can be realized. In
a high driver, M5: 25 μm × 12 = 300 μm and M6: 25 μm × 100 = 2500 μm.
According to the simulation results with an ESD compact model (Fig. 7.32), the
PMOS has a low gate bias potential and the current is directed into the power array
gates. In the case of the low-side driver: M5: 50 μm × 20 = 1000 μm and M6:
50 μm × 261 = 13,050 μm, and the current is directed to the M5 and M6, resulting
in damage of these devices.
7.3.2.2 Synchronous Buck Regulator

The same principles used for power train ESD design are applied to the more
complicated analog circuit of the synchronous buck switcher.
An example of a synchronous switcher that can accept an input voltage range
from 3 to 20 V and provide up to 4 A of output current is presented in Fig. 7.33.
In this particular simple switcher design, core logic is not in use and thus ESD
protection of the low-voltage (2.5 V) pins is not required.
The circuit can be implemented in the non-copper top and copper top versions. In
the case of copper top variety, all circuitry can be put under copper pads, including
the ESD.
The major circuit pins are similar to the power train example discussed above.
The internal chip has separate analog and power grounds that bond to the same
package pin to reduce the bond wire impact. The analog ground AGND supports
the global ESD bus across the analog block, while the power ground is decou-
pled using either the back-to-back or the stacked back-to-back NMOS clamp. The
switch pin can be optionally protected by a precision high-voltage clamp. The same
clamp is used in the 20 V-tolerant input pin. CBOOT pin protection is accomplished
VDD
LGATE
M5
M6
PGND
a)
CBOOT
HGATE
M6
M5
SW
b)
Fig. 7.31 Overstressed devices in 24 V high- (a) and low-side (b) drivers
relative to switch pin using a low-voltage isolated snapback NMOS. The same pro-
tection is implemented for the VDD pin that supports the external connection to the
decoupling capacitor.
The control high-voltage power good and enable pins are also protected by high-
voltage clamps.
The feedback pin requires an elevated voltage tolerance of up to 10 V and thus
uses a corresponding 10 V clamp.
The decoupling between the power and analog grounds is required since the
current through the power block is significant.
a)
b)
Fig. 7.32 Comparison of the 2 kV HBM simulation results for low-side (a) and high-side
drivers (b)
The source of the low-side NLDMOS is connected to the PGND. It is, how-
ever, required that the PGND signal be able to swing up to ±5 V relative to the
analog ground AGND. Respectively, the ESD test program includes an ESD zap
combination of PGND to AGND. The possible current path is formed through the
reverse-path diode or parasitic diode of the 5 V snapback NMOS clamp. In the case
VIN
VIN
Enable EN CBOOT
PGood PGood VOUT

SW
COMP
VDD
FB
GND
Fig. 7.33 Application for a simple switcher
of the opposite AGND to PGND zap combination, the reverse-path diode of the
isolated NMOS snapback clamp conducts the current.
Protection of the input voltage pin is illustrated in Fig. 7.34. In the case of the
input to VBOOT (similar to VDD) zap combination, the current path relies on the
PLDMOS body diode. In the case of AVIN to analog ground AGND zap, the high-
voltage snapback clamp is engaged (Fig. 7.34a).
Additional protection measures are required for the internal circuit, so 7 V
avalanche diode clamps are extensively used across the internal circuit to clamp
the gate–source voltages of the internal components (Fig. 7.34b), thus reducing the
risk of a significant reduction of the pulsed critical voltage and gate oxide damage.
Protection of the power side is illustrated in Fig. 7.35.
The level shift circuitry is the most sensitive, as it is exposed to the full CBOOT
voltage to AGND. To protect the internal voltage regulator, a separation resistor of
∼10 k is used (Fig. 7.36) or even an additional internal voltage clamp.
Another major concern is about the two high-voltage pins power good (PGOOD)
and ENABLE. The challenge involves reliable protection of small, internal high-
voltage devices interacting with the pin.
In this case, an adequate ESD protection window is needed to provide survivabil-
ity of protected devices with a total width in the range of 100–300 μm. If the process
technology provides an option of higher voltage components not optimized for the
switching operation, such components will be the best choice. Another alternative
includes circuit design based on stacked devices, in order to guarantee that the total
voltage limited by the ESD protection clamp will not exceed the pulse SOA limits
in the rather unpredictable conditions of gate coupling.
If none of this is possible, a separation polyresistor is the best choice (Fig. 7.37).
7.4 Controllers 357
200Ω
AGND VIN
PLDMOS
gnd
VDD
+ +
E20V E5V
AGND
AGND
a)
PLDMOS PLDMOS
PLDMOS
DZ
PLDMOS
DZ
DZ
b)
Fig. 7.34 Input voltage pin protection: AVIN to VDD and AVIN to AGND
7.4 Controllers
Controllers are integrated components that provide driver signals to the discrete
switching components in hybrid dc–dc voltage regulators. Different types of dc–dc
converters can be realized with the appropriate selection of the controller type, both
the asynchronous and synchronous, buck and boost varieties.
In principle, ESD protection of the controller pins is rather similar to that of
the integrated voltage regulator. The difference consists in the necessity of switch
pin protection for synchronous converters. The driver (DR) pin is connected to an
external discrete switch device. Often, output devices in the controller do not have
sufficient size to realize the self-protection capability. Therefore, an explicit power
clamp at the switch pin may be required. Under such conditions, it is important to
know what is the gate coupling of the array in ESD stress conditions and what is the
BOOST
PVIN
Internal Boot
Diode
agnd
E5VI
SW
E20V
VBOOP agnd
E5V E5VI E20V
E5VI PGND
E5V
AGND
Fig. 7.35 ESD protection for the power-side SW with optional HV clamp
CBOOT
PLD MOS
PLD MOS
NL DMOS NL DMOS
EN
NMOS NMOS
NMOS
AGND
Fig. 7.36 Level shifter circuitry
7.4 Controllers 359
20k
PGOOD
50k Internal
D
E20V G
B
S
AGND
Fig. 7.37 Power good (PGOOD) high-voltage pin and ENABLE pin protection
pulsed SOA of the device. Several aspects of controller protection are highlighted
below.
7.4.1 Asynchronous Buck-Boost (SEPIC) Controller

An example of the application and the functional diagram of a wide supply voltage
range (2.97–48 V) controller are presented in Fig. 7.38 [126]. Typical applications of
this controller include various distributed power systems, notebooks, PDAs, digital
cameras and other portable applications, offline power supplies, and set-top boxes.
The controller drives the low-side discrete N-FET in switching regulators. It is
suitable for use in topologies requiring a low-side FET, such as a boost, flyback,
SEPIC. The controller can be operated at extremely high switching frequencies (up
to 1 MHz) in order to reduce the overall solution size. The switching frequency is
adjusted by using a single external resistor or by synchronizing it to an external
clock.
Similarly, the current limit can be programmed with a single external resistor.
Other capabilities of the controller include standard protection features such as
thermal shutdown, short-circuit protection, and overvoltage protection. Another typ-
ical feature is a power-saving shutdown mode that reduces the total supply current
to 5 μA.
The controller uses a fixed frequency with pulse width modulation (PWM) and
current mode control architecture. In a typical application circuit, the peak current
through the external MOSFET is sensed through an external sense resistor. The volt-
age across this resistor is fed into the ISEN pin. This voltage is then level shifted and
fed into the positive input of the PWM comparator. The output voltage is also sensed
through an external feedback resistor divider network and then fed into the error
amplifier (EA) negative input (feedback pin, FB). The output of the error amplifier
(COMP pin) is added to the slope compensation ramp and fed into the negative input
of the PWM comparator. At the start of any switching cycle, the oscillator sets the
RS latch using the SET/blank-out and switch logic blocks. This forces a high signal
on the DR pin (gate of the external MOSFET) and the external MOSFET turns on.
VIN
VIN
VOUT
a)
VIN
b)
Fig. 7.38 Application circuit (a) and functional block diagram (b) of the low-side NMOSFET
controller
When the voltage on the positive input of the PWM comparator exceeds the voltage
on the negative input, the RS latch is reset and the external MOSFET turns off.
The voltage sensed across the sense resistor generally contains spurious noise
spikes. These spikes can force the PWM comparator to reset the RS latch prema-
turely. To prevent these spikes from resetting the latch, a blank-out circuit inside
the IC prevents the PWM comparator from resetting the latch for a short duration
after the latch is set. This duration, called the blank-out time, is typically 250 ns
and is specified as tmin (on) in the electrical characteristics section. Under extremely
light load or no load conditions, the energy delivered to the output capacitor when
the external MOSFET is on during the blank-out time is more than the energy
7.4 Controllers 361
delivered to the load. An overvoltage comparator inside the circuit prevents the out-
put voltage from rising under these conditions by sensing the feedback (FB pin)
voltage and resetting the RS latch. The latch remains in a reset state until the out-
put voltage decays to the nominal value. Thus, the operating frequency decreases at
light loads, resulting in excellent efficiency.
The specific pins for ESD protection of converters are somewhat similar to the
integrated switching regulators. VIN is usually decoupled from the power FET
devices and requires a high-voltage power clamp.
The control pins can in most cases tolerate an additional separation resistor in
the range of 1–5 k. Thus, the protection can be based on a two-stage approach
that takes into account a certain current that can be dissipated by the internal circuit
components connected to the control pin.
Similar to the integrated converters, the feedback pin usually experiences a rather
slow transient voltage and can be easily protected even by the dV/dt-triggered clamp.
The overall ESD protection network for the controllers is shown in Fig. 7.39.
The ESD protection network is composed of one 50 V high-voltage power clamp
at the VIN pin and a low-voltage clamp referred to the ground of the corresponding
(power or analog) domain.
E7V VIN E50V
E7V E7V
PGND
E7V E7V
E7V
E7V
AGND
Fig. 7.39 ESD protection network for an asynchronous controller
The low-voltage control pins of the analog domain include a current sensing pin
to detect the voltage generated across the external resistor, under voltage lockout
pin (UVLO), the compensation pin (COMP), the feedback pin (FB) connected to
the inverting input of the error amplifier, and the frequency adjust, synchronization,
and shutdown pins. All of these pins have local protection with a separation resistor
relative to the analog ground node.
The low-voltage pins in the power domain include the driver output (DR) and
driver supply voltage pins used for decoupling the capacitor connection (VCC).
Thus, the overall ESD protection solution is based on a classical local clamp
approach (Chapter 5). The most challenging pin is the high-voltage pin VIN. In
spite of a much lower transient voltage in the controller (in comparison with the
integrated voltage regulator), this pin still requires precise targeting of the clamp’s
turn-on voltage if the snapback solution is used. The targeting of precise ESD clamp
triggering is challenging in case of insufficient marginal SOA of the high-voltage

devices supported by the process. Under these circumstances, specifically device
level, ESD design is critical for the success of the protection in the final analog
circuit product.
7.4.2 Synchronous Buck Controller

An example of the application diagram for a synchronous buck controller is shown
in Fig. 7.40. This controller is intended for step-down regulator applications from a
high voltage or widely varying input supply. The control method is based on current
mode control utilizing an emulated current ramp. Current mode control provides
inherent line feed-forward, cycle-by-cycle current limitation, and ease of loop com-
pensation. The use of an emulated control ramp reduces noise sensitivity of the pulse
width modulation circuit, allowing reliable control of the very small duty cycles
necessary in high input voltage applications.
Fig. 7.40 Application diagram for 100 V synchronous controller
The operating frequency is programmable from 50 kHz to 1 MHz. The con-

troller drives external high-side and low-side NMOS power switches with adaptive
dead-time control. A user-selectable diode emulation mode enables discontinuous
mode operation for improved efficiency at light load conditions. A low quiescent
current shutdown disables the controller and consumes less than 10 μA of the total
input current. Additional features include a high-voltage-bias regulator, automatic
switchover to external bias for improved efficiency, thermal shutdown, frequency
synchronization, cycle-by-cycle current limiting, and adjustable line under-voltage
lockout.
7.4 Controllers 363
The ESD protection network for the controller can be constructed based on
the absolute maximum voltage limits for the circuit pins and the internal circuit
blocks, using local 7, 16, and 100 V clamps for ESD protection. The network, with
corresponding clamps E7V, E16V, and E100V, is shown in Fig. 7.41.
E16V
E16V
E16V
E16V
E100V
E100V
E16V
E7V
E100V
E7V
E100V
AGND
E7V E7V E7V E7V E7V

E16V E100V
AGND PGND
Fig. 7.41 ESD protection network for a 100 V synchronous controller composed of 7, 16, and
100 V local clamps
Even though both power and analog grounds are connected to the same PCB node
in the application, decoupling of the analog and power domains is required inside
the chip due to bond wire resistance and inductance. Correspondingly, the domains
are separated by the back-to-back ESD diodes. The remaining network structure is
based on local clamp protection, taking into account that each clamp incorporates
the reverse diodes.
There are three reference-ground-like levels by which local ESD protection is
organized. In the power domain, the VIN pin is functionally protected by the 100 V
power clamp relative to the power ground. This pin also connects the chip voltage
supply, which also connects the input voltage monitor and the input of the VCC
regulator. Therefore, all the devices connected to the pin should be evaluated for
alternative current paths.
Similarly, 100 V local clamp protection is applied to the VOUT and, optionally,
to SWITCH pin.
Other pins with 100 V absolute maximum ratings are the current sense (CS) and
enable (EN) pins, which use local 100 V protection relative to the analog ground.
These pins can usually tolerate a second-stage resistor.
The power-side 16 V pins include an optional input for the external power supply
VCCX, a VCC pin for local decoupling by the external capacitor, low- and high-side
driver outputs LO and HO, and a boost pin (HB). The low-side driver output LO is
protected using a power-ground-referenced 16 V clamp. The high-side driver and
boost are protected relative to the switch, treating this fast transient pin as a local
ground.
The other analog domain control pins are protected locally, relative to the analog
ground (Fig. 7.41).
One of the peculiarities of some synchronous controller circuit designs is the
requirement that a high-side switch pin must go below the power ground. In this
case, it is important to avoid parasitic injection of the carriers into the substrate. An
example of the solution is shown in Fig. 7.42. It provides for SW pin biasing below
ground at a low current and voltage of up to 3 V. This dual-direction protection can
be achieved by stacking high-voltage clamps that incorporate an additional base–
emitter diode provided by the NPN BJT.
CBOOT
E24V
SW
VCC
E24V
PGND
Fig. 7.42 Simplified block diagram for a 50 V controller circuit with ESD protection components
7.5 Light Management Units and LED Drivers
7.5.1 Analog LED Technology

Today, backlight light emission diode (LED) technology is replacing other light
sources for liquid crystal displays (LCD). RGB (red, green, and blue) LED improves
the color saturation on TVs and allows dynamic dimming. LED backlight tech-
nology has attracted a great amount of attention due to its wide color range, easy
7.5 Light Management Units and LED Drivers 365
application of dynamic driving, and scalability to larger sizes. Over time, cost-
efficiency issues are being resolved by successful commercialization, for example,
in LCD TV applications.
There are numerous applications for LED backlighting. For example, automotive
cockpit displays, rear seat entertainment systems, computer displays; mid-size LCD
TVs, digital frames; portable ultrasound, touch pads, industrial operator interfaces,
security monitoring and video.
There are several ways to produce backlighting in systems that provide different
spaces for LED drivers. Direct lit approach requires a small number of high-power
LED modules with a current of up to 5 A. This approach can be realized using a
large number of LED modules with small currents of 50 mA. Such an architecture
provides advantages in thermal and optical aspects, but makes control more com-
plex. The side-emitting approach requires a high brightness and a current of ∼0.5 A
when operating at some medium number of LED.
From an ESD point of view, the major concern created by the LED driver speci-
fications is the current level of the output current sink components. In the case of the
50 mA output, each high-voltage current sink will require individual ESD protec-
tion, while in the case of single-point direct-lit drivers, the current sink power array
might be already self-protecting.
The real light management units may be rather complex systems that include
several modules (Fig. 7.43). In addition to the dedicated LED light source, such a
system can include a multichannel color sensor, a temperature sensor, sophisticated
digital modules for backlight control, and integrated power supply modules.
Color Sensor
LCD Control
Mixing Light Guide and LCD Panel
Temp Sensor Light Source
Temper.
PWM Power Supply
Optical Light
Source
Display SYNC
Dynamic Controller
Controller
Backlighting Control
Brightness Color
Fig. 7.43 Simplified example of the closed-loop basic architecture for LED backlighting
7.5.2 LED Drivers

An example of the application and functional block diagram for a constant current
boost LED driver with internal compensation is presented in Fig. 7.44.
VIN
a)
VIN
b)
Fig. 7.44 Application (a) and functional (b) diagram of a LED driver circuit
The driver is intended for applications in the LED backlight current source,
backlight OLED and HB LED drivers, handheld devices, LED flash drivers, and
automotive uses.
The constant current LED driver is a monolithic, high-frequency, PWM DC/DC
converter with a design that requires minimum external components. It can deliver
typical 3 A peak currents with an internal 170 m NMOS switch. The switching fre-
quency is internally set to either 525 kHz or 1.60 MHz, allowing the use of extremely
small surface-mount inductors and chip capacitors. Even though the operating fre-
quency is high, efficiencies of up to 88% are easy to achieve. External shutdown is
included, featuring an ultra-low standby current of 80 nA. The driver utilizes current
mode control and internal compensation to provide high performance over a wide
range of operating conditions. Additional features include dimming, cycle-by-cycle
current limiting, and thermal shutdown.
The device operates similar to a voltage-regulated boost converter, except that it
regulates the output current through the LEDs (Fig. 7.44b). The magnitude of the
current is set with a series resistor.
The ESD protection network for this small pin count circuit includes low-voltage
VIN protection relative to the power ground and protection of the control and feed-
back pins relative to the analog ground. Due to a very high current of the output
NLDMOS array (over 2 A) the output (SW) pin does not require explicit protection.
The ESD protection network described above is shown in Fig. 7.45.
SW
E7V FB
E7V VIN
DIM E7V
AGND PGND
AGND PGND
Fig. 7.45 ESD protection network for LED driver
7.5.3 Light Management Units

Unlike powerful LED drivers, the backlighting LMUs (light management units)
are usually designed to control relatively low current levels through multiple LED
strings. Respectively, the most specific pins for ESD protection are the high-voltage
current sink pins. The LMU can be designed for both portable and integrated
systems. In portable electronics, the most typical requirement for LED drivers is
operation from a standard battery power supply of ∼3–5.5 V.
To provide high performance in backlighting, one of the most important param-
eters is the output voltage of the LED drivers. A higher voltage level provides a
corresponding advantage in the stacking of a larger number of LEDs. This allows
reduction of the number of wires in the portable system and simultaneously provides
an equalized intensity of the light emission across the LED string, simply due to the
same current level. Which in turn makes it possible to use the less-expensive LED
diodes with unmatched electrical characteristics.
The LMU uses an integrated current mode pulse width modulation (PWM) boost
converter, similar to the LED driver in portable applications. The converter becomes
functional at an input voltage from 2.7 to 5.5 V and provides a regulated, high-
voltage output to the LED string.
In addition to the voltage boost, the light management units (LMUs) can incor-
porate many features provided by the corresponding functional blocks, for example,
programming via an I2 C-compatible interface.
A simplified circuit diagram for the LMU system is presented in Fig. 7.46. The
driver includes an integrated boost voltage regulator to increase the battery input
voltage level to ∼40 V, in order to produce power for a string of LED. Simplified
internal circuit diagrams are shown in Fig. 7.47.
The module uses an external reference voltage source that is encoded by an
analog–digital converter (ADC). In each channel, the LED intensity is controlled
by the current sink circuits that use the pulse width modulation principle.
The LED bias supply circuit adaptively regulates the supply voltage of the LED
strings to maximize overall efficiency. The biasing circuit provides regulated current
3–5V
8-24V
FEEDBACK SWITCH LEDOUT

20V White LED Driver
REF
REF PWM NLDM20 LED NLDM20
+
- W=36mm W=1.3mm
Driver
REF
PGND Boost Regulator Iref
+
-
NMOS
W=1.3mm
Boost Switcher
VDD1 RND
4ohm
LEDGND
VIN
LDO
VDDA
VDD2
RLED
VIN RGB GLED
Digital Driver BLED
Analog Digital
Block Control Inter-
Circuit face
VDDIO
AGND RGBGND
SCL
SDA
Fig. 7.46 Application circuits and functional block diagram for LMU
to the current sink circuit. The maximum current per output is set via an exter-
nal low-power resistor connected to the control pins. An I2 C-compatible interface
allows for independent adjustment of the LED current in either output.
The core of the circuit is a PWM current mode boost converter. At the start of
each switching cycle, the internal oscillator sets the PWM converter. The converter
turns the NMOS switch on, allowing current to ramp up in the inductor while the
output capacitor supplies power to the LEDs. The error signal at the output of the
error amplifier is compared to the sensed inductor current. When the sensed inductor
current equals the error signal or when the maximum duty cycle is reached, the
NMOS switch turns off, causing the external Schottky diode to pick up the inductor
current. This allows the inductor current to ramp down, causing its stored energy to
charge the output capacitor and supply power to the load. At the end of the clock
period, the PWM controller is set again and the process repeats itself. When biasing
LED strings, the circuit maximizes efficiency by adaptively regulating the output
voltage.
The main ESD-specific requirement of LMUs is the protection of current sink
pins (Fig. 7.47). For portable applications, the current sink is usually designed to
supply a relatively low current of up to 30 mA to the LED stack. Thus, an open-drain
current sink device requires special measures for ESD protection.
VIN
+
–
VREF ADC PWM

+
–
+
DRIVER –
PWM
Control
Clock
Dimming
a)
VREF
LED 1
Force ON
PWM
LED 2
ENABLE
b)
Fig. 7.47 Simplified circuit diagrams for LED backlighting block (a) and current sink circuits (b)
7.5.3.1 Switch Pin Protection

If the LMU is designed to supply current to multiple LED strings, the boost switch
has a relatively large NLDMOS device. If the size of the output array is insufficient
for self-protection, then local switch-pin protection is used. In the case of a very lim-
ited ESD protection window, two high-voltage clamps can provide an appropriate
voltage limiting relative to both the power and analog grounds (Fig. 7.48).
VDDA
MP SWITCH
E5V MNH
E5V
E24V E24V
MN
ESDM
PGND
Fig. 7.48 Local ESD protection of the boost switch circuit
At the same time, since the array can transmit the ESD signal due to a high gate–
drain voltage/current, additional 5 V snapback or avalanche diode protection can be
used to clamp the switch gate of the NLDMOS and thus protect the driver circuit
components. Respectively, the driver’s output NMOS and PMOS transistors have
butted well ties to maximize the critical avalanche current in the whole gate bias
range (Fig. 7.48).
The switch pin experiences a very fast voltage transient when the switch is turned
off. Non-interference of the ESD device with normal operation is crucial for the final
analog product. The high-voltage clamp should be carefully designed or selected
with features that eliminate the possible dV/dt effect and guarantee the absence of
false triggering.
7.5.3.2 Feedback Pin Protection

Another fast transient circuit is the boost feedback circuit. It consists of a current
sink and a feedback resistor divider. The sink current is only approximately 5 mA.
This high-voltage circuit block requires a corresponding high-voltage local clamp
(Fig. 7.49).
FEEDBACK
VDDA
MNH
E5V
GATE
CLAMP
E5V E24V
MN
ESDM
LEDGND
Fig. 7.49 Simplified circuit diagram for boost feedback circuit with ESD protection
7.5.3.3 LED Driver Protection

The LED driver (Fig. 7.50) is the most specific pin for LMUs. Normally, LED driver
protection would require just a high-voltage clamp. However, in some cases, the
power-optimizing process may provide no other alternative but to implement a cas-
cade NLDMOS driver (Fig. 7.50). This is one of the direct examples of ESD and
analog circuit co-design. Essentially, the only reason for stacked driver design is
to enable protection of the low side NLDMOS driver under ESD stress. This cir-
cuit solution simply increases the absolute maximum voltage of the circuit, thus
increasing the ESD protection window.
LEDOUT
VDDA
MNH
E5V
GATE
CLAMP
E24V
REF MNH
ESDM
LEDGND
Fig. 7.50 Simplified circuit diagram for LED driver with ESD protection
Additional improvement toward robust protection includes the gate clamp

described below and internal driver protection.
7.5.3.4 Gate Clamp

Measures for current sink and switch protection can be insufficient due to excessive
coupling of the array. To resolve this issue, the capacitive coupled gate clamp is pre-
sented in Fig. 7.51; it is based on low-voltage NMOS devices MN1-MN3. During
ESD stress of the boost pin, the capacitive coupling provides on-state conditions for
the NMOS MN3 connected to the PGATE node, thus creating short-circuit condi-
tions for the power NLDMOS (not shown on Fig. 7.51) in the boost circuit. As a
result, the NLDMOS remains in the off-state during the stress, providing the highest
possible drain–source triggering voltage. In this case the ESD current path is pro-
vided by the additional explicit high-voltage power clamp (not shown on Fig. 7.51).
The CMOS diode of NMOS MN2 is used in the circuit to ensure that the excessive
coupling through the capacitor will not damage the NMOS gate.
VIN
PGATE
CMIM
MN3
VDDA
MN1 MN2
PGND
Fig. 7.51 Gate clamp
During normal powering sequence, in spite of the voltage change at the boost
pin, the coupling capacitor remains fully charged, while the VDDA signal keeps the
NMOS MN1 connected to the GATE node in the off-state.
7.5.3.5 RGB Driver

The RGB LED driver design includes cascaded NLDMOS (MNH) and low side
low voltage NMOS (MN) connected to the driver (Fig. 7.52). In this case, a low-
voltage NMOS is used due to the poor matching of the NLDMOS devices. In
addition to 24 V protection with snapback clamp (E24V) the 5 V snapback clamp
(E5V) is employed to ensure that the low-side low-voltage NMOS (MN) is protected
(Fig. 7.52).
A resistor is needed for the cascaded NLDMOS (MNH) to ensure the functional-
ity of the RGB driver (Fig. 7.52). This gate biasing is common for all RGB outputs.
A 5 V clamp added to the NLDMOS (MNH) gate, as well as a capacitive cou-
pled gate clamp. The capacitor is a capacitive connection from the output, with
metal-to-metal capacitance if required, of approximately 150 fF.
LEDOUT
VDDA
MNH
E5V
GATE
CLAMP
E5V E24V
MN
ESDM
RGBGND
Fig. 7.52 Simplified circuit diagram for the RGB driver with ESD protection
7.5.3.6 Control Pins

The approach for control pin protection in LMU is similar to that of other analog
circuits. Each specific pin design requires an analysis of the connected control cir-
cuit. In most cases, the control pins include a CMOS input buffer to provide the
input function and an open-drain NMOS device to provide the output function.
For the output driver, usually even a relatively small (50 ) well saturation
resistor between the pin and the NMOS drain provides a significant advantage
(Fig. 7.53).
The high-voltage control pins may have an open-drain current sink as an output
and an NLDMOS protected digital input.
VIN 24V
E5VI
-5V
SENSE
VDD
GAIN
5V VCC regula
Regulator
Regulator tor E20V
100 5K
VCCP(19V)
VDDP VCCN(5V)
SW
Active
E5VSNM Clamp
PGND
Bias
VDDA
OFF ADJ LDO
100 100 ENABLE
100
VDDA
NLDMOS
E5V E5V
E5V E20V
AGND
Fig. 7.53 Protection of the control pins

7.5.3.7 Current Sink Protection

The output of the main current sink is tied to the drain of a 2 mm NLDMOS device
with ∼4 source degeneration resistor RND . Additional attention is required for
isolation of the input devices so that they do not share composite. The shared com-
posite of the two NMOS devices connected to two different pins can be problematic
with respect to ESD due to paretic current path.
In the case of micro-SMD, the custom modifications of ESD cells include the
removal of pad plates and the implementation of appropriate connections to the
bump metal to avoid the current crowding effect.
Similar to boost regulators and controllers, protection of the boost pin can be
done relative to the switch pin using an isolated low-voltage clamp. An additional
alternative current path can be introduced by the overvoltage clamp between the
boot and the power ground pin. Similar to the switch driver, additional measures
have been used to increase the pulsed SOA of a 5 V NMOS device by butting the
P-well ties to the source.
An alternative option is to implement the self-protecting capability of the LV
NMOS by a drain extension of 1–3 μm. I/O latch-up prevention rules also apply,
including double guarding.
7.6 A Few More Case Studies

In this last section of the chapter three extra case studies are discussed to emphasize
one more time a non-trivial nature of the power management circuit design as well as
to demonstrate the value of mixed-mode simulation approach. The cases cover quite
unique events of transient latch-up and device failure under charged device model
(CDM) pulse. All of them occur due to an unexpected current path in power analog
circuit. The specifics of these cases are substantial dependence of the phenomena
on IC layout and even on PCB design.
7.6.1 Power Array–ESD Clamp Interaction
In general the challenges for ESD protection of the power circuits are not limited by
the ESD protection window targeting only. In principle the substrate current, gener-
ated by any internal circuit components, can impact critical triggering regime of the
snapback ESD device both in ESD pulse domain and under other conditions. The
scenario of the “accidental” ESD device turn-on in non-ESD time domain transient
conditions may result in circuit or even printed circuit board components failure dur-
ing functional testing. At the same time the event can be sensitive to the circuit board
design due to significant change of the pulsed voltage waveforms caused by induc-
tive and capacitive PCB load. In principle in case of ESD pulse regime a premature
turn-on of the ESD device in snapback mode due to additional injection from the
internal circuit components can be reversible and even favorable for the ESD circuit
7.6 A Few More Case Studies 375
operation itself. However, in actual operation with power supply voltage applied the
turn-on of the ESD device may cause damage due to excessive power dissipation at
high current.
A simplified circuit diagram with the switch pin (SW) is presented in Fig. 7.54a.
The original circuit met all specified ESD requirements and was designed with the
absolute maximum drain voltage of ∼45 V for both switch (SW) and other separate
control pins. The maximum operating voltage is specified at the level of ∼40 V. The
ESD clamp is based upon NDeMOS-SCR and designed for the triggering voltage of
∼50 V targeting the lower ESD protection window limit of ∼45 V. In the particular
design identical ESD protection clamps have been used for protection of both the
switch and the overvoltage protection pins. At the switch pin the topological separa-
tion of the clamp from the large power array was ∼200 μm (Fig. 7.54b). The space
is filled with the P-guard ring diffusion.
250 um separation SCR clamp
SW
Gate SCR
clamp
Power
50mm
Array DMOS Internal
Array Circuit
GND
a) b)
Fig. 7.54 Simplified application diagram for the part of the LED driver circuit (a) and layout of
the power array and ESD clamp (b)
The observed application problem was related to accidental turn-on of the clamp
in certain transient test regimes close to the absolute maximum limits.
The pulsed measurements demonstrate a significant change in the triggering volt-
age observed at the pin in comparison with the characteristics of the stand-alone
clamp.
At switch pin to ground combination (SW–GND) TLP test results were differ-
ent depending on the input pin voltage (VIN). With floating input pin the triggering
voltage was measured as low as ∼40 V (Fig. 7.55). Prior the triggering into snap-
back a significant difference of pin pulsed I–V characteristics was observed. Much
higher ∼3 A current level preceded the snapback mode (Fig. 7.55). This current
is representative of the on-state conditions of the array connected in parallel to
the pin.
Different VIN dc bias values resulted in different conditions at the power array
gate in TLP transient regime. With VIN pin bias at high level the array gate potential
is low and the clamp turn-on closely matches the stand-alone conditions. However,
even in this case it is hard to conclude to what extent the internal avalanche–
injection conductivity modulation mechanism of the clamp itself is involved. With
Pulsed Voltage (V)

0 20 40
1.E+01
Pulsed Current (A)

1.E+00
1.E-01
1.E-02
10ns TLP: Clamp
1.E-03 10ns TLP: SW-GND (VIN float)
10ns TLP: SW-GND (VIN 4.7V)
1.E-04
Fig. 7.55 Comparison of the TLP characteristics of the stand-alone ESD clamp with the
characteristics at SW pin at different VIN voltage levels
bias applied to the input pin TLP characteristics of the SW–GND pin combina-
tion show the triggering voltage ∼48 V much closer to the stand-alone clamp with
low pre-trigger current. However, even in this case switching DMOS draws over
∼150 mA current prior to the snapback mode (Fig. 7.55) indicating additional
current path.
At floating VIN node the array gate to drain coupling of the switch transistor is
high. This results in significant drain current and corresponding substrate current.
In spite of the large guard-ring protected distance between the power array and ESD
device the 10 V triggering voltage reduction is observed (Fig. 7.55).
Depending on the process, array design, and electrical regime the array substrate
current can reach a level of ∼30% of the drain current. The current path is real-
ized between the array drain and any bulk p+ -diffusion in the vicinity of the arrays
including P-guard ring of the ESD clamp. At the same time only a few milliamps of
the current injected into SCR clamp is sufficient to trigger snapback mode.
Independent from triggering scenario the holding voltage at each pin is observed
at the level similar to the stand-alone clamp. The 1–2 higher resistance in this
case can be attributed to extra ground bus circuit resistance. This fact confirms the
same current path in the high-current snapback mode.
The control circuit pin to ground zap combination demonstrated TLP character-
istics similar to the stand-alone device (Fig. 7.56).
This case demonstrated that despite layout isolation of SCR clamp, SCR clamp
is still likely triggered by the parasitic current from internal circuit.
The substrate current in DeMOS array is acting similar to the base current of the
parasitic NPN structure and provides lower turn-on voltage.
The following measures are usually considered to improve the design. For the
low-impedance switch pins, in case of large switching LDMOS arrays W > 30 mm
the circuit should rely on array self-protection. In this case corresponding design
optimization of the power array may be required. An alternative approach may
Pulsed Voltage (V)

0 20 40
1.E+01
Pulsed Current (A)

1.E+00
1.E-01
1.E-02
1.E-03 10ns TLP: OVP-GND
10ns TLP: Clamp

1.E-04
Fig. 7.56 Comparison of the TLP characteristics of the stand-alone ESD clamp with the
characteristics at OVP pin
involve application of non-snapback high holding voltage ESD clamp based on

lateral PNP or NPN device.
In case of small LDMOS arrays the gate clamp should be introduced to pro-
vide off-state for power array during ESD stress conditions and guarantee high
breakdown voltage; however, this measure might be inadequate in fast transient test
conditions.
For high-impedance control pins ESD clamp with pulsed triggering voltage sig-
nificantly above absolute maximum can be used in combination with the two-stage
protection approach. Typically this case is practical for the pins with large separation
resistor.
Finally in application the design measures are related to appropriate PCB design
that is minimizing the parasitic unductive components responsible for high-voltage
overshoots in transient regimes.
7.6.2 Nepi–Nepi Transient Latch-Up Scenario

Another case of transient latch-up is typical for high-voltage processes. This latch-
up current path can be realized either between two adjacent high-voltage pins
protected by high-voltage ESD devices or the ESD device and the internal circuitry.
The primary reason for latch-up path formation is parasitic NPN with variable base
potential or simply current injection in ESD test conditions.
The example of parasitic NPN and corresponding current path is illustrated in
Fig. 7.57 for the case of adjacent high-voltage control pin protected by NLDMOS-
SCR ESD clamp and the power pin protected by high holding voltage lateral PNP
ESD clamp. Both clamps can be referenced to the same power ground pin. However
during ESD test the power ground node is floating. This in general creates vari-
able conditions on the base of the parasitic NPN formed by Nepi regions of both
structures and P-isolation ring.
P-isolation ring
Nsinker epi ties
NLDMOS-SCR Clamp Lateral PNP Clamp
Power Ground
Fig. 7.57 Illustration of the alternative current path formed between two adjacent high-voltage
pins protected by NLDMOS-SCR and lateral PNP clamps in BCD process technology
The originally expected current path should be realized in each direction through
corresponding HV snapback clamp of positively biased pin and the reverse path
diode of the negatively biased pin. However if two clamps placed with minimum
isolation rules, an alternative current path is formed through the parasitic NPN
structure, described above.
This alternative current path remains open for ESD current. Therefore, the final
scenario depends upon the metallization routing and the internal circuit design. In
a favorable case the new current path can simply positively contribute to the ESD
protection level. However, in general an irreversible burnout may occur, since the
real current path through the internal circuit is hard to predict Fig. 7.58.
Several important conclusions from this case study can be derived. First of all the
high-voltage reverse path diode design in high-voltage clamps is critically impor-
tant. For example, often for the pin protection itself the body diode of the clamp
is sufficient. However, for the highlighted HV pin-to-HV pin combination the total
voltage may exceed the parasitic NPN turn on voltage especially in fast transient
mode, specifically due to excessive reverse path diode voltage drop.
A significant improvement of the body diode characteristics in high voltage
lateral LPNP clamp in BCD process technology for reverse ESD protection can be
accomplished by the following measures: enlarging n+ -base contact area, adding
Nsinker Epi Tie

Base contact
e
i ti
ep
er
nk
Si
N-
Emitter
2-contact base
Collector
1-contact base
Fig. 7.58 Partial layout view of the Lateral PNP Clamp with the device regions and comparison of
the negative TLP pulse characteristics for versions of the clamp with different N-base connection
base contacts or even by adding N-sinker epi tie to take an advantage of low buried
N-layer resistance.
The effect of these measures is presented in Fig. 7.59 for the case of 100 V lateral
PNP. All methods support 1.5 A current (>2 kV HBM) at less than 3.5 V voltage
drop. The highest current and lowest resistance achieved with N-sinker as expected,
which also improves the forward TLP I–V characteristics.
1.E+01
1.E+00
Pulsed Current (A)
1.E-01
1.E-02 b)
a)
L iso-iso L epi-epi 1.E-03
L iso-iso=10um, Vbe=0
1.E-04 L iso-iso=10um, Vbe float
L iso-iso=10um, Vbe=0.7V
L iso-iso=30um, Vbe=0
1.E-05
0 100 200 300
Pulsed Voltage (V)
L nbl-iso V be
c)
Fig. 7.59 Layout view of experimental structures for epi-to-epi isolation rules evaluation (a),
measured TLP characteristics for different spacing and isolation ring bias conditions (b), and
experimental dependence of the critical snapback voltage upon epi region separation (c)
Meanwhile, the main approach to suppression of the parasitic current involves

introduction of ESD layout guidelines for spacing limits between two epi- regions
connected to the high voltage. A simple three-terminal experimental device structure
to study the current path through parasitic NPN can be created and characterized to
determine the rules for different high-voltage epi regions isolation in ESD condi-
tions. According to the experimental data the critical voltage is a function of the
isolation region bias and the spacing between the epi-regions that represents the
base width.
7.6.3 CDM Case of the High-Voltage Pin Protection

Typically CDM ESD pulse is not of a major concern for small pin count analog
circuits due to smaller parasitic capacitance of the package. However, in some cases
CDM pulse directly impacts even small pin count high-voltage pins and special
measures should be taken.
Specific example below demonstrates the case of CDM failures in avalanche
diode-referenced CDM clamp (Fig. 7.60). The clamp is used to protect high voltage-
tolerant under-voltage lockout pin (UVLO). An external resistor divider from the
system input voltage sets the under-voltage turn-on threshold. An internal very low
current source provides hysteresis. This pin can be used for remote shutdown con-
trol. The ESD-related specific characteristic of this pin is a very low internal circuit
load.
ESDP
D
PE
G
B
S
R G
ESDM
Fig. 7.60 Simplified circuit diagram for NLDMOS-SCR clamp with high-side reference and
failure analysis photos of the damaged gate–source region of the NDeMOS-SCR clamp device
In a power product several pins have been protected by this clamp; however, only
the pins with low internal load, similar to UVLO, failed during ESD tests. Failure
analysis revealed gate–source region oxide damage.
Consequently the physical mechanism of the CDM failure is concluded from
failure analysis results to be gate oxide breakdown under CDM stress. The domi-
nant current path is through the high-side avalanche diode capacitance. To eliminate
ESDP
D
PE
G
B
S
RCDM
RG
ESDM
Fig. 7.61 An additional gate resistor RCDM design measure to eliminate CDM failure of the high-
side-referenced NLDMOS-SRC ESD snapback device
this undesired effect an additional resistor to protect the gate of the ESD protection
clamp is proposed (Fig. 7.61).
The additional experimental analysis for this case is presented below to explain
the process in the high-side-referenced NLDMOS SCR clamp under CDM stress.
The difference of the clamp characteristics in the CDM time domain can be revealed
by comparing the 100 ns standard TLP and 2 ns very fast TLP (vfTLP) measure-
ments. The difference is observed for both the drain–source and source–drain pulsed
I–V characteristics.
In case of reverse clamp path the clamp diode speed is not sufficient to clamp the
reverse path voltage low. For the particular package the 750 V CDM pulse roughly
corresponds to 5 A peak current level.
Comparison of the TLP and vfTLP (Fig. 7.62) shows significant difference in the
clamping voltage due to the difference in the measurement time domain.
Additional comparison using vfTLP measurements is made for the clamps with
and without high-side reference component. The result demonstrates much lower
maximum current for the clamp with high-side reference component (Fig. 7.63).
Further insight on this CDM failure is obtained using mixed-mode simulation
analysis. Based on simulation results it can be demonstrated that in the case of high-
side reference not only reverse, but even forward current path creates gate voltage
overstress.
Figure 7.64 provides comparison of the positive CDM pulse simulation for four
representative cases of the circuit: (i) with disconnected high-side avalanche diode
reference; (ii) with connected diode reference; (iii) the effect of the additional gate
protection resistor; and (iv) the effect of the additional capacitive load parallel to
the pin represented by reverse path diode. For positive CDM current direction the
presence of the reference components results in three times higher gate peak volt-
age that explains the gate oxide damage (Fig. 7.64c). This undesired effect can be
10–10 10–9 10–8
8
6
Current (A)
4
2
0
0 20 40 60 80 100
200
Voltage (V)
100
0
0 50 100 150 200

0 2 4 6 8
Current (A)
0 50 100 150 200

Time (ns)
Fig. 7.62 Comparison of the 2 ns vfTLP and TLP characteristics and illustration of the different
time domain measurement windows for the case of TLP and vfTLP
significantly reduced both by the additional gate protection resistor and by addition
of the load (Fig. 7.64c).
A different scenario is observed for negative stress. In this case the effect on
the reverse path diode or the body diode provided by internal circuit components
is critical. In case of 100 ns TLP or HBM time domain the parasitic high-voltage
diode formed by the drain and bulk Pbody or P+ -guard ring connection is usually
sufficient for reliable current path. A different behavior is observed in case of very
fast CDM event transient pulse. The waveforms for 500 V positive CDM stress with
and without resistor and with additional reverse path diode load are compared in
Fig. 7.65.
7.7 Summary 383
Current [A]
4
0
0 20 40 60 80 100 120 140 160
Voltage [V]
Fig. 7.63 Comparison of the 2 ns vfTLP characteristics for the clamp with and without reference
device
A significant overstress of the gate in CDM time domain is observed if the high-
side reference component is connected and there is no load. An improvement is
achieved by adding the gate resistor as well as in case of extra load provided by
body diode.
7.7 Summary
In this chapter, applications of ESD network principles and clamps are discussed
within the scope of different power circuits.
The specifics of power circuits significantly impact the ESD protection strategy.
Fast-switching and high-voltage pins, together with the lack of an ESD protection
window due to power-optimized output devices, create significant challenges and
often require circuit-dependent, custom ESD approaches. Self-protection aspects of
the integrated power devices in complex, multiple power domain networks are hard
to simulate and require substantial product-level experimentation.
In spite of similar basic principles for ESD protection network design and ESD
clamp solutions, a new characteristic feature of the power circuit consists in the
multiple current paths that can be realized during ESD event.
In the case of fast transient pins and output devices with pulsed absolute maxi-
mum voltages too close to operational limits, a significant effort is put into focused
design and selection of clamps with a precise voltage reference and a low dV/dt
triggering effect. The possibility of transient latch-up on transient pins or in case
of a hot plug-in circuits requires countermeasures. Among them are the design and
application of rather challenging high holding voltage solutions (especially in the
case of high-voltage circuits) and utilization of the self-protection capabilities of
the internal circuits. At the same time, realization of the two-stage ESD protection
network becomes rather challenging due to the impossibility of putting a separation
R1 R4 L0
1e8*(1-... 10 1u
ESDP
R5*
R3 C0*
1e8*(1+... 1e-12 R7*
100M
D0* 100M
W=100 M1
V0 $Vpulse...
D1
W=100
R6*
0.0001 W=200
R2*
1e4
ESDM
a)
b)
c)
Fig. 7.64 Mixed-model simulation circuit with positive 2000 V CDM pulse (a) and pulsed wave-
forms observed at the drain (b) and at the gate (c) of the ESD device for the cases with and without
high-side reference, gate protection resistor, and reverse path diode load
7.7 Summary 385
R1 R4 L0
1e8*(1-... 10 1u ESDP
R3 C0*
1e8*(1+... 1e-12 R5*
R7*
100M
V0 $Vpulse... D0* 100M

W=100 M1
D1
W=100
R6*
0.0001 W=200
R2*
1e4
ESDM
Fig. 7.65 Mixed-mode simulation circuit with negative CDM pulse (a) and pulsed waveforms
observed at the source (b) and at the gate (c) node of the ESD device (relative to the ground) for
the cases with and without high-side reference, gate protection resistor, and reverse path diode load
resistor on the power node. As a result, the strategy for protection changes from rail
based protection to individual device protection.
A general methodology of implementing appropriate ESD protection contains
several steps. It depends upon whether the ESD protection network is created for an
existing, active design, or, conversely, a co-design of the internal voltage regulator
circuit, the ESD protection, and even the components can be accomplished. The
latter is apparently preferable.
The fist step of the methodology is to identify the pin types: control, power,
switch, boost, the functionalities of feedback pins, voltage tolerance. Transient sig-
nal, and other specification parameters determine the most suitable network design
and the most appropriate clamps.
The second step requires a significant amount of “navigation” from each pin
inside the circuit block hierarchy in order to reveal the most exposed devices that
will provide an alternative current path. The following step is focused on estimation
the reversible current level in ESD conditions which can be supported by the inter-
nal circuit without any damage to the active devices or interconnects. To achieve a
suitable level of confidence in the estimation, understanding of both the pulsed SOA
of the components and the circuit operation is critical.
These two steps usually require substantial effort and experience with similar
products. They involve a number of assumptions and provide relatively low initial
accuracy of estimation for unique circuit design cases. However, this is the most
widely used approach among industry experts. Part of the needed experience is
illustrated with the case studies presented in this chapter.
However, the most effective way is to combine this valuable experience-based
empirical approach with simulation tools. Examples of the mixed-mode and com-
pact model simulations with ESD snapback models have been presented across this
chapter and described extensively in Chapter 5.
The following steps involve choosing the ESD protection network compo-
nents and selection of clamps with the most appropriate characteristics, as well as
increasing the self-protection capabilities.
When the ESD protection network schematic is finalized, finding an appropriate
ESD layout is the next critical step. In addition to the obvious backend layout design
that supports the ESD current level at a low voltage drop, more thorough work is
required to eliminate possible cross-talk (due to the injected current into the sub-
strate) between the internal circuit and the ESD clamp. This is usually achieved by
both the guard ring structure and an appropriate arrangement of the ESD clamp on
the chip. For example, in most cases, the high-voltage ESD clamp should be suf-
ficiently separated from the power array to avoid the false triggering by the high
substrate current provided by the array. Other layout-related efforts are focused on
the elimination of the parasitic coupling between the metal layers that can provide
undesirable gate coupling of the components.
For integrated power arrays, the self-protection capability is a layout-dependent

parameter [136]. Recent research work [113] demonstrates the possibility of esti-
mating the self-protection operation of power arrays using a 2.5D approach to
extract an equivalent circuit of the array combined with simulation [137] and
visualization [138] tools.
When the circuit and layout design work is completed, the incoming test results
are the most important factor for further understanding of the ESD network and
circuit coupling. If reliability improvement steps have to be taken because of prod-
uct failures under ESD test, the partition test, combined with reprocessing-level FA
results, becomes very important. Usually, the failure location is repeatable and the
major challenge is in creating an appropriate solution. Product-level validation of
the alternative solutions is usually a rather expensive approach. This is why both
empirical analysis and numerical evaluation of the proposed solution may greatly
contribute to the level of confidence in its success.

Example 7.1 Output Stage of Buck DC–DC Voltage Regulator

Library Name: Examples7_Power_Management_Analog_Circuits
Project Name: E7.1a_5 V_Synchronous_Buck_No_Bond_Wire_Inductance
Project Name: E7.1b_5 V_Synchronous_Buck_Normal_Operation
Project Name: E7.1c_5 V_Synchronous_Buck_Transient_Latchup_Case
This set of examples presents case studies for the simplified version of the output
stage of the synchronous 5 V buck DC–DC converter. In one example, the bond wire
inductance effect is shown to produce a high-voltage transient at the power pins.
In normal operation regime, depending on circuit parameters, this may result in a
transient latch-up event due to triggering of the NMOS ESD clamp into snapback.
The mixed-mode simulation examples for the output stage of synchronous DC–DC
converter can also be used to understand the principles of operation of the voltage
regulator and the effect of the physical parameters of the finite element components
on the output stage characteristics, including switching speed and losses (Fig. E7.1).
L1
VIN
2n
R1* M1*
PGATE_CLK
R4*
1
V1* $VDD*pu... 1
L2 L4
W=40000 SWITCH VOUT V2 $Vdd*pu...
M0*
2n 10u
C3*
R3* 10u R2*
NGATE_CLK 1000
1
V3* $VDD*pu... W=20000 L3
PGND
2n
Fig. E7.1 Mixed-mode simulation circuit for simplified output stage of 5 V DC–DC converter
including bond wire inductances and simulation results for normal operation regime; mixed-mode
circuit for transient latch-up simulation and results of analysis demonstrating induced parasitic
snapback in the ESD protection clamp device
Example 7.2 5 V Boost DC–DC Converter and Transient Latch-Up

Project Name: E7.2a_5 V_Boost_HNMOS_Normal_Operation
Project Name: E7.2b_5 V_Boost_HNMOS_Transient_Latchup
This set of examples presents case studies of the simplified version of the output
stage of the asynchronous 5 V boost DC–DC converter with the switch pin protected
by a snapback NMOS ESD clamp. The example demonstrates both the principle
of voltage regulation and possible transient latch-up event due to snapback of the
NMOS ESD clamp (Fig. E7.2).
SWITCH ESD_Clamp D0*

L4
10u
M0* W=1000
M1
R4*
0.1 VOUT
R3*
NGATE_CLK
D1*
1 W=10000 W=50 C3
V3* $Vdd*pu... 10u R2*
V2 $VIN*pu... W=400
1000
R1*
10K
PGND
Fig. E7.2 Mixed-mode simulation circuit for the simplified output stage of the 5 V DC–DC
boost converter and simulation results for the normal operation regime showing output and switch
voltages for different VIN voltage levels
Example 7.3 High-Voltage Boost Output Stage

Project Name: E7.3_20 V_Boost_NLDMOS
This example provides a starting point for mixed-mode analysis of a simplified

version of the output stage of the asynchronous 20 V Boost voltage converter
(Fig. E7.3). An optional high-voltage ESD protection clamp can be added to analyze
parasitic effects during ESD stress and under normal operation.
VIN SWITCH D0*

L4
10u M0*
W=1000
R4*
0.1
R3*
NGATE_CLK
1 W=10000
V3* $Vdd*pu...
V2* $VIN*pu... VOUT
C3
10u R2*
1000
PGND
PGND
Fig. E7.3 Mixed-mode simulation circuit for a simplified output stage of the 20 V DC–DC boost
converter, simulation results for the normal operation regime showing output voltage VOUT at
different VIN voltage levels, gate drive voltage and diode current waveforms
Example 7.4 100–600 V Boost Output Stage with Vertical DMOS

and IGBT

Project Name: E7.4a_100 V_Boost_VDMOS
Project Name: E7.4b_100 V_Boost_IGBT
These examples present mixed-mode analysis of a simplified version of the output
stage of the 100–600 V boost voltage regulator. The output stage is assembled using
parameterized vertical DMOS and IGBT devices that represent discrete components
(Fig. E7.4).
Example 7.5 Power Array with Gate Clamp Example

Project Name: E7.5_Gate Clamp
L4 D0*
10u
W=1000
R4* IGBT0*
0.1
R3
NGATE_CLK
1
V3* $Vdd*pu...
V2* $VIN*pu... W=10000 VOUT
C3
10u R2*
1000
PGND
Fig. E7.4 Mixed-mode simulation circuit for a simplified output stage of the boost converter and
simulation results for the normal operation regime showing output voltage VOUT at different VIN
voltage levels, gate drive voltage and VDMOS drain current waveforms
The example presents a version of the so-called gate clamp implementation. The
purpose of the gate clamp is to ensure a relatively low gate bias of the NMOS
array in the ESD pulse time domain, but enable switching of the device in normal
operation mode. A simulation with two HBM circuits is performed in parallel to
demonstrate the transient gate voltage of the power device with and without the gate
clamp (Fig. E7.5) and the higher drain voltage in the circuit with the gate clamp.
Additional optional snapback clamps can be added in parallel to the power array to
observe the alternative ESD current paths through the clamp and though the array.
Example 7.6 Serial Data Line Pin Case

Project Name: E7.6_SDA_pin
C2 C6
2.6e-12 2.6e-12
R3 SW1 L0 R1 SW2 R4 HBM R5 L1 R2 R6
12.6e-6 1.5K 12.6e-6 1e8*(1-... 1.5K

1e8*(1+... 1e8*(1-... 1e8*(1+... C5
C1
C0 1e-11 C4 1e-11
V0 $Vpulse... 1e-10 V2 $Vpulse... 1e-10
SWITCH M1 SW
M4
C3* M3
1p W=20000 W=20000
M5 R7
M6 W=100 100M
V1 0 W=5
W=1 PGND
Fig. E7.5 Mixed-mode simulation circuits for power array with and without gate clamp and
simulated gate and drain voltage waveforms
This example represents the case of the serial data line pin ESD protection with a
snapback NMOS clamp. Different current paths through the clamp and the inter-
nal NMOS device can be observed in the simulation, depending on the passive
component values and active device parameters (Fig. E7.6).
C2
2.6e-12
L0 R1 R4
12.6e-6 1e8*(1-... 1.5K C1

R3 C0 1e-11
1e8*(1+... 1e-10
SDA
R5*
V0 $Vpulse...
50
M2* M1*
C3
1.5p
W=400 W=100
R2*
1e4
Fig. E7.6 Mixed-mode simulation circuit and transient current through the internal circuit NMOS
M2 for different width of the protection device M1
Chapter 8
System-Level and Discrete Components ESD
This final chapter presents material on a rather cross-disciplinary subject related

to system-level ESD robustness. System-level ESD requirements are defined by
different standards and specifications than component-level ESD requirements.
Component-level ESD standards are specified to ensure ESD robustness of the
integrated circuits and components during manufacturing and handling inside the
controlled, protected ESD environment until the components are incorporated into
the system. ESD specifications on the component level are circuit specific and pack-
age specific, practically until the component is mounted on the print circuit board.
It is expected that when the component is mounted in the properly designed sys-
tem, the system’s design will guarantee the absence of events that result in stress
of the components above the absolute maximum limits, even when the system itself
experiences an ESD event.
System-level ESD robustness is required to ensure robustness of the whole sys-
tem during system manufacturing and operation. This can be achieved both by
different system design architecture that involves transient voltage suppressors and
by implementation of the on-chip ESD protection for the pins. The focus of this
chapter is specifically on the on-chip system-level protection.
The chapter is composed as follows. First, Section 8.1 provides an introductory
portion of the material related to the understanding of the standards and ESD system
specifics. Section 8.2 brings the human–machine model approach that enables on-
chip system design for ESD clamps and protection networks. In Section 8.3, several
application cases for on-chip ESD protection are discussed to demonstrate on-chip
protection peculiarities and the general approach. A solution to one of the particular
cases of system requirements related to hot swap and plug-in is discussed in Section
8.4. Finally, in Section 8.5, the chapter is complemented by a discussion of the ESD
robustness of discrete components.

396 8 System-Level and Discrete Components ESD
8.1 System-Level Specifications and Standards
8.1.1 Meaning of ESD Robust System

ANSI/ESD S20.20 is an ESD control program that provides a systematic approach
to protecting components and equipment susceptible to ESD damage starting from
HBM discharges of >100 V. A set of standards related to the measurement of static
control in footwear, packaging, flooring, conveyor belts, ionizers, etc. However, it
is difficult to maintain this level of control 100% of time and everywhere due to the
cost associated with these measures.
The component-level ESD standards protect the ICs from damage within a static
controlled environment. These standards assume basic ESD control practices that
are typical for a 2 kV human body model, a 200 V machine model, and a 750 V
charge device model. This applies to IC’s before they are mounted on a board.
Thus, the focus of ESD protection shifts from chip hardening to system hard-
ening. The previously mentioned CDM, MM, and HBM standards apply to chip
protection, but do not guarantee system level protection. The end system is tested
according to system-level standards with an “ESD gun.” For example, the surface of
the device is divided into a grid. Each grid space is zapped a minimum of 10 times
(Fig. 8.1a). Failure can be catastrophic damage, latch-up (device requires power-off
to reset), or “soft” damage, for example, a display flicker.
(a)
(b)
Fig. 8.1 System testing wit ESD gun (a) and CE mark image (b)
In principle, successful system design requires understanding beyond the chip

level. In particular, ESD is not just a handling issue. For system-level protection
design, it is important to know what circuitry is connected on the board to the chip
in the final system design, if the particular current path is directly connected to
the external system pins, and what kind of connection and cable is in place. More
details may be required to identify if the connection can be probed or touched during
system repairs or maintenance. Thus, understanding of the relevant system (ESD)
qualification standards can directly affect the design.
8.1 System-Level Specifications and Standards 397
ESD robustness is one of the quality features of the system. At the same time,
the system is the final application of an integrated circuit component. A robust ESD
system ensures reliable operation during and after ESD stress.
8.1.1.1 CE Mark
Depending on the criticality of the application, different levels of functionality have
to be maintained during or after ESD stress. Passing appropriate standard ESD tests
is a legal requirement for systems that udergo corresponding certification proce-
dures. For example, the IEC 61000-4-2 standard has to be met to obtain the CE
certification mark. Since 1996, all medical consumer electronics and many other
systems must conform to the 89/336/EC (EMI and ESD) and display the CE mark
to be sold in Europe.
The CE mark (Fig. 8.1b) is a label affixed to a product to show its compliance
with the electromagnetic compatibility (EMC) directive and with the appropri-
ate safety directive: the machinery directive for the safety of machines or the
low-voltage directive for the safety of electronic products. The manufacturer is
responsible for selecting the safety directive most appropriate for his product. Many
products include both the electrical and mechanical components.
The standards hierarchy for EU includes Product or Product Family: tests, test
level, failure criteria; Generic EMC: simulators, test setups, procedure, if a product
standard does not exist; Basic EMC: simulators, test set-ups, procedure.
8.1.1.2 Basic EMC Standards

Practically every manufacturer in the system development business must perform
tests to satisfy the following Basic EMC Standards:
ESD immunity EN 61000-4-2
EFT – electrical fast transients EN 61000-4-4
Surge immunity EN 61000-4-3/6
RF emissions Various
Harmonic/flicker tests EN 61000-3-2/3
Recently, European Norm (EN) naming conventions have been used for existing
standards: EN61000-4-x = IEC 61000-4-x, etc.
The object of the IEC 61000-4-2 standard is to establish a common reproducible
basis for evaluating the performance of electrical and electronic equipment when
subjected to electrostatic discharges; typical waveform of the discharge current; the
range of test level; test equipment; set-up; test procedure.
8.1.1.3 Automotive Industry Standards

Automotive industry applies the ISO 10605 standard. Functional Status according
the ISO 10605 includes the following classes:
– Class A: all functions of a device or system perform as designed during and after
exposure to interference.
– Class B: all functions of a device/system perform as designed during exposure;

however, one or more may go beyond the specified tolerance. All functions return
automatically within normal limits after exposure is removed. Memory functions
shall remain class A.
– Class C: one or more functions of the device or system do not perform as designed
during exposure, but return automatically to normal operation after exposure is
removed.
– Class D: one or more functions of the device or system do not perform as designed
during exposure and do not return to normal operation until exposure is removed;
the device or system is reset by a simple operator/use action.
– Class E: one or more functions of the device or system do not perform as designed
during and after exposure and cannot return to proper operation without repair or
replacement of the device or system.
8.1.1.4 IC and System-Level Comparison

It is important that the tests and protection methods are applied specifically to ESD
robust systems (like IEC 61000-4-2), rather than components. In principle, the com-
ponents formally do not necessarily need to meet any special requirements beyond
the safe packaging and handling if the system design is compliant with the system-
level specifications. Thus, for components, there are other ESD discharge models
that might be more relevant: HBM, MM, charged device model (CDM), charged
board model (CBM), and the cable discharge event (CDE) [139].
However, the trend set by the system designers is to elevate the passing level for
component ESD tests on the chip in order to help system-level ESD design.
Manufacturing and application cycles of the IC component can be subdivided
into the IC component-level ESD and system-level ESD (Fig. 8.2). During the
component-level ESD, the IC is always handled inside the ESD protected area
(EPA), where ESD protection methods and design are implemented in a well-
controlled environment. In this environment, presence of ESD events with pulse
amplitude above the component-level ESD (2 kV HBM, 200 V MM, 1 kV CDM)
is practically excluded. In reality, the levels of ESD are much lower. This is why
reduction of the component-level minimum ESD requirements to 1 kV HBM and
100 V MM is recently in discussion.
Thus, practically, on-chip ESD protection compliant with several hundred volts
HBM guarantees safe handling. Protection realized between any pin combinations
covers the life cycle from backend manufacturing to board mounting inside EPA.
Outside EPA, the protection methods and design of the system include several
measures. For example: shielding, pre-running ground, application of the system-
level ESD components, for example, transient voltage suppressors (TVS), to achieve
the desired system-level specification.
In general, system-level ESD protection is necessary for satisfying the corre-
sponding much higher levels of ESD currents, in comparison with the package-level
requirements. At the same time, properly designed ESD protection on the packaging
level can protect components even outside EPA. Usually, the enhanced protection
(a)
(b)
Fig. 8.2 Typical sign to designate ESD-protected area (EPA) (a) IC and system-level ESD
protection and measures (b) [139]
design applies only to the very few exposed pins with system-level requirements,
rather than to the whole component. In this case, IC-level ESD protection design for
the non-exposed pins does not influence system-level ESD robustness.
A typical system-level ESD exposure during assembly can include different sce-
narios. For example, in a car assembly, a discharge may occur when a cable is
plugged into a control unit. In a mobile assembly, a discharge is typical when boards
are handled or sub-boards are connected to the main board. Typical system-level
ESD exposure situations during end-user operation usually involve antenna pins
overstress; picking up energy of an ESD discharge in a nearby surrounding area;
discharge of cables and connectors during plug-in or wiring; damage of the battery
supply pin during changing of batteries; discharge inside a system due to board or
housing design flaws.
In effect, any metallic part which can be touched by the end-user and which is
close to the IC pins or the board can expose system to an ESD event. System-level
ESD, according to IEC 61000-4-2, mimic a discharge to a system (outside an EPA)
from a handheld metallic tool.
For the system level ESD, direct discharge to “open” connectors is excluded.
For communication systems, discharges through cables and equipment may be real
threats (CDE – cable discharge event). Examples include USB, FireWire interface
in PC/camera/peripherals; Ethernet (cables may be very long), BNC cables; auto-
motive wiring harness. Different discharge types may contain an “initial discharge”
that corresponds to the connection of the charged cable to the pins, a “secondary
discharge” that corresponds to a change in the connection, for example, a short cir-
cuit at the other end of the cable, or a combination of discharge from both cables
and devices [140].
8.1.2 System-Level ESD Pulse and Model
8.1.2.1 System-Level ESD Test for ICs

The application board capacitor has already been emphasized for the case of the
FDP-link (Section 6.3.3). Thus, at system-level IC testing, it is important in principle
to have a special application board specifically for this type of testing. The role of
the application board is to mimic the final system board size and the ground planes.
In addition, special modifications to the board are often required to allow zap access
for the pins, due to standard system-level gun dimensions. In most cases, only some
specific pins are required to undergo system-level tests while other pins are not
subject to such requirements. One of the challenges is correctly interpreting these
requirements: what ports are to be tested; what is the system board design.
The board design itself can also help pass the system-level test. For example, in
addition to the decoupling capacitors effect, the board inductances will attenuate the
peak current (Fig. 8.3).
Typically for on-chip system-level protection there is an area and performance
trade-off. On chip protection provides an advantage in reduction of component
count, but at the same time creates an overhead in silicon area, and complication
in test/qualification. The complications may require running several learning cycles
to correctly target the “ESD window.” Thus, component and system tests must be
carefully specified in advance. The real target is the result in the end system.
8.1.2.2 IEC ESD Pulse Waveforms

System-level ESD pulse has two-peak waveform (Fig. 8.4a, b) according to IEC
61000-4-2. On the physical level, the first peak is caused by a metallic part that
touches the system-level pin. This peak has a fast rise time and high amplitude, but
is short due to the low resistance of the metallic part and limited charge. The first
peak is followed, with corresponding delay, by a discharge that physically represents
400
300
Voltage (V)
200
100
0
10 20 30 40 50 60 70 80
Time (ns)
Fig. 8.3 Various inductance components on representative board trace and comparison of the
measured waveforms at ESD suppressor and at input pin of an IC
a charged body of an end-user or a tool. This peak has lower amplitude due to higher
resistance versus the metallic part and a corresponding delay and width, because the
object discharged is much larger.
Standard documents usually define a waveform, not the equivalent circuit. For
the IEC 61000-4-2 pulse, the waveform is defined by several representative points
of current amplitude during the transient event. For the first peak, the specific current
per 1000 V of the precharge level is I1st peak = 3.75 A/kV; the second specific current
level per precharge level is defined after 30 ns from the first peak: I30 ns = 2 A/kV;
the next current level is I60 ns = 1 A/kV. In addition, the rise time for the first peak
of the pulse is defined within trise =100-300 ps and full width ≈ 1 ns.
To replicate the IEC 61000-4-2/ISO 10605 standard, one of the main compo-
nents of the equivalent tester circuit can be roughly represented by the Rsys Csys
network with parameters 330 and 150 pF. However, such a network will not pro-
vide a desired waveform by itself. A more accurate equivalent circuit is discussed
in Section 8.2 for the human machine model (HMM).
A comparison of HBM and HMM pulses of the same amplitude is presented in
Fig. 8.4b. As can be seen, the differences between the waveforms are in both the
pulse shape and peak current level.
(a)
(b)
(c)
Fig. 8.4 Typical output waveform of ESD simulator (4 kV) (a) experimental waveforms for
IEC61000-4-2 and comparison with HBM (b) and setup for IEC61000-4-2 system-level test (c)
From a practical ESD design point of view, a larger ESD current level is not
a major problem in the case of system-level ESD stress. It can be handled by a
corresponding width scaling of the ESD clamps to support system-level ESD cur-
rent. However, the task that provides major challenges for on-chip system-level ESD
design is withstanding ESD stress under power-on conditions.
While component tests are only performed with the IC unpowered, according to
standards, the system test is done with both the unpowered system and with power
applied to the functioning device. The subsequent requirements are that there must
be no physical damage and the device must keep working normally after the stress
has ended. This means latch-up must not occur on either the power or signal pins,
as that will require system reset or power-down to bring the system back to normal
operation.
Finally, unlike HBM testing, IEC stressing is done in two ways: contact dis-
charge and air-gap (spark) discharge. Due to energy dissipation (loss) to the air, the
discharge voltage requirement is higher for the air-gap test.
– Class 1 – 2 kV contact, 2 kV air discharge

8.1.2.3 System-Level Test Setup

A calibrated waveform is usually realized by commercial testers. The test setup for a
system test is defined by the standard document and is rather non-trivial. It includes
the gun that provides discharge to the system, which is placed on the isolation mat on
top of the table with the metal plate grounded through the 470 K resistor (Fig. 8.4).
It is clear that such a setup is based upon the coupling of the system with the ground
plane. While the dielectric mat for the test is defined by the standard, the coupling
will depend on system size and ground isolation. In particular, a portable system
with a battery power supply might provide a totally different ESD current path than
a similar system with a grid-connected power supply module that provides a low-
impedance ground connection.
A comparison between the component and the system-level ESD stress is
summarized in Table 8.1.
In each IC component used in the design of the system, there are different
classes of pins that can be subdivided according to specific system-level ESD
robustness (Fig. 8.5). The most common pins are intended for the on-board inter-
chip connection. These pins are not critical for system ESD robustness and only
need to meet the component ESD requirements. The second group of pins is
attached to the signal busses connecting several PCBs. These pins are poten-
tially exposed during repair and maintenance, which can be done violating EPA
requirements.
Table 8.1 Comparison of device and system-level ESD stress
Function Device level ESD test System level ESD Test (IEC)
Stressed pin group All pin combinations Few special pins

Supply Unpowered Powered & unpowered
Test methodology Standardized Application specific
Test set-up Commerical tester & Application specific
sockets
Typical qualification 1 ...2 kv JEDEC HBM 8 kv Contact (IEC
goal 61000-4-2) 15kv Air (IEC
61000-4-2)
Corresponding peak 0.65 ... 1.3 A > 20 A
current
Fallure signature Destructive Functional or destructive
Fig. 8.5 Illustration for the three groups of pins with different system-level ESD requirements
The third group of pins is the most critical for system-level ESD; these are the
pins directly or indirectly attached to the external connectors, for example, exposed
USB pins. Depending on other connected components, these pins will experience
either the whole or a partial level of system-level ESD stress. Typically, for such a
group of pins, some increased ESD protection level is required.
8.1.2.4 Cable Discharge Event (CDE)

Cable discharge events occur during operation in the ports of systems, both in pow-
ered and in non-powered conditions. In real life, conditions for the primary charge
can usually be formed due to the triboelectric charging of the cables handled across
the facility prior to installation or during transportation from the manufacturer.
The standards for CDE are not yet established and therefore specific custom
requirements, involving the twisted cable pair precharge level and the length of the
cable (for example 2 kV and 200 m, respectively), are often applied to define the
test setup. The CDE is usually composed of two-stage events. The first part repli-
cates the connection of the charged cable to the pin and the second part replicates
the creation of short-circuit conditions on the opposite end of the connected cable.
When charged cables are connected to electronic products (or vice versa), the
discharge is observed through the relatively low impedance of the cables. Cable
discharge events (CDE) generally have similar waveforms for the initial discharge
to the IEC, which represents the fast discharge of the low impedance part of the
cable closest to the pin, and the discharge of the remaining part of the cable with a
larger delay and impedance. Due to reflections from the unterminated cable ends, the
follow-up current from the energy stored in the cable is in an oscillatory waveform,
the frequency of which depends on the cable length. Today, ESDA WG14 is tasked
with developing a standard practice for CDE, followed by draft and review.
An additional comparison of different standard pulses and the system tests are
presented in Table 8.2. In particular for the presented data, it is clear that different
system-level electrical overstress requirements may greatly exceed the accumulated
charge level of the component-level specification by several orders of magnitude.
8.1.3 Transient Latch-up During a System-Level Event

One of the typical examples of different solutions for the IEEE 1284 interface
standard is presented in Fig. 8.6. The original solution requires an external tran-
sient voltage suppressors (TVS) discrete ESD protection component for interface
reliability. Plugging and unplugging printer cables can subject the printer ports to
high electrostatic discharge conditions, which may ultimately render the printer
inoperable.
External protection diodes add cost and consume PCB space, increasing the size
of system blocks and eventually the cost of the system itself. Therefore, a more
advanced solution represents system-level ESD protection integrated into the chip
(Fig. 8.6).
Thus, introduction of system-level protection for the specified pins is an
admirable goal. However, a new challenge in meeting the test specifications for a
powered system arises due to the transient latch-up induced by the ESD devices and
other parasitic on-chip components, as a result of the high current injected into the
components.
As can be seen in the previous chapters of this book, in the case of component
ESD protection requirements, local ESD protection can often be successfully real-
ized using snapback clamps. These clamps simply turn off at the end of the ESD
discharge event. This allows application of small footprint ESD protection clamps
with holding voltages below the power supply level, or voltage levels at the pin in
operation conditions.
However, in the case of power-on conditions, self turn-off may not occur. Self-
turn-off will generally be a function of the load resistance at the pin, the difference
between the minimum holding voltage and power supply level, and the critical turn-
off current of the clamp.
406
Table 8.2 Comparison of device and system-level ESD stress for different pulses [141]
Standard/Pulse Application V[V] Duration (10–90%) # Pulses Rl [] C [pF] Ipeak [A] Charge Charge rel. to HBM
RC discharge generated pulses

“HBM” JESD22-114 Component 8k 150 ns 1 15k 100 5.3 800 nC 1
“MM” JESD22-116 Component 400 16 MHZ 1 – >8 0.8 nC 1 0E−03
“CDM” JESD22-101 Component 75 0.8 ns 1 1 8.5–17 – –
“GUN” IEC 61000-4-2 System 8k 120 ns 10 330 150 33 1.2 μC 1.5
ISO/TR 10605 inside System/Vehicle 8k 1μs 3 2k 330 30 2 64 μC 3.3
ISO/TR 10605 outside System/Vehicle 8k 360 ns 3 2k 150 30 1.2 μC 1.5
Voltage generated pulses
ISO 7637: 1 System −100 2 ms 5000 10 – 10 20 mC 2.5E + 04
ISO 7637: 2 System 100 50 μs 5000 10 – 10 0.5 mC 6.3E + 02
ISO 7637: 3a System 150 100 ns 3.6E6 (lh) 50 – 3 300 nC 0.375
ISO 7637: 3b System 150 100 ns 3.6E6 (lh) 50 – 2 200 nC 0.25
ISO 7637: 5 (load dump) System 87 40–400 ms 1.10 2 – 43 17 C 2 lE+06
8 System-Level and Discrete Components ESD
Fig. 8.6 Integration of the external protection diodes into the chip to provide protection in
accordance with IEC 61000-4-2 on the connector interface pins
Some of these parameters are rather hard to measure. The minimum holding
voltage measured using a 50 TLP system may not correspond to the real
value at a higher load resistance. Short pulse measurements of the turn-off current
may not correlate with the actual current at long pulse conditions, where current
filamentation phenomena may take over the current redistribution effect.
Thus, product verification of the on-chip system solutions is rather critical. In
practice, the ESD protection cell is expected to provide an appropriate system-level
current capability with the holding voltage of the used ESD device above the power
supply level. In the opposite case, depending on the load resistance of the pin, a
transient latch-up may occur. In this case, after an ESD event, the current through
the snapback device in the high-current state will be continuously supported by the
power supply, keeping the ESD device in high-conductivity state. The following
shunting of the pin may result in loss of functionality or heat generation, and the
high current density may result in accelerated degradation and permanent burnout
of the ESD device in the latch-up state.
The waveforms for such an event are illustrated in Fig. 8.7 for the latch-up event
induced in a parasitic SCR structure formed by a guard ring structure connected to
VDD and VSS busses [140]. At the pulse amplitude of up to 18 V, the device always
returns to the original low current state. However, if the voltage is increased to 20 V
and then reduced down to a lower level of ∼3 V, the latch-up between VDD and VSS
lines is observed, due to high-current path formed.
Additionally, a numerical simulation for devices capable of hot plug-in (dis-
cussed in Section 8.3) also illustrates the transient latch-up effect.
An additional important comment should be made: in real application, many sys-
tems may be exposed to other kinds of electrical overstress with significantly higher
power than ESD pulses.
Fig. 8.7 Illustration of the transient latch-up in the chip with a parasitic p–n–p–n structure
ESD structures (both on-board and on-chip) are first hit by these pulses. In
general, these devices are not designed to withstand arbitrary overstress pulses.
Therefore, robustness against EOS pulses has to be specified and taken into account
explicitly in IC and board design.
8.1.4 System-Level Protection Components

To help system-level design, there is a great variety of transient voltage suppressors
(TVS) available on the market. These suppressors essentially represent a discrete
version of the ESD protection clamp with parameters defined by the datasheet. In
the recent years, more sophisticated modules that combine arrays of suppressors,
targeting specific automotive, high speed, low parasitic, integrated EMI filters, and
other requirements became available. An example data sheet for such a component
is presented in Fig. 8.8.
The TVS are used on PCB to provide a primary current path for the system
terminal connected to the IC system pins, and often include additional separation
components to realize a two-stage ESD protection network.
One of the alternative ways to manufacture suppressors in silicon process tech-
nology is based upon polymeric material with nanoparticles that provide extremely
low parasitic capacitance, in comparison with semiconductor-junction base suppres-
sors [142].
8.2 On-Wafer Human Metal Model Measurements 409
Fig. 8.8 Example of the datasheets for transient voltage suppressor
8.2 On-Wafer Human Metal Model Measurements

The material of this chapter is focused specifically on the on-chip system-level
protection, as it is a major trend in portable system design. On-chip system-level
protection creates advantages in both cost and appeal of small form factor products.
At present, characterization methods for system-level and component-level ESD
stresses exist in parallel, without a real correlation between the testing results. In
[143, 144], the correlation between CDM, HBM, and electrical overstress testing at
the component level and tests that use different discharge networks at the system
level has been studied. A correlation between system-level and component-level
ESD characterization has been found. However, these results are obtained for only
one type of technology and cannot be generalized for other applications.
When system-level pins are identified, corresponding design actions are taken
to create an ESD cell with an adequate width toward the appropriate current level.
The cell must not only target the appropriate ESD protection window, but must
also meet the size and holding voltage requirements. Such design activity involves
a significant amount of test chip-level experimentation and comparative analysis of
the stand-alone cells. It would be rather inefficient to conduct such analysis on the
IC or, especially, the system level.
At the same time, testing of the individual semiconductor components is essen-

tially outside the scope of the IEC61000-4-2 standard. One of the typical decisions
applied by manufacturers in overcoming this issue during component development
is the fabrication of IC’s with an elevated component ESD protection requirements
for system-level interacting pins, for example, 4, 8, or 15 kV HBM.
However, as will be shown in this section, the correlation between component
and system-level ESD pulses is not a given.
Another alternative is to test packaged devices using system-level tools. In this
case, perhaps only the worst-case scenario can be reproduced for the component pin
discharged to the ground, since the capacitive coupling of the system through the
dielectric mat is hard to replicate.
Therefore, the human metal model (HMM) alternative has been proposed to
create a recommended practice of testing semiconductor components and devices
using the IEC (EN) 61000-4-2 basic. To allow the IC manufacturer to predict the
ESD performance of their products under system-level stress conditions, the ESD
Association (ESDA) has proposed a new measurement method: the HMM [145,
146, 147, 148]. The objective of the HMM initiative is to establish a common repro-
ducible basis for evaluating the performance of electrical and electronic equipment
when subjected to electrostatic discharges. In this section, HMM tester for on-wafer
measurements is described.
8.2.1 On-Wafer HMM Tester and Equivalent Circuit of the Pulse

HMM reproduces the ESD discharge caused by a human touching a pin of a
grounded electrical component with a metal tool.
The goal of this new HMM standard practice (SP) document is to simulate a
system-level type of ESD stress at the component level. The focus of this SP is
on components that have a direct connection to the stressed pin, such as interface
circuits, external ports.
Studying the ESD performance of such components during the design phase
requires on-wafer ESD measurements. Using an on-wafer HMM testing setup
enables the study of protected ESD components under system-level stress conditions
without prior packaging.
A version of the HMM ESD tester is presented in [144] using an on-wafer testing
method. With the proposed setup, new results are obtained from the devices manu-
factured in both advanced CMOS and high-voltage technologies. The characteriza-
tion results are further compared with those of the commonly used HBM testing.
HMM simulates a system-level type of ESD stress at the component level.
Stress waveforms with a shape defined in the IEC 61000-4-2 standard document
are applied to the device under test (DUT). The HMM standard practice, currently
developed by the ESDA standards committee WG 5.6, uses an ESD gun as the
pulse generator. Such an ESD gun contains a discharge circuit that conforms to the
IEC61000-4-2 system-level ESD standard.
ESD guns have several disadvantages when used to test components. One issue is
the electromagnetic field produced by an ESD gun during the application of a stress.
The spread of the radiated field is not known and varies strongly between different
gun models [145]. Also, the form factor of most of the available gun models is not
suitable for the integration of such a gun into a component-level measurement setup.
An original on-wafer HMM testing setup using an ESD gun as the pulse source
is presented in the standard (Fig. 8.9). Recently, this setup has been improved by
replacing the ESD gun with a novel HMM testing module (Fig. 8.9b). This module,
mounted on a standard wafer prober, delivers a discharge stress waveform accord-
ing to the IEC61000-4-2 standard. Due to the specific hardware, electromagnetic
disturbances do not occur during the discharge.
a)
b) voltage probe
V
HANWA IHMM
HMM DUT current
tester probe
module
Fig. 8.9 Wafer-level HMM measurements setup with a commercially available gun tool (a) and
specially developed HMM module (b) [143]
8.2.1.1 Equivalent Circuit for HMM Simulation

With this improvement and the addition of voltage and current probes, it is possible
to simultaneously measure voltage and current waveforms during an HMM stress.
The current is measured as a function of time using a Tektronix CT-2 current probe
(Fig. 8.9). In [149], the current probe is capable of accurately measuring the HMM
R1 L1 R2 L2
C1 C2 CB DUT
Fig. 8.10 Equivalent circuit for HMM pulse with two precharged capacitors C1 and C2
current waveform up to the required 30 A of peak current for a HMM precharge

voltage of 8000 V.
The equivalent circuit for the HMM pulse is presented in Fig. 8.10, with param-
eters of the components summarized in Table 8.3. The ESD pulse equivalent circuit
contains two precharged capacitors: C1 with a full precharge and C2 with 60% of
the precharge level from the original C1 precharge.
Table 8.3 Parameters of the equivalent circuit for HMM pulse circuits (Fig. 8.10)
R1 210 Defines initial current peak

C1 120 pF Capacitance with full precharge
L1 3 μH Defines quasi-static current level
R2 120 Defines quasi-static current level
C2 30 pF Capacitance with 60% level of original precharge
L2 100 nH Defines rise time
CB 30 pF Board capacitance
Such an equivalent circuit adequately correlates with both the specified standard
parameters for IEC61000-4-2 waveforms mentioned above and the experimental
data (Fig. 8.11).
8.2.2 HMM-HBM Component Correlation
Using the advanced wafer-level setup [143], typical ESD protection structures have
been characterized in order to demonstrate a correlation between device responses
to system-level ESD stress and to component-level ESD stress. For component-level
stress, HBM pulse results have been taken for comparison. It has been found that
the correlation factor strongly depends on the type of device.
The available commercial voltage probes are not capable of measuring high
voltages with a high bandwidth. A new type of voltage probe was developed for
this purpose. This voltage probe allows the measurement of an about 2000 V
peak voltage with the bandwidth required for HMM measurements and enables the
measurement of the turn-on behavior of an ESD protection device under HMM
stress.
a) 4.0
3.5
Measurement
Current [A] 3.0 Simulation
2.5
2.0
1.5
1.0
0.5
0.0
–1.0E-08 1.0E-08 3.0E-08 5.0E-08 7.0E-08 9.0E-08
Time [s]
b)
4
3.5
3
Current [A]
2.5
2
1.5
1
0.5
0
0 10 20 30 40 50 60 70 80
Time [ns]
Fig. 8.11 HMM simulation for short-circuit conditions of 1 kV HMM pulse compared to mea-
sured HMM current waveform from the setup with commercial gun (Fig. 8.8a) (a) and advanced
setup with Hanwa HMM module (Fig. 8.8b) (b)
Pass–fail measurements have been carried out on typical ESD protection struc-
tures, such as diodes, silicon-controlled rectifiers, and high-voltage ESD clamps,
to study the capabilities of the HMM on-wafer measurement setup. To enable a
comparison of the various methods, fresh samples of the same device have been
characterized with the commercial HBM on-wafer HANWA HED-W5000M tester.
Only the waveforms of the maximum precharge level before device failure dur-
ing HBM and HMM stress are compared below. These levels are equal to the
maximum stress a device can withstand without being damaged or becoming non-
functional. Thus, this methodology guarantees the comparability of measurement
results when the correlation between the stress levels of two ESD testing methods is
unknown.
To obtain the correlation factor, the precharge voltages obtained at device fail-
ure during HBM and HMM testing are divided. Voltage waveforms have been
captured to study the transient device behavior under different stress conditions.
To remove the additional voltage drops due to parasitic elements in measure-
ment setups, calibration methodology [144] is applied to the on-wafer HMM
setup.
8.2.2.1 Diodes Under HMM Stress

For diodes, the current waveform and DC characteristic are taken from devices
manufactured in a 90 nm CMOS process technology.The diodes have a width of
50 μm with a diffusion isolation space of 0.5 μm.
The functional leakage test between the HMM pulses of elevated amplitude and
corresponding current waveform at the step before failure is presented in Fig. 8.12.
The results demonstrate a correlation factor of 3.7 between HBM and HMM testing.
In HMM, the device can withstand a maximum current of about 7.8 A, while the
maximum current in HBM is about 4.9 A (Fig. 8.12b).
8 8
Pre-charge voltage [kV]
7 HBM 7400V
6 HBM 6
Current [A]
HMM 5
4 4
3
2 2
1
0 0
10–11 10–10 10–9 10–8 10–7 10–6 10–5 0 20 40 60 80
Leakage Current [mA] a) Time [ns] b)
Fig. 8.12 Leakage evolution at negative 0.2 V bias during HBM and HMM tests (a) and overlay
of HBM and HMM current waveforms obtained from a diode right before device failure (b)
8.2.2.2 LVTSCR
A different correlation has been obtained from test results of LVTSCR devices man-
ufactured in the same 90 nm CMOS process. The devices have a width of 50 μm
and a gate length L=0.25 μm. In this case, a correlation factor of ∼5 between HBM
and HMM testing has been measured.
The overlay of HBM and HMM current waveforms just prior to device failure
(Fig. 8.13b) shows that a particular device can withstand a maximum current of only
∼3 A in both cases. Unlike for the diode, the maximum thermal stress during HBM
and HMM is different for the LVTSCR. The device is less robust under HMM stress
due to the much faster rise time of the HMM pulse, which causes higher overshoots
to occur during device turn-on. One of the possible explanations for this effect is
4
5
HBM 4500V
4 3 HMM 900V
Current [A]
HBM
3 HMM 2
2
1
1
0 0
10–7 10–6 10–5 10–4 10–3 10–2 10–1 100 0 20 40 60 80
Fig. 8.13 Leakage evolution at 0.6 V bias during HBM and HMM testing (a) and overlay of HBM
and HMM current (b) in LVTSCR devices
that the higher overshoots due to the first peak cause a failure in the device, which
is already at lower current levels in HMM than in HBM.
8.2.2.3 High-Voltage ESD Clamps

A different correlation factor has been observed in high-voltage ESD clamps
manufactured in a 50 V BCD technology and stressed according to the same
methodology as low-voltage devices.
The measured correlation factor between HBM and HMM is about 1.5. This
factor is significantly different from that of the measured low-voltage CMOS
diodes. A comparison of the current waveforms captured just prior to device fail-
ure (Fig. 8.14b) reveals more differences. During the first 70 ns, the HMM current
is significantly higher than the HBM current. After 70 ns, the HMM current falls
below the level of the HBM current. In HMM, the device can withstand a maximum
current of more than 6 A (initial peak), while the maximum current in HBM is only
about 1.7 A, resulting in a higher robustness during HMM testing.
4
7
6 HBM 2900V
3 HMM 1900V
5
Current [A]
2 4
3
1 HBM
2
HMM
1
0 0
10–7 10–6 10–5 10–4 10–3 10–2 10–1 0 20 40 60 80
Fig. 8.14 Leakage evolution at 10 V bias during HBM and HMM testing (a) and overlay of HBM
and HMM current (b) obtained from a lateral PNP device at the step prior to failure
100
80
HBM
Voltage [V] 60
40
HMM
20
0
0 500 1000 1500 2000
Time [ns]
Fig. 8.15 Overlay of HBM and HMM current waveform obtained from a lateral PNP device right
before device failure, stress level: HBM: 2500 V, HMM: 1800 V
This significant difference is due to the different behaviour of the device under
different stress conditions. Figure 8.15 shows the voltage waveforms in the device
during HBM and HMM testing. After turn-on, the voltage drop on the device is
about 25 V during HMM stress and about 60 V during HBM stress, indicating dif-
ferent conduction mechanisms during both stress types. This creates less generated
heat during HMM stress, thus making the device more robust, in comparison with
HBM.
Thus, the correlation between system and component-level pulses is not univer-
sal and depends on device type and process technology. However, on-wafer HMM
measurement setup can provide rather reliable information for the expected perfor-
mance of system-level on-chip protection by recreating a similar stress environment
to one that would be tested with an ESD gun as the pulse source.
8.3 On-Chip Design for System-Level Pins
8.3.1 Examples of Circuits with System-Level Protection

Historically, the first, most common products with system pins are products with
the RS-232 interface standard. This standard has been designed to handle commu-
nication between two devices for short-distance (max 50–100 ft). Other examples
include dual transceivers released to meet class 4 EN-61000-4-2; miniature electret
microphone; high-speed bi-directional level shifters, ultra-low dropout voltage reg-
ulators; power management units, and display drivers, numerous automotive [153]
and medical applications [154].
An example of the system-level ESD protection solution with EMI suppression is
presented in Fig. 8.16. The circuit includes high-level ESD protection accomplished
8.3 On-Chip Design for System-Level Pins 417
a)
50 Ohm VDD_Internal
VDD
VDD 2 Ohm
D D b)
G B G B
DDESD S S
ESDM
ESD diode
Two 5V NMOS Snapback Clamps
Fig. 8.16 EMI filtering with system-level ESD protection. Block diagram for EMI filter (a) and
ESD protection network with dual-direction device DDESD (b)
with a dual-direction ESD device. The dual-direction functionality is required to

avoid generation of EMI harmonics. After the EMI filter components, the second
stage of ESD protection includes a 50- separation resistor and an NMOS snapback
clamp.
Another example of on-chip system-level protection requirement is for the
input/output pins to be tolerant of a much higher voltage than what is usually applied
to circuits fabricated in a low-voltage BiCMOS process. These cases can be found
in a variety of analog applications for displays, automotive, and communications
products.
In this case, it is desirable to provide ESD protection for circuits with a high-
voltage tolerance in a low-voltage BiCMOS process without adding additional
process steps to avoid increasing the cost.
One example is presented in Fig. 8.17. The microchip is fabricated in a 5 V
BiCMOS process. The chip is intended to calculate the current through the induc-
tive load by measuring a small differential voltage across the shunt resistor RS . The
common-mode voltage on RS can be very high when the power FET is on. The volt-
age drop across the input resistors RI1 and RI2 shields the 5 V active components in
the differential amplifier (DA) from this high-input, common-mode voltage. When
the power FET is off, the current through the load returns through the freewheel-
ing diode (D). This results in a negative-input, common-mode voltage on the shunt
High Voltage
PowerFET
+5V
Integrated circuit
+in
RS
RI2
out
DA
–in
RI1
gnd
D
Fig. 8.17 Current measurement circuit in the application with an inductive load using a differential
amplifier (DA) and on-chip resistors RI1 and RI2
resistor. As a result, the full, required voltage range for the input pins is –5 V to
+60 V.
Presence of the body diode inside typical BiCMOS ESD protection devices NPN
BJT, BSCR, NMOS, LVTSCR, or active clamps with grounded P-substrate prevents
proper chip performance. For example, if the circuit (Fig. 8.17) the current in the
inductance (motor) circuit is measured by sensing a small differential voltage on the
small shunt resistor RS . High current through the body diode will create a large volt-
age drop on the circuit interconnects, making accurate circuit operation impossible.
As a result, one of the basic devices for dual-direction operation should be similar
to a diode AC switch (DIAC) [14]. The BiCMOS version of this dual-direction ESD
protection device has been proposed in [150] and recently studied further in [42,
151, 152].
To improve voltage tolerance of ESD devices beyond process capabilities, a new
method for increasing the breakdown voltage is proposed in [151], and is motivated
by the goal of keeping the ESD devices as free structures, i.e. structures that can be
realized using only layout changes. This method is based on the use of a “spotted”
NISO blocking mask to create a corresponding spotted implant pattern. It is impor-
tant that the NISO region is formed early in the process, before the NEPI deposition,
and followed by the high-temperature anneals.
The total anneal produces an NISO diffusion profile of ∼7 μm. As a result, even
for the case of a spotted implant pattern (Fig. 8.18a, b) defined within the minimum
dimension design rules (2 μm), the final doping profile is essentially quasi-uniform
(Fig. 7.18c). This new, “diluted” NISO layer has an average doping that is lower
Pwell Pwell
Nwell
a) b)
10–1
100
10–1
1017
PHOSPHORUS (1/cm3)
CURRENT (A/um) 10–2

10–3
NO NISO
SOLID NISO 10–4
1016 10–5
PERIODICALLY
MASKED NISO
10–6
10–7
10–8
1015 10–9
10–10
10–11
1014 10–12
0 2 4 6 8 10 0 10 20 30 40 50 60
Y(μm)(0--the surface level) VOLTAGE(V)
c) d)
Fig. 8.18 The cross-section for the phosphorus profile after NISO implant (a), the final doping
profile for the NISO region (b, c), and the calculated isothermal I–V characteristics (d) for a PNP
triggered DIAC [42]
than the initial NISO implant (Fig. 8.18d) and can be used to form asymmetrical
blocking junctions with higher breakdown voltages.
According to TCAD analysis, 50% blocking of the NISO implant by using the
2 μm × 2 μm mask creates a spotting pattern that produces approximately half of
the NISO doping level, in comparison with the NISO region that is formed in stan-
dard processes (Fig. 8.18c). This can be used to substantially increase the breakdown
voltage of the DIAC device to above 60 V in experimental design (Fig. 8.19).
If dual-direction voltage tolerance is required for the input pad, only the symmet-
rical device (Fig. 8.18b) can be modified to an asymmetrical bi-directional device
(Fig. 8.20), saving additional space on the chip.
Thus, a voltage tolerance that significantly exceeds that of devices supported by
the process technology can be achieved for system-level solutions.
A generic method for increasing the breakdown voltage consists of creating a
diluted doping layer using a spotted implant mask, followed by anneals that result
in a quasi-uniform doping layer formation. The diluted doping layer can also be
4
STRUCTURE TYPE
2
CURRENT(A)
–2
–4
–40 –20 0 20 40 60
VOLTAGE(V)
Fig. 8.19 Measured TLP characteristics of the reference devices with full NISO layer, diluted
NISO layer, and the cascaded version
NEGATIVE CURRENT (A/um)
100
10–1
10–2
Pwell Pwell 10–3
NEGATIVEPAD BIAS
Nwell 10–4
POSITIVEPAD BIAS
10–5
10–6
10–7
10–8
10–9
10–10
10–11
10–12
0 10 20 30 40
NEGATIVE VOLTAGE (V)
a) b)
Fig. 8.20 The cross-section with the net doping profile and the calculated isothermal I–V
characteristics of the asymmetrical dual-direction DIAC structure
used to obtain blocking junctions with a higher breakdown voltage than available in
the BiCMOS process. The overall value of this technique is in providing a greater
variety of ESD protection devices and leading to improved ESD protection window
targeting for both BiCMOS and submicron CMOS processes with a triple well.
8.3.1.1 On-Chip and Suppressor ESD Device Interaction

The case of on-chip and discrete system-level protection interaction is illustrated in
Fig. 8.21 [139]. An IC pin with normal ESD protection is protected by an additional
PCB suppressor diode. Respectively, the majority of current during the ESD event
is expected to flow through the PCB diode. The peculiarity of such an approach
is related to the comparison of the characteristics of the PCB diode and the ESD
protection on the chip. When the voltage drop on the PCB diode is lower than on
the IC protection, the expected current path is valid. However, if the critical for
failure voltage drop on the internal IC pin is lower than the level provided by the
PCB diode, the internal ESD circuit can fail even if the PCB ESD diode itself can
withstand ESD pulse.
Fig. 8.21c illustrates the possible effect of bus resistance, which may result in an
unexpected burnout of the bus unless it is engineered to conduct system-level ESD
current.
a) b)
c)
Fig. 8.21 Combination of the on-chip and system-level PCB protection [139]
With the on-chip system-level protection design, there is an area and performance
trade-off. The performance of the end system is the real target. A learning cycle may
be required to correctly target the “ESD window.” Component and system tests must
be carefully specified in advance. Additional external components may be required
to prevent a latch-up event.
8.4 Hot Swap and Hot Plug-in

The problem of power sequencing and ground detection occurs in many systems.
For example, mobile systems using a battery supply have a power up sequence that
is limited by the input capacitance and the high resistance of the battery. Hot swap
controllers provide limiting of inrush current and short-circuit protection during
operation.
A particular aspect of system-level ESD protection problem is related to a hot
swap and plug-in requirements. It is usually assumed that hot swap and plug-in
events are performed in a condition that is at least partly EPA compliant.
The high holding voltage lateral PNP device has been previously experimen-
tally discussed in Chapter 3 of this book. The disadvantage of such a device is its
relatively large footprint.
As it has been mentioned above, the most reliable way to guarantee system stabil-
ity against transient latch-up is to require a minimum on-state holding voltage above
the power supply voltage level. This requirement also maintains system stability
against short-term electrical overstress (EOS) events above absolute maximum volt-
age limits. Alternatively, the ESD protection clamp or network may be designed to
guarantee both critical snapback and the minimum holding current levels above the
current level that can be provided by the power supply at normal operation and test
conditions. In this case the conditions for latch-up cannot be realized in principle.
A new compact PNP-SCR device-level solution utilizing a two-stage snapback
approach is proposed in [49].
8.4.1 The Concept of Two-Stage SCR ESD Devices

In some hot plug-in and hot swap applications, as well as in some applications
where system-level protection is required, the EOS and power supply current can
be rather limited. Therefore, rather than define a minimum specified holding volt-
age for robustness against transient-induced latch-up, a minimum snapback current,
for example, 1 A, may be chosen above the range of the power supply current.
Below this current, the device will be robust against latch-up due to the high power
required to induce snapback.
A consequence of the high holding voltage of a lateral PNP device is a
required large footprint (1000 μm total width) due to its high on-state resistance
in the conductivity modulation mode. This contrasts with the low-holding-voltage
8.4 Hot Swap and Hot Plug-in 423
CATHODE ANODE
C E B G
p+ p+ n+ B S P D
p+ n+ p+ n+
Pwell Nwell Pwell
Nwell
P-substrate P-substrate
a) b)
c) d)
Fig. 8.22 Simplified cross-sections of lateral PNP (a) and DeMOS-SCR (b) devices, and the
experimental TLP I–V characteristics (c) and (d), respectively
DeMOS-SCR device, which can be implemented in a smaller footprint. The lat-

eral PNP also produces a bell-shaped voltage waveform during ESD stress that can
challenge internal circuit protection.
Both the lateral PNP and DeMOS-SCR devices presented in Fig. 8.22 have ben-
eficial aspects; therefore, it is desirable to design a two-stage device with the high
holding voltage characteristic of a PNP at low current, and the low on-resistance
characteristic of an SCR at high current.
To achieve such a two-stage operation in a practical manner, an additional
isolation-limited n+ emitter can be embedded in the p-collector region of the lateral
PNP device. This can be achieved by modification of original device topology
(Fig. 8.23a) by enclosing the additional n+ emitter diffusion in the p+ collector dif-
fusion (Fig. 8.23b) or by creating interdigitated n+ emitter regions (Fig. 8.23c). In
both cases, the triggering current at which the device switches from PNP current
modulation mode to SCR mode can be controlled by the n+ emitter topology. In the
topology of Fig. 5b, an increase in the enclosure parameter LE should increase this
triggering current. Similarly, in the topology of 5b, varying the number and relative
lengths of the n+ emitters will control the triggering current.
The cross-section of the device (Fig. 8.23a) is already shown in Fig. 8.22a. The
cross-section for the topology (Fig. 8.23b) corresponds to the process simulation
cross-section (Fig. 8.24a). The cross-section for the topology in Fig. 8.23c is roughly
Pwell Nwell
Pwell Nwell Pwell Nwell
LE p+
Interdigitated Region
n+
p+
P+-Collector
P+-Collector
P+-Emitter
P+-Emitter
P+-Emitter
n+
p+
n+
n+
n+ n+ p+ n+
Emitter
Base Base Base
a) b) c)
HV PAD
RNW
pnp
npn
RPW
GND
d)
Fig. 8.23 Simplified original topology for the high-voltage lateral PNP ESD device (a) and imple-
mentation of the two-stage PNP-SCR devices using enclosed (b) and interdigitated (c) n+ -emitter;
the equivalent circuit of the device (d)
a superposition of the two cross-sections referred to above, depending on the cutline

position.
The equivalent circuit of the new device is still the same as for any SCR structure
(Fig. 8.23d). For example, it is similar to the cross-section of the NLDMOS-SCR if
the polygate region is excluded. However, the major difference is in the selection of
the parasitic n–p–n and p–n–p structures that form the SCR and, in particular, the
additional value of the two-stage snapback capability.
In the case of the NLDMOS-SCR, the triggering in the first stage is controlled by
the parasitic n–p–n structure that has an already rather low holding voltage. At the
given load characteristics during the device turn-on, the critical current is sufficient
to initiate double-injection conductivity modulation from the n+ - and p+ -emitter
junctions.
In the case of one of the topologies properly implemented for the two-stage
SCR (Fig. 8.23a–c), the lateral p–n–p is the embedded component that first controls
a) b)
2.5
Pad P2, 0:1 nEmitter-Collector Ratio
2
Pulsed Current [A]

1.5
0.5
0
0 10 20 30 40 50 60
Pulsed Voltage [V]
c)
Fig. 8.24 Two-stage PNP-SCR device cross-section (a), with simulated I–V characteristics (b),
and experimental dependence of TLP characteristics on n+ /p+ interdigitation composition (c)
operation in avalanche–injection conditions. Until the current density in the struc-

ture becomes sufficient, the n+ -emitter junction is closed and device operation is
similar to that of the lateral PNP device. The potential of the emitter junction rises
over ∼0.8 V and the injection of electrons becomes possible only at a certain crit-
ical current, followed by positive feedback for the double-injection conductivity
modulation process from the p+ and n+ emitters similarly to the final state of the
NLDMOS-SCR.
Thus, the sequence of the embedded parasitic pnp and npn components provides
differentiation between the two-stage SCR proposed in this study and the known
NLDMOS-SCR solution.
One of the expected tunable characteristics of the device is the critical triggering
current in SCR mode. The higher the isolation of n+ -emitter, the higher the trig-
gering current expected. The isolation level of the n+ -emitter is a layout-dependent
figure of merit. For example, in layout topology (Fig. 8.23b), larger spacing LE
causes a higher isolation level. A similar statement is true for the reduction of
the total amount of interdigitated minimum dimension n+ -emitter islands in the
topology (Fig. 8.23c). Thus, an important degree of freedom is provided on the
topology level to control the desired current and voltage before snapback into SCR
conductivity modulation mode.
The suggested concept of the two-stage PNP-SCR clamp has been validated by
physical process and device (TCAD) simulation (Fig. 8.24a, b), with successful
experimental implementation in a 5 V 0.18 μm DGO CMOS process with 40 V
extended drain devices (Fig. 8.24c).
A cross-section of the simulated device is presented in Fig. 8.24a. The structure
is similar to the lateral PNP in the base and emitter region (at left), with the n+
emitter embedded and isolated by the p+ collector region (at right), to form an SCR
device. The level of n+ emitter isolation depends on the device topology and can be
simulated. Greater n+ emitter isolation results in a higher triggering current between
the PNP and SCR stages (Fig. 8.24b).
In [49], experimental devices have been implemented using the interdigitated
approach of Fig. 8.23c. The interdigitated regions of the n+ emitter and p+ collector
form one stripe of the composite region. A decrease in the triggering current and
holding voltage with an increasing n+ collector to p+ ratio observed in the TLP
measurements of Fig. 8.23c is consistent with the simulated results of Fig. 8.24b
At the same time, all devices provided an on-state current suitable for system-
level tests.
A mixed-mode simulation has been performed in order to demonstrate how the
new solution will perform in power surge electrical overstress (EOS) events and in
HBM operation. The output characteristics of the devices used in the simulation are
presented in Fig. 8.25a. Two mixed-mode circuits used for validation are presented
in Fig. 8.25b, c.
In the EOS event, a DC voltage source of VIN = 35 V is applied to the two-stage
devices A and B with characteristics presented in Fig. 8.25d. The EOS pulse is
produced by the current source IEOS in a time domain close to fast ESD conditions.
According to the transient characteristics with two-stage device “B” (with a low
triggering current) remains in the low holding voltage latch-up state, while device
“A” (with a much higher triggering current) returns to low leakage state (Fig. 8.25d).
The 8 kV HBM operation of both devices provides rather similar waveforms
suitable for high-level ESD protection (Fig. 8.25e).
Thus, for hot swap and hot plug-in requirements, use of the two-stage HV PNP
device with an embedded parasitic SCR structure is a rather promising solution. It
enables design parameters that are controllable by layout for a two-stage operation
in the PNP and SCR conductivity modulation modes.
Practical implementation of this new PNP-SCR clamp is achieved by simple
topological control of the first-stage triggering current to a desired level.
10–1
ESD DEVICE CURRENT (mA/μm)

DEVICE
10–2
10–3 A
–4
10
10–5
10–6
10–7
10–8
0 10 20 30 40 50
CLAMP VOLTAGE (V)
a)
RH LT HV PAD
HV PAD
RNW
RNW
Two stage pnp

pnp CT
SCR RL
npn
npn IEOS CP
CH
RPW
RPW VIN
GND
GND HBM Pulse Circuit Two-stage SCR
b) c)
DEVICE 60 DEVICE
40
CLAMP VOLTAGE (V)
CLAMP VOLTAGE (V)
A 50 A
30 40
30
20
20
10
10
0 0
100 101 102 103 100 101
TIME (ns) TIME (ns)
6
CURRENT (A)
CURRENT (A)
101
4
100
10 –1 2
–2
10 0
100 101 102 103 100 101
TIME (ns) TIME (ns)
d) e)
Fig. 8.25 Simulated DC I–V characteristics for two two-stage SCR’s with different critical cur-
rents for SCR turn-on (a). Mixed-mode simulation circuits used to validate the hot plug-in
solution in case of electrical overstress (b) and HBM ESD pulse (c); corresponding transient pulse
waveforms (d) and (e)
In contrast to an SCR device, this solution is inherently insensitive to electri-

cal overstress upto a desired level, as long as the overstress current is below the
triggering current for transition into the SCR mode.
8.5 System-on-Package (SOP) Protection
To optimize the cost of the integrated components, a system-on-package (SOP)

approach is used. With (SOP), the final packaged device is composed of two or more
dies bonded in the same package. A typical example of the SOP is the integration of
switching voltage regulators or power arrays in one package with digital blocks to
create complex power management units (PMU’s). A PMU essentially combines a
co-processor unit made on expensive silicon with an inexpensive power module to
support high-density digital design using 0.45–0.13 μm CMOS processes, for exam-
ple, in the 0.5 μm BCD process. Similarly, because of simple cost considerations,
the modules may include a combination of dies from high-voltage processes where
the dense core logic devices are simply unavailable. A stacked die SOP example is
External package Bottom die Top die

bond wires
Inter die
bond
wires
Fig. 8.26 Illustration of the shared interdie ESD protection
Since the ESD protection specification is defined for packaged parts, SOP design
can be advantageous. First of all, the die pins not connected to the lead frame of the
package do not need any protection at all.
An advantage can also be taken from the external pins of such an SOP. In partic-
ular, some of the more expensive external die pins can be protected either fully or at
the first stage by the ESD network of the less expensive die.
Similarly, the least expensive silicon die can protect system-level pins, or
additional dies with a TVS function can be introduced into the package.
8.6 ESD Robustness of Discrete Components 429
8.6 ESD Robustness of Discrete Components
8.6.1 Discrete Components in High Reliability Systems

Electronic components form an integral and critical part of on-board systems and are
present in vehicle controls, communications, navigation, radar systems, etc. Future
aircraft systems, including electric aircrafts and Unmanned Aerial Vehicles, or the
Next Generation Air Transportation Systems, will certainly rely more on electric and
electronic subsystems and components. This new application will also result in the
increase of the number of electronic faults with perhaps unanticipated fault modes.
In addition, the move toward lead-free electronics and Micro-Electro-Mechanical
Systems will further result in unanticipated anomalous behavior.
To improve the reliability of systems such as aircraft, assure adequate in-flight
performance, and reduce maintenance costs, it is therefore imperative to provide
system health awareness for the electronic components. To that end, an understand-
ing of the behavior of deteriorated components is needed, as well as the capability to
anticipate failures and predict the remaining life of the embedded electronics. The
development and advancement of this capability is also relevant to multiple NASA
exploration systems and long-endurance robotic space missions.
Generally, an understanding of the intrinsic and extrinsic degradation mech-
anisms of discrete devices is crucial for the adoption and application of health
management to systems. Within the field of electronics, knowledge of semicon-
ductor degradation under various system and environmental scenarios may be
coupled with prognostic algorithms to predict the future state and time-to-failure
of semiconductor components.
The existence of measurable extrinsic degradation precursors, pertaining to
device packaging, has been well established in literature concerning power tran-
sistor devices. Intrinsic degradation precursors related to the physical properties of
the semiconductor materials have also been observed. However, it is not widely
known how degradation propagates as a function of the environmental conditions
and various stressors, in particular, ESD stress. The attainment of such knowledge
is critical for advancement in the field of power electronics health management and
prognostics.
8.6.2 ESD Requirement for Discrete Components

ESD and EOS stress are major sources of damage and failure of discrete electronic
devices, especially field effect transistors and IGBTs, which are extremely sensitive
to such stress. In the case of low-voltage (<100 V) and low-power devices (<1 W),
even package-level ESD events usually result in an immediate irreversible burnout.
This event is easy to detect by a simple electrical parameter test. For high-power
MOSFET and IGBT devices, the scenario is different. As shown below, at some pin
combination, these devices may indeed withstand not only the standard package-
level stress but also the system level. However, even in this case, ESD robustness of
the control electrode is negligible and requires an ESD protected area environment
to guarantee the absence of handling damage.
Independent from the levels of ESD pulse that can cause local damage of the
power device, another question is the final impact of the ESD events on the long-
term reliability parameters of the stressed device. Depending on the nature of the
induced defect, an accelerated degradation and corresponding reduction of the time
before failure can be expected, due to the change of electric field distribution, current
density, and lattice temperature in the location of the defect.
At the same time, ESD protection requirements are originally defined for inte-
grated circuit components. A ESD specification of the most commercially available
discrete components, for example, IGBT and MOSFET, cannot be found even for
the control electrode.
Often, an intuitive assumption about ESD robustness of a discrete component is
based upon the power and high-voltage parameters of the component in comparison
with the ESD pulse. For example, with a component current level over 10 A, which
significantly exceeds the package-level ESD current or the operational voltage of
the component in a kilovolt range, the expectation is rather high.
To address the ESD robustness of a discrete component, test results for a discrete
600 V 16 A Insulated Gate Bipolar Transistor (IGBTs) device with an integrated
flyback diode are presented in this section. It is demonstrated in [155] that ESD
robustness depends on the terminal of this device. In the case of collector stress, the
device can indeed easily outperform the ESD level up to a 16 kV ISO discharge.
However, if the gate of the device is stressed, the device can only pass the 5 kV
HBM test. Thus, the robustness of discrete power components is rather device and
terminal specific.
It is demonstrated that ESD pulse stresses may create latent local damages that
act similar to defects in the structure of the power device. Such stress-induced
defects may not result in immediate parametric failure due to the change of the
measured parameters from the datasheet range. While some deviation of the param-
eters can be observed within the datasheet limits, the long-term reliability of the
component is significantly impacted. Thus, stress-induced defects do not result in
formal parametric failures per datasheet specifications, but can be detected by the
substantial change in the electrical characteristics when compared with the original
device parameters measured prior to ESD stress.
8.6.2.1 ESD Effects on Power Transistors

Some results obtained by system-level stress tools are presented below for the
commercially available 600 V 16 A IGBTs [155].
In power transistors such as power MOSFETs, the gate oxide is extremely sus-
ceptible to ESD events. Gate oxide stress above the critical voltage can cause
structural damage to the gate oxide. This damage leads to leakage current levels
beyond the datasheet limits. Alternatively, it could also cause a moderate change in
the device structure that may result in failure during future operation.
Thus, ESD damages are categorized as two types: (a) catastrophic and (b) latent.
Devices with catastrophic damage are those that have been instantly destroyed at
the time of the ESD event and therefore will not pass the functional tests that are
part of the manufacturing or repair test process.
Latent damage occurs in devices when the ESD degrades the device, but not to
the point of destruction. In this case, the changes in the power device structure are
local and deviations of characteristics due to changes in the local region are hard to
detect. Therefore, these devices could generally pass the functional tests, especially
after only a single pulse ESD event. However, the reliability of components with
latent damage is in question. One of the major resulting expectations is failure due
to further ESD events or other forms of electrical overstress (EOS). Another possible
impact is on the safe operating area, thereby reducing device robustness in standard
application regimes.
The most common result of gate oxide overstress due to ESD events is the
breakdown of gate dielectric material.
Usually, immediate failure does not occur until the gate-to-source voltage
exceeds the rated maximum by an amount that may be in the range of two to three
times the rated maximum.
During the manufacturing of the device, it can be assumed that structural defects
are well controlled. Nonetheless, a small probability of a priori defect remains [156,
157, 158]. For discrete power device arrays, the detection of these defects is hard.
An acceleration of the degradation processes can be expected for devices with prior
defects that manifest themselves in an increased electric field, current density, and
Joule heat generation. In contrast, defects that are induced during the operation of
the device are hard to predict.
8.6.2.2 Statistical Approach for ESD and Reliability Parameters Verification

Integrated electronic components are required to pass standard ESD tests. Long-
term reliability parameters (for example, the middle time before failure) of the
components are verified by statistical methods on a limited set of components.
Likewise, safe operating area (SOA) parameters are confirmed only on a limited
amount of samples.
Since most of the above testing methodologies involve subjecting the device to
extreme electrical overstress events, it is advised that the components subjected to
the tests are no longer used in the field, even if they pass the tests. The intuitive justi-
fication for this is the possible impact of the test on long-term reliability parameters.
For example, if the component has been tested for pulsed SOA or zapped with an
ESD tester, then it incurs significant damage. In the case of power devices, the dam-
age might often be latent and can be detected by measurements of some secondary
parameters that produce different initial I–V characteristics. Thus, the stressed com-
ponent may pass the parametric test within the datasheet specification limits while
harboring latent damage.
Thus, the verification of ESD robustness and reliability parameters is provided

by statistical methods applied to similar devices, while only nominal datasheet
electrical parameters are verified for the devices shipped to customers.
To the best of our knowledge, no technology exists that would enable a repeated
diagnostic of such electrical damages, since the impact of a sequence of diagnostic
tests on the operational lifespan and long-term reliability of the device is not yet
well understood. Only a statistical sampling of a group of devices can currently be
done using existing technologies.
It is logical to assume that if during application or assembly a discrete device
will experience an ESD event above a certain critical level, which will form induced
latent damage in the device structure, then the device’s long-term operation and
individual time to failure may be impacted. In particular, the individual time to fail-
ure will be significantly different from the value measured by statistical methods on
non-stressed devices.
Formation of the latent defects in power components is discussed below using
numerical simulation analysis and results of ESD stress experiments.
8.6.3 Preliminary Numerical Analysis for Devices with Defects

and the Two-Transistor Model
A numerical simulation model for the device has been used to demonstrate an effect
of latent defect formation [155]. To represent the latent damage of the device, a
superposition of two device structures with different width scaling factors has been
used. The first device has a large width scaling factor and represents the unchanged
part of the array. The second device, with a small scaling factor, had some deviation
of the original parameters. The characteristics of the two-transistor mixed-mode
circuit (Fig. 8.27) have been used as a physical representation of the power device
array with latent damage or a local process technology defect. The last case assumes
that the deviation of the device parameters is due to process technology defects.
Fig. 8.27 Physical representation of the device with local defects using a two-transistor mixed-
mode circuit
The physical model (Fig. 8.27) enables a more comprehensive understanding of

the damage caused by ESD and is further used for comparison with experimental
data in the following section.
For simplification of the numerical analysis, an isothermal case is considered,
thereby ignoring the local temperature rise during ESD stress. The upper and lower
part of the 2D device cross-section is shown in Fig. 8.28a, b. It displays architecture
typical for a vertical gate (trench gate) IGBT cell. The upper region (Fig. 8.28a) of
the device has a shallow source and body n+ and p+ diffusion regions, along with the
second p-body diffusion that forms the well of the vertical N-channel MOS device.
The poly-silicon gate is separated from the silicon by a thin gate oxide. The right
side of the upper cross-section (Fig. 8.28a) represents the active device, while the
left part represents the lateral termination of the array. Figure 8.28b shows the lower
part of the device, which forms the heavily doped p++ collector region separated
from Nepi region by the n-buffer.
0 45
n+ p+
46
1
47
PBODY
2
48
3 N-BUFFER 49
P-COLLECTOR
50
1 1.5 2.0 2.5 3.0 3.5 0 1 2 3 4
X(μm) X(μm)
a) b)
Fig. 8.28 Simplified simulation cross-section for the (a) upper and (b) lower part of the IGBT
device used for numerical analysis of the experimental test data
The principle of device operation at switching is based upon the formation of

an accumulation channel in the MOS structure at positive gate bias, which pro-
vides electron current through the epi region followed by hole injection from the
p-collector.
In a more physically correct way, the p-collector should be called the emitter.
Historically, the current terminology came in place because IGBT devices have been
used as replacements for vertical double-diffusion MOSFET devices, where the n+ -
drain region that connects the epi-layer at the bottom represents the collector of the
parasitic NPN structure.
The controlled conductivity modulation of the Nepi region provides a rather high-
current level above the typical DMOS levels, making the IGBT device an important
component of power electronics.
The switching operation of the numerical IGBT model (Fig. 8.29) delivers the
expected switching behavior of IGBT devices.
200
150
VOLTAGE (V)
100
VCE
50
VGE
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
TIME (μs)
5
CURRENT (A)
4
IG
3
IC
2
1
0
0.0 0.5 1.0 1.5 2.0 2.5 3.0
TIME (μs)
Fig. 8.29 Switching characteristics of the standard devices
Output characteristics for devices with different epi doping levels at a 5 V collec-
tor voltage are shown in Fig. 8.30c. They demonstrate a threshold voltage of ∼0.9 V
and a saturation current iC of ∼5e–5 A.
The transconductance characteristics of the device for different PBODY charac-
teristic lengths are shown in Fig. 8.30b and demonstrate the effect of the change of
the threshold voltage upon this parameter.
The two-transistor model (Fig. 8.27) of the device with a latent defect can
be represented by a superposition of a cross-section with original parameter
XPBODY = 0.1 μm and a cross-section with parameter XPBODY = 0.075 μm
(Fig. 8.30b). Such defect can be artificially formed as a result of ESD event. To
compose a physical equivalent of the defective device, the two-transistor model
incorporates components with different scaling factors.
The change of the transconductance characteristic is presented in Fig. 8.32.
Depending on the width scaling factor for both cross-sections, a correspond-
ing peculiarity will be observed on the two-transistor model transconductance
characteristics.
If such a defect is formed, the local region with XPBODY = 0.075 μm will remain
in the on-state when the major part of the array is in the off state, thus providing
conditions for an accelerated degradation process. The mixed-mode simulation case
for a defective device can be represented by a cross-section with a width scaling
GATE BIAS (V)

10–3

5V
10–4 3V
10–5
10–6
10–7
10–8
10–9
10–10
10–11 0V
10–12
10–13
0 50 100 150 200 250 300
COLLECTOR-EMITTER VOLTAGE (V) a)
0.020
Peak Profile XC parameter
0.015
0.1(default)
0.02
0.05
0.010
0.005
VCE = 5V
0.000
0 1 2 3 4 5
GATE-EMITTER VOLTAGE (V) b)
N-epi uniform doping

0.030 VCE = 5V
0.020 5x1014cm–3(default)
0.010
0.000
0 1 2 3 4 5
GATE-EMITTER VOLTAGE (V) c)
Fig. 8.30 Output IC –VCE characteristics at different gate bias of the initial device (a) and change of
the original transconductance IC –VGE characteristics upon different characteristic PBODY profile
lengths (b) and N-epi doping levels (c)
factor wi , and an additional cross-section with the parameter deviation for defective
devices wd .
Similarly, change in other parameters of the structure results in change of the
simulated characteristics. For example, change in the N-epi region (Fig. 8.30c)
provides no change in the threshold voltage, while the saturation current is signifi-
cantly changed. Deviation due to either process technology defects or ESD damage
of the local doping levels only from 5 × 1014 cm–3 to 1 × 1015 cm–3 results in
corresponding change of the current.
Another critical impact of local defects on device reliability is the change of
the safe operating area (SOA). Output IC –VCE characteristics of devices with peak
doping level change in the n+ -buffer region are presented in Fig. 8.31. When the
original peak doping level of 1 × 1019 cm–3 in the buffer region is changed to
1 × 1018 cm–3 , a significant reduction of critical voltages in the whole range of gate
bias is observed (Fig. 8.31), compared to the characteristics of the original device
(Fig. 8.30a). In the case of the two-transistor model for defective devices, the small
width scaling factor of the structure will determine the limitations of the whole
device SOA.
10–3 GATE BIAS (V)

5V
10–4 3V
10–5
10–6
10–7
10–8
10–9
10–10
10–11 0V
10–12
10–13
0 50 100 150 200 250 300
Fig. 8.31 Output IC –VCE characteristics of devices with N-epi doping deviation
The effects of local faults, i.e., localized N-epi doping deviations, are obtained
through a mixed-mode simulation of the two-transistor model. An example of the
change in output and transconductance characteristics for an original device with a
PBODY defect is presented in Fig. 8.32.
8.6.4 Experimental Evaluation of Discrete Components

Robustness
The device chosen for experimental stress testing is the IRG4BC30KD. This device
has a typical structure of a vertical gate (trench gate) IGBT cell (Fig. 8.33), with a
0.0020
Peak Profile XC parameter

0.0018
VCE=5V
0.0016
0.0014
0.0012
0.0010
0.0008
0.0006 0.1(default)
0.0004
0.0002
0.0000
0 2 4
GATE-EMITTER VOLTAGE (V)
Fig. 8.32 Two-transistor model analysis: change in transconductance of 100 μm faulty device
with 1 μm localized PBODY doping deviation
package-integrated flyback diode. The Figure shows two separate dies, one of which
contains the IGBT, while the other forms the protective diode.
8.6.4.1 TLP Stress

At first TLP measurement methodology has been applied to determine the gate
ESD robustness. The gate–emitter TLP characteristics are presented for positive
and negative stress (Fig. 8.34).
During TLP testing, the gate oxide is overstressed, but no change in the leakage
current is observed until the critical current is ∼2–3 A. No gradual changes in the
device structure are observed until this level of current is reached. At the critical
current level, the device fails irreversibly. The catastrophic nature of both positive
and negative TLP gate overstress can be explained by the discharge of the whole
charge accumulated in the gate–emitter MOS capacitor through the local damage
area formed in the dielectric.
The device holds approximately three times the amount of voltage specified by
the maximum limit rating, with a pulsed current that at least satisfies package-level
ESD requirements.
The experimental results of catastrophic failure are repeatable for a 50- TLP
load.
At the same time, the source–drain TLP characteristics in discrete power com-
ponents are rather robust (Fig. 8.34). In the case of the HEXFET, the device has a
rather strong body diode that provides for an avalanche current above the channel
saturation current in pulse conditions (Fig. 8.34). Similar avalanche energy robust-
ness in the IGBT component is achieved by means of an integrated high-voltage
Schottky diode (Fig. 8.35).
Fig. 8.33 De-capped view of the IGBT die
8.6.4.2 ISO System-Level Pulse Test

ISO gun is used in order to overcome TLP system and package-level limitations
during the collector–emitter ESD stress test of the IGBT device. Both modes of
operation of the gun – direct electrical contact discharge and air-gap discharge –
have been used.
For the initial pool of samples, the leakage level observed after ESD zapping was
consistently lower by several orders (0.01 μA vs. 250 μA) than the maximum limit
of the datasheet parameters. The breakdown voltage varied around two different
levels, 650 and 665 V at room temperature, above the datasheet maximum rating
limit of 600 V. No failures were observed below a certain ESD pulse threshold level,
even when up to 100 pulses were applied.
ESD pulse was applied to the collector–emitter junctions. The reverse current
paths in IGBTs are protected by the package-level integrated diode and hence can
withstand a high level of electrical voltage. The diode can withstand stress up to
4 4
GATE BIASING CIRCUIT
GATE-EMITTER
3 3
TLP CURRENT (A)

2 2
1 1
0 0
0 10 20 30 40 50 60 70 80 90 10–1110–1010–9 10–8 10–7 10–6 10–5 10–4 10–3 10–2
TLP VOLTAGE (V) GATELEAK.CURRENT(A)
Fig. 8.34 Positive-pulsed (TLP) gate–emitter characteristics of IGBT devices subjected to gate
ESD overstress
the 30 kV limit of the ISO gun. Most of the exhibited collector–emitter leakage
is at hundreds of microamps to several milliamps, while most of the devices still
had breakdown voltages above 600 V; several devices also showed reduction in
breakdown voltages.
Although the device exceeded the datasheet leakage current limit (∼250 μA) by
a factor of 3, the samples remained fully functional with no additional gate leakage
current. Primarily, two typical modes of fault generation were observed for ESD
pulses above a certain threshold level. These involved localized defect formations
that caused deviation in device parameters but no catastrophic failures. These defect
modes are described in the following section.
In the case of contact zap, the gate–emitter (G–E) ESD pulse combination dam-
aged the device irreversibly and catastrophically. The gate collector zap combination
(G–C) provided parametric failure.
8.6.4.3 Collector–Emitter ISO Test

Similarly, collector–emitter ISO ESD stress above some critical limit results
in either deviation of the previously measured electrical characteristics within
datasheet limits, or parametric failure over the datasheet limits or catastrophic
irreversible failure up to complete damage of device functionality.
Comparison of the electrical data before and after the C–E zap is presented in
Fig. 8.36. As expected, the majority of the samples have leakage-breakdown char-
acteristics that correspond to the formation of a localized region with a current level
of ∼1 mA. This localized defect can form in the vertical emitter–collector junction
25 GATE BIAS (V)

12
20
15
8
10 6
0
0 20 40 60 80 100 120
(a) PULSED DRAIN-SOURCE VOLTAGE (V)
PULSED COLLECTOR CURRENT (A)
GATE BIAS (V)

25 10
8
20 4
15
10
5
0
0
0 100 200 300 400 500 600 700
(b) PULSED COLLECTOR-EMITTER VOLTAGE (V)
Fig. 8.35 Output TLP characteristics for discrete HEXFET with avalanche energy 90 mJ (a) and
16 A 600 V discrete IGBT with integrated Schottky diode (b)
between either the PBODY -N-epi regions, or it can correspond to the damage of the
n-buffer region.
Two typical modes of non-catastrophic parametric failure were observed for
pulses above a certain ESD threshold level.
Characteristics for the first mode of parametric deviation within datasheet limits
are shown in Fig. 8.36a, b. In the first mode, change in the device output collector–
emitter characteristics occurs only after a critical voltage of ∼300 V (Fig. 8.36a)
and, at the same time, a part of the transconductance characteristics represents a
local change in the threshold voltage (Fig. 8.36b).
Combining this experimental data with the numerical simulation analysis of the
two-transistor model, this first type of damage is expected in the CMOS body region,
which changes certain voltage-dependent blocking capabilities of the CMOS part of
the device.
10–9

10–7 10–8
–7
10
10–6 10
–6
–5
–5 10
10
10–4
10
–4 BEFORE STRESS 10–3
–2
10
BEFORE STRESS
10–3 10
–1
0 100 200 300 400 500 600 700 0 1 2 3 4 5 6 7

a) COLLECTOR-EMITTER VOLTAGE (V) b) GATE BIAS (V)
–9
10
–7

10 10
–8
–7
–6 10
10
10–6
–5
10 10–5
BEFORE STRESS
10–4
–4
10 10–3
10–2
–3 BEFORE STRESS
10
0 100 200 300 400 500 600 700 10–1
0 1 2 3 4 5 6 7
c) COLLECTOR-EMITTER VOLTAGE (V)
d) GATE BIAS (V)
Fig. 8.36 Comparison of the collector–emitter output characteristics for first mode of non-
catastrophic failure in the first (a) and second (c) mode of parametric failure and the corresponding
comparison of the transconductance characteristics (b) and (d) for the same device
A different change of output characteristics is observed (Fig. 8.36c, d) for the

second mode of parametric deviation within datasheet limits, where no change in
the threshold voltage (Fig. 8.36d) is accompanied by a change in the leakage current
in the whole voltage range. In this case, a local leakage can be assumed through the
peripheral termination region of the device.
Parametric failures above datasheet limits also result in change of the SOA
characteristics when the critical voltage of the whole device is below 600 V.
8.6.4.4 Gate–Collector ISO Tests

For gate–collector stress, the comparison of electrical test (ET) data before and after
ESD stress leads to the conclusion that the failure signature is completely different
from C–E stress. The ET data have been collected at a ∼1 mA peak current limita-
tion, based upon curve trace data. The output characteristics are shown in Fig. 8.37.
Most of the samples have elevated leakage current due to the current path through
the gate, while the breakdown voltage remains unchanged. This indicates that the
MOS structure, rather than the collector junction, has incurred damage.
The level of change in the collector current, as well as in the transconductance,
indicates more severe damage as compared with the results obtained from ESD zaps
applied to the CE junction.
–1
10

–7
10 10
–2
–3
–6 10
10 –4
10
–5
–5 BEFORE STRESS 10
10
–6
10
–4 –7
10 10
–8
10
10–3 –9 BEFORE STRESS
10
0 100 200 300 400 500 600 700 0 1 2 3 4 5 6 7
a) COLLECTOR-EMITTER VOLTAGE (V) b) GATE BIAS (V)
Fig. 8.37 Change of the (a) output characteristics and (b) transconductance as a result of gate–
collector air-gap stress
Thus, for the case of discrete components in a certain range of ESD pulse mag-
nitude, latent damage is observed prior to catastrophic failure both as a deviation
of the ET parameters within the datasheet limits and as a change of the electri-
cal parameters above the datasheet limits, under the preserved functionality of the
stressed components. When the history of the device is unknown, previously expe-
rienced ESD stress events may result in latent damage that acts as a local defect in
the device. In spite of the full functionality of the device, these damages may result
in the change of SOA or the time to failure of the device in application regimes, due
to the possible acceleration of the degradation processes in the region of the latent
defect.
8.7 Summary
Robust ESD systems have to be insensitive to ESD discharge during operation. This
robustness needs to be realized by the combination of EMC design methods both
on-board and on-chip.
In addition, packaging and handling requires protection against ESD damage.
This damage should be distinguished between events within EPA, where on-chip
protection is sufficient in most cases, and outside EPA, where the IC is already
mounted and embedded in a component.
In effect, positive IC (package) HBM or CDM stress results can be completely
inadequate as criteria for designing a robust ESD system.
A careful characterization of the on-chip and on-board protection is a neces-
sary part of system-level ESD robustness. To achieve an optimum IC design for
robust ESD systems, a co-design should be performed including both chip-level and
board-level protection. Finally, sufficient latch-up robustness even beyond JEDEC
requirements should be provided at the exposed pins, especially for life-critical sys-
tems. At the same time, both overshoot and undershoot at supply lines should be
minimized during ESD.
For a few exposed pins, a significantly increased ESD robustness is required and
can often be provided on the chip using a specially designed ESD cells. On-chip
system-level ESD solutions are based upon the same principles as package-level
ESD protection devices, with specific attention to width scaling and the measures
providing a holding voltage above the power supply level.

Example 8.1 System HMM Pulse Simulation

Library Name: Examples8_System_Level_and_Discrete_Components
Project Name: E8.1_HMM
The example provides an equivalent circuit for human–machine model (HMM) ESD
stress. This stress is considered as an equivalent of the system-level stress. Mixed-
mode examples with NLDMOS-SCR (Fig. E8.1) demonstrate the specific features
of the HMM pulse waveform and can also be used to explore system-level pulsed
operation of different snapback clamps from the examples above.
Example 8.2 HMM Simulation with PCB Components and TVS

Project Name: E8.2_HMM_with_PCB_and_TVSuppressor
The example demonstrates interaction of the parasitic printed circuit board (PCB)
components L2–L4 with the snapback ESD device, resulting in voltage suppression
at the input IC pin. The voltage waveforms observed at different circuit nodes are
presented in Fig. E8.2, demonstrating significant reduction of the voltage amplitude
at the internal system pin.
Example 8.3 Power Switch

Project Name: E8.3a_DMOS_Switch; E8.3b_IGBT_Switch
Fig. E8.1 Mixed-mode simulation circuits for an HMM pulse applied to NLDMOS-SCR with
current and voltage waveforms observed at 16 kV pulse amplitude
These examples provide a basic mixed-mode simulation circuit for VDMOS and
IGBT switches operating under voltage VIN of up to 600 V. The voltage and current
waveforms observed during turn-on are shown in Fig. E8.3.
References
1. Amerasekera A, Duvury C (1995) ESD in Silicon Integrated Circuits. Wiley, West Sussex.
2. Dabral S, Maloney TJ (1998) Basic ESD and I/O Design. Wiley, West Sussex.
3. Estmark K, Gossner H, Stadler W (2003) Advanced Simulation Methods for ESD Protection.
Elsevier, Amsterdam.
4. Voldman SH (2006) ESD RF Technology and Circuits. Wiley, Chichester.
5. Voldman SH (2004) ESD Physics and Devices. Wiley, Chichester.
6. Voldman SH (2005) ESD Circuits and Devices. Wiley, Chichester.
7. Vashchenko VA, Hopper P (2005) Bipolar SCR ESD devices. Microelectron Reliab 45:
457–471.
8. Proceedings of EOSESD Symposiums (1997–2007). ESD Association, www.esda.org
9. Vashchenko VA, Sinkevitch VF (2008) Physical Limitations of Semiconductor Devices.
Springer, Berlin.
10. Stadler W (2007) HBM MIL883-STD, Method 3015.7; State-of-the-Art in ESD Standards.
1st International ESD Workshop, South Lake, Tahoe, 127–147.
11. ESD Association (1994) HBM-ESD Sensitivity Testing: Human Body Model (HBM)-
Component Level, ESD Association Standard, S5.1.
12. ESD Association (1994) MM-ESD Sensitivity Testing: Machine Model (MM)-Component
Level, ESD Association Standard, S5.2.
13. IEC 61000-4-2 (2008-12) Electromagnetic Compatibility (EMC) – Part 4-2: Testing and
Measurement Techniques – Electrostatic Discharge Immunity Test.
14. Sze S (1981) Physics of Semiconductor Devices. Wiley, New York, NY.
15. Vashchenko VA, Sinkevitch VF (1996) Current instability and burnout of HEMT structures.
Solid-State Electron 39:851–856.
16. Vashchenko VA, Kozlov NA, Martynov YB, Sinkevitch VF, Tager AS (1996) Negative dif-
ferential conductivity and isothermal drain breakdown of the GaAs MESFET. IEEE Trans
Electron Dev 43: 513–518.
17. Vashchenko VA, Martynov YB, Sinkevitch VF (1996) Physical limitation on drain voltage
of power PM HEMT. Microelectron Reliab 37:1137–1141.
18. Vashchenko VA, Martynov YB, Sinkevitch VF, Tager AS (1996) Current instability in GaAs
n+ -i-n+ structures as a limitation of the maximum drain voltage of power MESFET’s. Solid-
State Electron 39:1027–1031.
19. Vashchenko VA, Martynov YB, Sinkevitch VF, Tager AS (1996) Electrical current instability
at gate breakdown in GaAs MESFET. IEEE Trans Electron Devices 43(12):2080–2084.
20. Vashchenko VA, Martynov YB, Sinkevitch VF (1997) Electrical instability and filamentation
in ggMOS protection structures. Solid-State Electron 41(11):1761–1767.
21. Tazzoli A, Danesin F, Zanoni E, Meneghesso G (2007) ESD robustness of AlGaN/GaN
HEMT devices, 29th Electrical Overstress/Electrostatic Discharge Symposium (EOS/ESD),
EOS/ESD 16–21 September 2007. 4A.4-1–4A.4-9. Anaheim, California, USA.
22. Sleiman A, Di Carlo A, Lugli P, Meneghesso G, Zanoni E, Thobel JL (2003) Channel thick-
ness dependence of breakdown dynamic in InP-based lattice-matched HEMTs. IEEE Trans
Electron Devices 50(10):2009–2014.

DOI 10.1007/978-1-4419-6565-3, C Springer Science+Business Media, LLC 2010
448 References
23. Bonch-Bruevitch VL, Zviagin NP, Mironov AG (1975) Domain Electrical Instability in
Semiconductors. Nauka, Moskow.
24. Nikolis R, Prigogine I (1979) Self-organization in nonequilibrium systems. Mir, Moscow.
25. Haken H (1980) Sinergetics. Mir, Moscow.
26. Kerner BS, Osipov VV (1994) Autosolitons. A new approach to Problems of Self-
Organization and Turbulence. Kluwer Academic Publishers, Dordrecht, Boston, London.
27. Scholl E (1987) Nonequilibrium Phase Transition in Semiconductors Self-organization
Induced by Generation and Recombination Process. Springer, Heidelberg.
28. Moens P, Bychinkin S, Reynders K, et al. (2004) Effect of hot spot hopping and drain bal-
lasting integrated vertical DMOS devices under TLP stress. Proc. Int. Rel. Phys. Symp.,
393–398. Phoenix, AZ, USA.
29. Gossner H, Muller-Lynch T, Eastmark K, Stecher M (1999) Wide range control of sustaining
voltage of ESD protection elements realized in smart power technology. Proc. EOS/ESD
Symp., 19–27. Phoenix, AZ, USA.
30. Keppens B, Mergens M, Armer J, et al. (2003) Active-area-segmentation (AAS) technique
for compact ESD robust, fully silicided NMOS design. Proc. EOS/ESD Symp., 250–258.
31. DECIMMTM Angstrom Design Automation (2010) Release 2.0, www.analogesd.com
32. Hanwa (2009) HBM, MM Waveform Capture System. Hanwa Electronic IND. Co., Ltd.,
Japan.
33. Scholz M, et al. (2007) Calibrated wafer-level HBM measurements for quasi-static and
transient device analysis. Proceedings of EOS/ESD Symposium, 89–94.
34. Grutchfield HB, Moutoux TJ (1966) Current mode second breakdown in epitaxial planar
transistors. IEEE Trans Electron Devices 13:743–747.
35. Oxner ES (1985) Power Field Effect Transistor and Its Application. Prentice-Hall,
Englewood Cliffs, NJ.
36. Vashchenko VA, Kuznetsov V, Hopper P (2007) Implementation of dual-direction SCR
devices in analog CMOS process. Proc. EOSESD Symp., 75–79.
37. Concannon A, Vashchenko VA, ter Beek M, Hopper P (2003) A device level negative
feedback in the emitter line of SCR-structures as a method to realize latch-up free ESD pro-
tection. 41st Annual 2003 IEEE International Reliability Physics Symposium Proceedings,
30 March–4 April 2003, 105–111 .
38. Russ C, Mergens M, Verhaege K, et al. (2001) GGSCRs: GGNMOS triggered silicon con-
trolled rectifiers for ESD protection in deep sub-micron CMOS processes. Proc. EOS/ESD
Symp., pp. 22–31.
39. Wolf S (1990) Silicon Processing for VLSI Era, Vol. 2, Process Integration. Lattice Press,
Sunset Beach, CA.
40. LaFonteese D, Vashchenko VA, Korablev KG (2009) Breakdown voltage walkout effect in
ESD protection devices. IRPS.
41. Wu J, Rosenbaum E (2004) Gate oxide reliability under ESD-like pulse stress. IEEE Trans
Educ 51(7):1192–1195.
42. Vashchenko VA, Kindt W, Hopper P, ter Beek M (2004) Implementation of 60 V tolerant
dual direction ESD protection in 5V BiCMOS process for automotive application. Proc.
EOS/ESD Symposium.
43. Wunsch D, Bell R (1968) Determination of threshold failure levels of semiconductor diodes
and transistors due to pulse voltages. IEEE Trans Nucl Sci NS-15:244–259.
44. Dwyer V, Franklin A, Campbell D (1991) Thermal failure in semiconductor devices. Solid-
State Electron 33:553–560.
45. Tasca D (1970) Pulsed power failure modes in semiconductors. IEEE Trans Nucl Sci NS-
17:364–372.
46. Charitat G (2001) Voltage handling capability and termination techniques of silicon power
semiconductor devices. IEEE BCTM 10.5, 175–183.
47. Vashchenko VA, ter Beek M (2005) ESD protection window targeting using LDMOS-SCR
devices with PWELL-NWELL super-junction. Proc. IRPS, April 2005, 612–613.
References 449
48. Chung Y, Xu H, Ida R, Baird R (2006) Snapback breakdown dynamics and ESD susceptibil-
ity of LDMOS. Proc. Int. Rel. Phys. Symp., 352–355; Steinhoff R, Huang J-B, Hower P,
Brodsky J (2003) Current filament movement and silicon melting in an ESD-robust
DENMOS transistor. Proceed. EOS/ESD Symp., 220–268.
49. Vashchenko VA, Lafonteese D (2009) System level and hot plug-in protection of high
voltage transient pins. Proc. EOS/ESD Symposium.
50. Vashchenko VA, Concannon A, ter Beek A, Hopper P (2002) Technology CAD evaluation of
BiCMOS protection structures operation including spatial thermal runaway. Proc. EOSESD
Symposium, 101–110.
51. Chatterjee A, Polgreen T (1991) A low voltage triggering SCR for on-chip ESD protection
at output and input pads. IEEE Electron Device Lett 12(1):21–22.
52. Chen JZ, Amerasekera A, Vrotos T (1995) Bipolar SCR ESD protection for high speed
submicron bipolar/BiCMOS circuit frequency integrated circuits. IEDM, 337–340.
53. Vashchenko VA, LaFonteese D (2008) Lateral PNP BJT ESD protection devices. BCTM,
53–56.
54. Li J, Gauthier R, Chatty K, Kontos D, et al. (2005) PMOSFET-based ESD protection in
65 nm bulk CMOS technology for improved external latch-up robustness. Proc. EOS/ESD
Symposium.
55. Gendron A, Renaud P, Besse P, Salamero C, Bafleur M, Nolhier N (2006) Area-efficient
reduced and no-snapback PNP-based ESD protection in advanced smart power technology.
Proc. EOS/ESD Symposium, September 2006, 69–76.
56. Renaud P, Gendron A, Bafleur M, Nolhier N (2007) High robustness PNP-based structure
for the ESD protection of high voltage I/Os in an advanced smart power technology. Proceed.
BCTM, 226–229.
57. Vashchenko VA, LaFonteese D (2009) System level and hot plug-in protection of high
voltage transient pins. EOS/ESD Symposium.
58. Voldman S, Giessier S, Nakoe J, et al. (1998) Semiconductor process and structural opti-
mization of shallow trench isolation-defined and polysilicon-bound source/drain diodes for
ESD networks. Proc. EOS/ESD Symp., 151–160.
59. Concannon A (2005) National Semiconductor, Internal Engineering Analysis Report.
60. Merrill R, Issaq E (1993) ESD design methodology. Proc. ESD/EOS Symposium, 233–237.
61. Stockinger M (2003) Boosted and distributed rail clamp networks for ESD protection in
advanced CMOS technologies. EOS/ESD Symposium Proceedings.
62. Anderson W, Krakauer D (1998) Cross referenced ESD protection for power supplies. Proc.
EOS/ESD Symp.
63. Croft G (1996) Transient supply clamp with variable RC time constant. Proc. EOS/ESD
Symp.
64. Dabral S, Aslett R, Maloney T (1994) Core clamps for low voltage technologies. Proc.
EOS/ESD Symp.
65. Juliano P, Anderson W (2003) ESD protection design challenges for a high pin-count alpha
microprocessor in a 0.13um CMOS SOI technology. Proc. EOS/ESD Symp.
66. Li J, Gauthier R, Rosenbaum E (2004) A compact, timed shut-off, MOSFET-based power
clamp foe on chip ESD protection. Proc. EOS/ESD Symp.
67. Poon SS, Maloney T (2002) New consideration for MOSFET power clamps. Proc. EOS/ESD
Symp., 1–5.
68. Stockinger M, Miller J (2004) Advanced ESD rail clamp network design for high-voltage
CMOS applications. Proc. EOS/ESD Symp.
69. Worley ER, Gupta R, Jones B, et al. (1995) Sub-micron chip ESD protection schemes which
avoid avalanching junctions. Proc. EOS/ESD Symp., 13–20.
70. Maloney T, Kan W (1999) Stacked PMOS clamps for high voltage power supply protection.
Proc. EOS/ESD Symp., 70–77.
71. Smedes T, Heringa A, van Zwol J , de Jong P (2002) The impact of substrate resistivity on
ESD protection devices. Proc. EOS/ESD Symp., 354.
450 References
72. Vashchenko VA, Concannon A, ter Beek M, Hopper P (2003) Increasing the ESD protection
capability of over-voltage NMOS structures by comb-ballasting region design. 41st Annual
2003 IEEE International Reliability Physics Symposium Proceedings, 30 March–4 April
2003, 261–268.
73. Vashchenko VA, Concannon A, ter Beek M, Hopper P (2003) ESD-implant effect on pro-
tection capability of NMOS structures. 33rd Conference on European Solid-State Device
Research, 2003. ESSDERC’03. 16–18 September 2003, 565–568.
74. Quittard Q, Mrcarica Z, Blanc F, Notermans G, Smedes T, van Zvol H (2006) ESD protection
for high voltage CMOS technologies. Proc. EOS/ESD Symp., 77–86.
75. Vassilev V, Lorenzini M, Jansen P, Vashchenko VA, et al. (2003) Snapback circuit model
for cascoded NMOS ESD over-voltage protection structures. 33rd Conference on European
Solid-State Device Research, 2003. ESSDERC’03, 16–18 September 2003, 561–564.
76. Smedes T, Li Y (2003) ESD phenomena in interconnect structures. Proc. EOS/ESD Symp.
77. Boselli G, Duvvury C, Reddy V (2003) Efficient pnp characteristics of pMOS transistors in
sub-0.13um ESD protection circuits. Proc. EOS/ESD Symp., 257–266.
78. Duvvury C, Ramaswamy S, Amerasekera A, Cline RA, Andresen BH, Gupta V (2000)
Substrate pump NMOS for ESD protection applications. Electrical Overstress/Electrostatic
Discharge Symposium Proceedings, 26–28 September 2000, 7–17.
79. Vashchenko VA, Concannon A, ter Beek M, Hopper P (2002) Emitter injection control in
LVTSCR for latch-up free ESD protection. Proc. MIEL 2:741.
80. Meneghesso G, Tazzoli A, Marino FA, Cordoni M, Colombo P. Development of a new high
holding voltage SCR-based ESD protection structure. IEEE International Reliability Physics
Symposium, IRPS 2008, April 27–May 1 2008, 3–8.
81. Vashchenko VA, Concannon A, ter Beek M, Hopper P (2004) High holding voltage cascoded
LVTSCR structures for 5.5-V tolerant ESD protection clamps. IEEE Trans Device Material
Reliab 4(2):273–280.
82. Synopsys (2002) “TSUPREM-4” Two-Dimensional Process Simulation Program, Version
2002.4.0.
83. Synopsys (2002) MEDICI Two-Dimensional Device Simulation Program, Version 2002.4.0.
84. Chatty K, Peachey N, Quek C, et al. (2010) SCR-based power clamp for low leakage
applications. Proceedings of the International ESD Workshop, Poster 2.7.
85. Vashchenko VA, Farrenkopf D, Hopper P (2007) Active control of the triggering character-
istics of NPN BJT, BSCR and NLDMOS-SCR devices. 19th International Symposium on
Power Semiconductor Devices and IC’s, ISPSD’07, 27–31 May 2007, 41–44.
86. Avery L (1983) Using SCR’s as transient protection structures in integrated circuits.
Proceedings of ESD/EOS Symposium, 27–29.
87. Mergens M, Russ C, Verhaege K, et al. (2003) Diode-triggered SCR (DTSCR) for RF-ESD
protection of BiCMOS SiGe HBTs and CMOS ultra-thin gate oxides. Intl. Electron Devices
Meeting Technical Digest, 515–518.
88. Di Sarro J, Vashchenko VA, Rosenbaum E, Hopper P (2008) A dual-base trig-
gered SCR with very low leakage current and adjustable trigger voltage. 30th
Electrical Overstress/Electrostatic Discharge Symposium, EOS/ESD, 7–11 September 2008,
242–248.
89. Olson N, Vashchenko V, Rosenbaum E, Hopper P (2008) Small footprint trigger voltage con-
trol circuit for mixed-voltage applications. 30th Electrical Overstress/Electrostatic Discharge
Symposium, EOS/ESD, 7–11 September 2008, 196–203.
90. Vashchenko VA, Olson N, Farrenkopf D, Kuznetsov V, Hopper P, Rosenbaum E (2007)
Mixed device-circuit solution for ESD protection of high-voltage fast pins. 45th Annual
IEEE International Reliability Physics Symposium, Proceedings, 15–19 April 2007,
602–603.
91. Di Sarro J, Chatty K, Gauthier R, Rosenbaum E (2007) Evaluation of SCR-based ESD
protection in 90nm and 65nm CMOS technologies. Proc. Intl. Reliability Physics Symp.,
348–357.
References 451
92. Tan PY, Indrajit M, Li PH, Voldman S (2005) RC-triggered PNP and NPN simultaneously
switched silicon controlled rectifier ESD networks for sub-0.18 μm technology. Proc. Intl.
Symp. on the Physical and Failure Analysis of Integrated Circuits, 71–75.
93. Gerdemann A, Bhatia K, Rosenbaum E (2007) A Kelvin transmission line pulsing system
with optimized oscilloscope ranging. Proc. EOS/ESD Symp., 80–88.
94. Chen J, Zhang X, Amerasekera A, Vrotsos T (1996) Design and layout of a high ESD perfor-
mance NPN structure for submicron BiCMOS/bipolar circuits. Proc. Int. Rel. Phys. Symp.,
227–232.
95. Coffing D, Ida R (1998) Analysis of a Zener-triggered bipolar ESD structure in a BiCMOS
technology. Proc. BCTM, 227–232.
96. Vashchenko VA, Kuznetsov V, Hopper P (2007) ESD protection of fast transient pins in
bipolar processes. Bipolar/BiCMOS Circuits and Technology Meeting, BCTM’07, IEEE,
222–225.
97. Chen JZ, Amerasekera A, Vrotos T (1995) Bipolar SCR ESD protection for high speed
submicron bipolar/BiCMOS frequency integrated circuits. IEDM, 337–340.
98. Vashchenko VA, Hopper P (2006) A new principle for a self-protecting power transistor
array design. Power Semiconductor Devices and IC’s, ISPSD 2006. IEEE International
Symposium on 4–8 June 2006, 1–4; Vashchenko VA, Hopper PJ (2006) Self-protecting
arrays for open drain circuits. 44th Annual IEEE International Reliability Physics
Symposium Proceedings, 26–30 March 2006, 637–638.
99. Trajkovic T, Udugampola N, Pathirana V, Mihaila A, Udrea F, Amaratunga GA, Koutny B,
Ramkumar K, Geha S (2009) High frequency 700 V PowerBrane LIGBTs in 0.35μm bulk
CMOS technology. 21st International Symposium on Power Semiconductor Devices & IC’s,
ISPSD 2009, 14–18 June 2009, 307–310.
100. Udrea F (2007) SOI-based devices and technologies for high voltage ICs. IEEE
Bipolar/BiCMOS Circuits and Technology Meeting, BCTM’07, September 30, 2007–Oct.
2, 2007, 74–81.
101. Scholz M, Linten D, Thijs S, et al. (2010) Transient characterization of high voltage
switching devices. Proceedings of International ESD Workshop, Poster 2.4.
102. Vashchenko VA, Hopper P (2005) BSCR ESD protection in 250 V process taking into
account the turn-off effect. Proceedings of the Bipolar/BiCMOS Circuits and Technology
Meeting, 9–11 October 2005, 70–73.
103. Linten D, Vashchenko VA, Scholz M, et al. (2008) Extreme voltage overshoots in HV
snapback devices during HBM ESD stress. Proc. EOS/ESD Symp.
104. Eherton M, Khazhinsky M, Miller J, et al. (2009) EDA tool for checking ESD specific I/O
ring integration rules. International ESD Workshop.
105. Concannon A (2007–2008) National Semiconductor, Internal Engineering Analysis Report.
106. Anderson WR (2009) I/O Design Concepts for the ESD Engineer, International ESD
Workshop, South Lake, Tahoe, 2009 Seminar #1.
107. Joshi S, Rosenbaum E (2002) Compact modeling of vertical ESD protection NPN transistors
for RF circuits. Proc. EOS/ESD Symp., 289–295.
108. Vassilev V, Groeseneken G, Jenei S, Maes H (2002) Modeling and extraction of RF per-
formance parameters of CMOS electrostatic discharge protection devices. Proc. EOS/ESD
Symp., 111–118.
109. Russ C, Verhaege K, Bock K, et al. (1996) A compact model for the grounded gate NMOS
behavior under CDM ESD stress. Proc. EOS/ESD Symp., 302–315.
110. Mergens M, Wilkening W, Mettler S, et al. (1999) Analysis and compact modeling of lateral
DMOS power devices under ES stress conditions. Proc. EOS/ESD Symp., 1–10.
111. Vassilev V, et al. (2001) Analysis arid improved compact modeling of the breakdown
behavior of sub-0 25 micron ESD protection ggNMOS device. Proc. EOSESD Symp.,
62–70.
112. Vassilev V, Vashchenko VA, Jansen P, et al. (2005) ESD circuit model based protec-
tion network optimisation for extended-voltage NMOS drivers. Microelectron Reliab 45:
1430–1435.
452 References
113. Aliaj B, Vashchenko V, Cui Q, Liou J, Tcherniaev A, Ershov M, LaFonteese D (2009)

2.5-Dimensional simulation for analyzing power arrays subject to ESD stresses. Electrical
Overstress/Electrostatic Discharge Symposium Proceedings, 3A.2.1–3A.27.
114. Monticelli D (2004) The future of complementary bipolar. BCTM, 21–25.
115. Denison M, Murtaza S, Steinhoff R (2007) 25V ESD NPN transistor optimized by dis-
tributed emitter ballasting using emitter contact area segmentation. Proc. Int. Rel. Phys.
Symp., 604–605.
116. Jansen P, Thijs S. Linten D, et al. (2005) RF ESD protection strategies – the design and
performance trade-off challenges. Proceedings of the IEEE 2005 Custom Integrated Circuits
Conference, 21–21 September 2005, 489–496.
117. Stadler W, Brodbeck T, Gartner R, Gossner H (2006) Cable discharges into communication
interfaces. Proc. EOS/ESD Symposium, 144–151.
118. Pommerenke D (2001) Charged cable event. IEEE 802.3 Cable Discharge Ad-Hoc, February
2001.
119. Pischl N (2005) ESD transfer through Ethernet magnetics. Proc. Intl. EMC Symposium,
356–363.
120. Deatherage J, Jones D (2000) Multiple factors trigger cable discharge events in Ethernet
LANs. Electron Design 48(25):111–116.
121. Ram C (2007) Cable Discharge Event (CDE) Stress Simulation Test Procedure for Integrated
Circuits. Texas Instruments. TX, USA.
122. Sandler SM (2002) Switchmode Power Supply Simulation with Pspice and SPICE3.
McGraw-Hill, New York, NY.
123. Erickson R, Maksimovic D (2001) Fundamentals of Power Electronics. Kluwer Academic
Press, Dordrecht.
124. Mitchell D (1988) Dc-dc Switching Regulator Analysis. McGraw-Hill, New York, NY.
125. Basso CP (2008) Switch-Mode Power Supplies. McGraw-Hill, New York, NY.
126. Moss J, Ranucci P, Bikulcius S (2009) Optimizing power FET design. Innovation Solution
Conference, National Semiconductor.
127. Goyal A, Whitfield J, Hong C, et al. (2008) Unique ESD failure mechanism of high voltage
LDMOS transistors for very fast transients. IEEE IRPS, Phoenix, 673–674.
128. Chung Y, Xu H, Ida R, Min WG, Baird B (2005) ESD scalability of LDMOS devices for
self-protected output drivers. Proceedings of IS PSD, 351–354.
129. Chung Y, et al. (2006) Proc. Int. Rel. Phys. Symp. 352; Steinhoff R, et al. (2003) Proc.
EOS/ESD Symp., 220.
130. Parthasarathy V, Khemka V, Zhu R, et al. (2002) A multi trench analog+logic protection
(M-TRAP) for substrate crosstalk prevention in a 0.25 μm smart power platform with 100V
high-side capability. IEEE Electron Device Lett 23(4):212–214.
131. Vashchenko VA, et al. (2010) International ESD Workshop, in print.
132. Hower P (2010) Private Communication.
133. Mergens M, Mayerhofer MT, Willemen JA, Stecher M (2006) ESD protection considerations
in advanced high-voltage technologies for automotive. Proc. EOSESD, 54.
134. Chen W, Ker M-D, Jou Y, et al. (2009) Source-side engineering to increase holding voltage
of LDMOS in a 0.5um 16-V BCD technology to avoid latch-up failure. IEEE Proc. of 16th
IPFA; High-voltage nLDMOS in Waffle-layout style with body-injected technique for ESD
protection. IEEE Electron Device Lett 30(4):389–391.
135. Aliaj B, Vashchenko VA, et al. (2009) EOS/ESD Symp., 3A.2-7126; Moss J, Ranucci P,
Bikulcius S (2009) Optimizing power FET design. Innovation Solution Conference, National
Semiconductor.
136. Lafonteese D, Vashchenko VA, Linten D, et al. (2009) Self-protection capability of power
arrays. Proceedings EOS/ESD Symposium.
137. Virtuoso Spectre Circuit Simulator User Guide, Version 7.0.1, June 2008.
138. Chapra SC (2006) Applied Numerical Methods with Matlab. McGraw-Hill, New York, NY.
EOS/ESD placeholder for reviewer suggestion Chapter 3.
References 453
139. Gossner H (2009) ESD robust systems–IC design contribution, South Lake, Tahoe,
Seminar #2.
140. Gossner H, Drüen S, Esmark K, et al. (2005) SoC. A real challenge for ESD protection.
Proc. EOSESD.
141. Mergens M, Mayerhofer MT, Willemen JA, Stecher M (2006) ESD protection considerations
in advanced high-voltage technologies for automotive. Proc. EOSESD, 54.
142. Bonfert D, Gieser H, Bock, K, Svasta P, Ionescu C (2009) Pulsed behavior of polymer pro-
tection devices. Electron Technol. ISSE 2009. 32nd International Spring Seminar on 13–17
May 2009, 1–6.
143. Scholz M, Linten D, Thijs S, Sawada M, Nakaei T, Hasebe T, Groeseneken G (2008) On-
wafer human metal model. System-level ESD stress on component level. Proc. RCJ.
144. Scholz M, Linten D, Thijs S, Sawada M, Nakaei T, Hasebe T, Lafonteese D, Vashchenko
VA, Vandersteen G, Groeseneken G (2009) On-wafer human metal model measurements for
system-level ESD analysis. EOSESD Symposium.
145. Ashton R (2007) System level ESD testing – the waveforms. Conformity Magazine
12(10):28–34.
146. Stadler W, et al. (2004) From the ESD robustness of products to the system ESD robustness.
Proceedings EOS/ESD Symposium, 67–74.
147. Peachey N (2008) ESD Open Forum – ESD stressing of devices and components.
Conformity Magazine 13(6):22–23.
148. Wang K, et al. (2004) Characterization of human metal ESD reference discharge event
and correlation of generator parameters to failure levels – part II: Correlation of generator
parameters to failure levels. IEEE Trans EMC 46(4):505–511.
149. Scholz M, et al. (2007) Calibrated wafer-level HBM measurements for quasi-static and
transient device analysis. Proceedings EOS/ESD Symposium, 89–94.
150. Wang AZ, Tsay CH (1999) A compact square-cell ESD structure for BiCMOS IC. IEEE
BCTM 3.1, 46.
151. Vashchenko VA, ter Beek M, Kindt W, Hopper P (2004) ESD protection of the
high voltage tolerant pins in low-voltage BiCMOS processes. Proceedings of BCTM,
277–280.
152. Vashchenko VA, Concannon A, ter Beek M, Hopper P (2003) Multi-port ESD protection
using bi-directional SCR structures. Proc. BCTM, 137–140.
153. Besse P (2010) ESD/EMC in an automotive environment. Proceedings of International ESD
Workshop, Seminar #2.
154. Paddan P, Best M, Ferrer E, et al. (2010) An ESD robust wireless sensor chip for medical
application. Proceedings of International ESD Workshop, Keynote Presentation.
155. Wysocki1 P, et al. (2009) Effect of electrostatic discharge on electrical characteristics of
discrete electronic components (technical brief). PHM Conference, Anaheim.
156. Davidson S (2006) Searching for clues: diagnosing IC failures. IEEE Design Test Comput
23(1):67–68.
157. Desineni R, Blanton RD (2005) Diagnostics of arbitrary using neighborhood function extrac-
tion. Proceedings of 23rd IEEE VLSI Test Symposium, 366–373; Khatir Z, Skipedj J,
Lallemand R (1994) Main technological causes and some mechanisms of high power
GTO’s failures. Proceedings of Fifth International Conference on Power Electronics and
Variable-Speed Drives, 319–324.
158. Li TS, Huang CL (2009) Defect spatial pattern recognition using a hybrid SOM-SVM
approach in semiconductor manufacturing. Expert Syst Appl Int J 36(1):374–385.
159. Salman AA, Beebe SG, Emam M, Pelella MM, Ioannou DE (2006) A novel device for ESD
protection in deep sub-micron SOI technologies. International Field Effect Diode (FED)
Electron Devices Meeting, IEDM’06, 11–13 December 2006, 1–4.
160. Ille A, Stadler W, Kerber A, et al. (2006) Ultra thin gate oxide reliability in the ESD time
domain. Proc. EOS/ESD Symp.
454 References
161. Khazhinsky M, Stockinger M, Miller J, Weldon JC (2004) Engineering single NMOS and
PMOS output buffers for maximum failure voltage in advanced CMOS technolo-gies. Proc.
EOS/ESD Symp.
162. Salman A, Gauthier R, Wu E, et al. (2002) Electrostatic discharge induced oxide breakdown
characterization in a 0.1 um CMOS technology. 40th International Reliability Phys. Symp.
Proceedings, 170–174.
163. Stathis J, Maria D (1998) Reliability projection for ultra-thin oxide reliability in CMOS
devices and circuits. IEDM.
164. Weir B, Leung CC, Silverman P, Alam M (2005) Gate dielectric breakdown ion the time-
scale of ESD events. Microelectron Reliab 45:427–436.
165. Hsueh KK, Sin-Hao K, Lee J, Rosenbaum E (2008) UVeriESD: an ESD verification tool for
SoC design. IEEE Asia Pacific Conference on Circuits and Systems, 2008. APCCAS 2008,
November 30, 2008–December 3, 2008, 53–56.
166. Vashchenko VA, Scholz M, Jansen Ph, et al. (2006) Turn-off characteristics of the CMOS
snapback ESD protection devices – new insights and its implications. Proceed. EOS/ESD
Symp., 39–45.
167. Worley E, Bakulin A (2002) Optimization of input protection diode for high speed
applications. Proc. EOS/ESD Symp., 62–72.
168. Concannon A, Vashchenko VA, ter Beek M, Hopper P (2004) ESD protection of double-
diffusion devices in submicron CMOS processes. Proc. ESSDERC, 261–264.
Index
A Bipolar output
Absolute maximum limits, 7–8, 87, 89, 184, driver protection, 234
290, 340, 375, 395 protection, 234–235, 290, 292–293
Active clamp, 1, 3, 5–6, 118, 158–160, Bipolar SCR (BSCR), 7, 87, 116–117, 135,
215–217, 219–229, 232, 249, 151–152, 183, 187–188, 190,
259–262, 272–279, 295–296, 192–193, 210–211, 294, 418
299–300, 305–307, 310, 338–339, Breakdown, 5–68, 71–72, 75, 78–79, 86–87,
341–342, 373, 418 89–90, 94, 98, 102–103, 107–108,
Amplifiers, 139, 143, 146, 160, 214, 216, 232, 110–122, 126–132, 141–142, 145,
235, 281–295, 301, 307, 312, 368, 150–151, 161–163, 164–166, 169,
417–418 172, 177, 182–189, 195, 203, 207,
Analog–digital converter (ADC), 296–301, 209, 217, 222–223, 232, 236, 240,
307, 367–368 265–268, 290, 294, 313, 327–330,
Analog ESD design, 2–4, 16–17, 333, 336–337, 340, 350–352, 377,
139–140, 142 380, 418, 441
Avalanche voltage walkout effect, 122
breakdown, 5–7, 10–12, 14, 16, 23–31, 35, Buried diode, 113
37, 40, 45, 54, 57–59, 67, 89, 115, Buried oxide (BOX), 83
119, 126, 150, 161, 172, 182, 184, BVCEO-referenced BJT clamp, 184
187–189, 195, 203, 217, 222, 236,
265, 266, 290, 340, 352 C
voltage reference, 24, 182, Cable discharge events (CDE), 91, 301–307,
187–189, 340 398, 400, 404–405
diodes, 7, 23, 24, 27, 29–30, 33, 47, 79, CE certification mark, 397
85, 111–113, 116, 121, 126–127, Charge device model (CDM), 2, 8, 87, 91, 160,
183–187, 190–192, 194, 207–210, 207, 215, 219, 240–242, 282, 296,
248, 255, 340, 347, 356, 370, 300, 305–307, 374, 380–385, 396,
380–381 398, 406, 409, 442
–injection conductivity modulation, 11, Checker, 2, 218, 222
19, 35, 63, 103–104, 107, 117–118, CMOS diodes, 140–141, 242, 415
123, 163, 198, 250, 333, 375 CMOS input, 235–241, 261, 290, 294,
300, 373
B CMOS process, 1–2, 33, 44, 55, 67, 70–83,
Back-to-back diode clamp, 259–260, 292, 352 86–87, 98, 102–113, 117–132,
BCD process, 15, 33, 72–74, 83–87, 92–96, 134–135, 140–142, 150, 161,
112, 120–121, 123, 127, 149, 163–166, 169–170, 177, 194, 198,
150–151, 168, 190, 287, 320, 353, 211, 219, 223, 225, 227–229, 237,
378–379, 428 240, 250, 277, 279, 281, 284, 297,
Bipolar differential input, 232–234 306, 414, 417–418, 420, 426, 428
455
456 Index
Conductivity modulation, 5–7, 9–11, 13, Emitter injection control, 174

15–69, 89–90, 93, 98, 102–104, Enhanced Zener clamp, 117, 186, 188
107, 114, 115, 117–119, 121–123, ESD clamps, 3, 5–6, 52, 67, 81, 97, 112,
126–127, 129–131, 143, 147, 150, 143, 147, 155–212, 240, 262, 272,
155, 157–158, 160, 163, 166, 172, 287, 305, 317, 320, 348, 395, 403,
186–188, 198, 216, 219, 250, 253, 412–416, 443ESD compact models,
282, 332–333 263–272
Controllers, 281, 284, 296, 317–318, 321, 323, ESD pad ring, 2–3, 156, 213–214, 219
328, 357–364, 374, 422 ESD protected area (EPA), 398–400, 403, 422,
Current filamentation, 9, 18–20, 30, 41, 430, 443
51–52, 60–66, 93, 97–98, 100, ESD protection clamp, 5, 13, 94, 100, 151,
104, 407 156, 158, 166, 213, 251, 254–255,
effect, 18, 30, 60, 63–66, 93, 271, 288, 356, 375, 381, 389–390,
97–98, 100 405, 408, 422
Current sink, 365, 367–370, 372–374 ESD protection window, 7–8, 21, 67, 79,
Current stratification effect, 99–100 87–88, 94–95, 115–116, 155–156,
164, 166, 170, 177, 184, 192,
D 195, 198, 214, 240, 254, 282–283,
DD-BJT device, 136 290–291, 293–295, 317, 319, 340,
DD-SCR structure, 130 353, 356, 370–371, 374–375, 383,
DECIMMTM , 3, 14, 27, 68, 149, 204, 215, 409, 420
263, 272, 307, 387, 443 ESD SOA, 91–97, 102, 199
simulation examples, 14, 68, 148, 204, 272, External avalanche diode, 190
307, 387, 443
Deep N-isolation layer, 73, 128 F
Deep trench isolation (DTI), 71–72, Field electrode, 94, 120, 122, 169, 183
85–86, 321 Filamentation, 9, 18–20, 41–43, 51–52, 60–63,
DeMOS, 70, 75, 118–121, 123, 194–195, 334, 100, 104, 407
377, 423–424 effect, 18, 30, 60, 63–66, 93,
DeMOS-SCR, 118–121, 194, 423–424 97–98, 100
DIAC, 74, 131, 132, 134, 138, 262–263, 305, FOX device, 106–107, 110, 171, 235
418–420 FOXSCR, 109–111, 116, 150
Digital-to-analog converter (DAC), 127, 281, FOX Snapback device, 169–170
296–299, 301, 306–307
Digital circuits, 1, 3, 213, 295 G
Digital ESD Design, 2–3 Gate coupling effect, 169, 204, 268
Diode-triggered SCR (DTSCR), 177–178, 182, Gated diodes, 77, 141–142
184, 211–212 Gate oxide, 70, 75–77, 84–86, 102, 105–106,
Discrete components, 429–442 111, 177, 180, 212, 240, 246, 250,
Distributed systems, 59–60 294, 297, 356, 381, 430–431, 437
3-D numerical simulation, 197, 239 GGSCR, 177–178, 184
DNWELL, 73–74, 76–78, 80–82 Grounded gate NMOS, 103, 161, 177,
Dual-base triggered SCR, 178–182 340, 451
Dual-direction devices, 87, 128–135, 306
Dual-gate oxide (DGO), 70–83, 166, 281, 297, H
299, 426 HBM waveform measurements, 97
High-speed amplifier, 282–283,
E 287–288, 292
Ebers circuit, 55 High-speed I/O pins, 256–258
EEPROM, 106, 217, 222, 250–253, 274–275, HMM pulse, 401, 412–414, 443
281, 287, 299, 319, 341 Hot plug-in, 52, 118, 132, 160, 221, 290, 317,
Electro-thermal conductivity modulation, 338, 383, 407, 422–428
89, 98 Hot spot, 12, 14, 100, 105
EMC Standards, 397 Hot-swap, 395, 422–428
Index 457
Human body model (HBM), 8, 18, 21, 60, M

87, 91, 95, 97, 100–101, 104–105, Machine model (MM), 8, 21, 60, 91, 160,
159–160, 163–164, 176, 196, 163–164, 215, 219, 272–273, 305,
197–198, 204, 205–211, 215, 350, 395–396, 398, 401, 406, 443
219, 223, 226, 239–240, 242–244, Mixed device-circuit approach, 177
246, 250, 252, 268–269, 273–279, Mixed mode device-circuit solution, 191
304–305, 308–311, 313–315, Multiple voltage domains, 258–263
329–332, 334–338, 351, 355, 379, Multipoint LVDS (MLVDS), 305–306
382, 391, 392, 396, 398, 401–404,
406, 409–410, 412–416, 426–427, N
430, 442 N+ –n–n+ diode, 33–36, 68
N+ –p–n–n+ structures, 37
I NDeMOS, 70, 73–74, 76–78, 80–82, 94, 118,
I 2 C-compatible interface, 367–368 266–269, 271, 335, 337–338
IEC 61000-4-2, 8, 304, 397–398, 400–402, NDeMOS-SCR, 120, 190, 267, 268–269, 271,
404, 406–407, 410 375, 380
Instability, 9–14, 18–20, 24, 29, 33, 37–41, NDRIFT, 94–95, 119–120, 121, 122,
43–49, 51–52, 56, 59–66, 88–90, 126–127, 190
93–98, 102, 104–105, 107, Negative differential resistance, 12, 17–20,
115–116, 161, 163, 184, 195–196, 29–31, 33, 42, 45, 47, 50–51, 52,
200, 332–334 59–66, 115, 123, 167, 336
Insulated gate bipolar transistors (IGBT), Negative feedback, 7, 19–20, 52, 59–60, 74,
53, 150–151, 329, 390–391, 429, 129–131, 134, 148, 155, 203, 307
433–434, 436–440, 446 NISO, 73–74, 84–86, 96, 109, 164, 418–420
ISO gun, 306, 438–439 NLDMOS, 60, 83, 94, 97, 160, 372–373
Isolated Snapback NMOS, 134, NLDMOS power array, 2, 198, 269
164–166, 354 NLDMOS-SCR, 97, 120, 153, 183, 190,
192–193, 194, 199–200, 203,
207–208, 251, 266–267, 269, 275,
J
378, 380, 424–425, 444
JEDEC, 8, 404, 442
Non-self-aligned approach, 182
Non-uniform states, 19, 59, 62, 336
L NPN BJT devices, 15, 44–45, 49, 114–116,
Lateral PNP, 118, 121–126, 193, 298, 377–379, 124, 135–136, 278–279, 294
415–416, 422–426 NSINKER, 84–86, 144, 187, 331–332,
structures, 190 378–379
LC filter, 201, 347, 351 N-type buried layer (NBL), 84–86, 93, 95–96,
LED backlight technology, 364 103, 106, 108–109, 113, 115,
LED driver, 317, 328, 364–374 120–122, 124, 127, 143–145, 164,
Light emission diode, 364 168, 244, 379
LMU, 367–374, 373
Local ESD protection, 96, 151, 161, 215, O
247–248, 250, 256, 257, 261, 289, Open drain circuit, 164, 270–272
352, 363, 370, 405
Low-voltage CMOS output, 295 P
Low-voltage differential signal Parasitic n–p–n structure, 19, 71, 93, 103–104,
(LVDS), 222, 296–299, 302, 106–107, 110, 119, 157, 161,
305–306 163–164, 169, 219, 330, 332,
Low-voltage silicon controlled rectifier 334, 424
(LVTSCR), 7, 109–111, 119–120, Passivation, 82
150–151, 171–176, 205, 265–, 268, PBODY diffusion, 94, 331
305, 310, 414–415, 418 PDeMOS, 70, 73–74, 76–78, 80–82
model, 266 P–i–n structure, 16, 30–33, 64–65, 68
LVDS-to-CMOS converter, 298–299 P-isolation region, 86
458 Index
PLDMOS, 73, 83, 92, 94, 120–121, 183, 351, SCL-diode, 35

356–358 Self-protection, 5, 19, 96, 148, 158, 166,
PLDMOS-SCR, 120–121, 183 196–200, 234, 293, 314, 317, 319,
PMOS clamp, 158, 167–169, 184, 273–274 328–334, 341–342, 350, 357, 370,
PNP ESD device, 121, 424 376, 383, 386
P–n–p–n structures, 53–59, 67, 87, 109, 129 Serial data line (SDA), 248–250, 291, 296,
P–n–p structure, 16, 52–53, 56, 59, 61–63, 74, 368, 393
109–110, 116, 119, 122, 129, 134, Shallow trench isolation (STI), 72–74, 76–78,
167, 172, 190, 424 80–82, 84–86, 103, 106, 108–109,
Polyemitter, 86, 94, 99, 135, 293 111, 119–122, 125–127, 136–137,
Positive feedback, 7, 11, 18–21, 24, 30, 36, 140–141, 144, 242–245
40–41, 54, 56–60, 97–98, 110, Silicide exclusion mask, 80–81
115–116, 131–132, 147, 155, 172, Silicon controlled rectifier (SCR), 53, 109,
180, 265, 425 118, 265, 413
Power NLDMOS array, 332, 347 “Slave” clamps, 158
Power over Ethernet (PoE), 122, 318–319 Snapback BJT, 168, 189–191
P+ –p–n+ structure, 26–30, 68 Snapback model, 263–264, 267–268, 386
P-type buried layer (PBL), 73, 84–86, 93, 96, Snapback NMOS, 7, 19, 60, 62, 71, 73, 80–81,
124, 143–145, 168 103–105, 109–110, 134, 152, 157,
Pulsed SOA, 7–8, 19, 22, 24, 71, 87–88, 161–167, 171–175, 204–206, 217,
90–92, 94–96, 103, 184, 200, 214, 222, 235–236, 244–246, 250–251,
255, 282–283, 293, 330, 340, 349, 255, 262–265, 269, 308, 310–314,
359, 374, 374, 386, 431 339–342, 347, 353–355, 389, 393
characterization, 95 clamp, 161–167, 169, 173, 204–206, 217,
Pulse width modulation (PWM), 318, 352, 222, 235, 255, 262, 308, 310–311,
360, 365–369 313, 339–340, 347, 352, 355, 393
PWELL, 50, 73–82, 92, 94, 102–109, Space charge neutralized (SCN) diode, 35
111–113, 119–123, 127–130, 132, Spacer region, 79
140, 143–144, 163–165, 169–171, Spatially non-uniform states, 19, 59–62
179, 197, 234, 235, 238, 240, Stacked NPN BJT clamp, 135
264–268, 293–294, 308, 310, 338, STI device, 106
343, 423–424 Surface avalanche diode, 111–113
Synchronous buck
Q controller, 362–364
Quasi 3-D structure, 99–100 regulator, 351, 353–357
R T
Rail-based ESD protection, 215–247, 261, 290, Thermal breakdown, 11–12
308, 317 Thermo-electrical current filamentation, 104
Rapid thermal anneal, 86 Thick field oxide (TFO), 72, 86, 95,
RC-timer, 2, 158–160, 219–221, 224, 228–231, 105–108, 169
262, 273, 300, 338 Thin film resistors (TFR), 70, 84, 143,
RC-triggered SCR, 179 146–147, 247, 291
RESURF, 27, 94, 112, 120, 127, 155 Transient voltage suppressors (TVS), 395, 398,
405, 408, 428, 443–444
S Transmission line pulse (TLP), 8, 17, 21–22,
Safe-operating area (SOA), 2, 5, 7–8, 15, 91–99, 104, 113–115, 117–119,
19, 22, 69, 71, 87–101, 107, 143, 122–123, 130–132, 134–135,
184, 189, 196–198–200, 214, 236, 137–138, 143, 145–146, 165, 168,
255–256, 282–283, 293–294, 330, 171, 174–176, 180–181, 186–195,
333, 340, 349, 356, 359, 362, 374, 200, 202, 225–227, 232–233, 266,
386, 431–432, 434, 441–442 292–293, 330, 333, 349, 375–377,
Saturation resistors, 33, 79, 94, 104, 143–146, 379–383, 407, 420, 423, 425, 426,
173, 241, 263, 373 437–440
Index 459
Tunneling diode, 112 X

Two-stage PNP-SCR clamp, 426 “40x” rule, 147, 166–167
Two-stage protection, 5, 240–247, 253, 340,
352, 377
U Z
Ultra-high voltage (UHV), 92, 200–203 Zener diode, 16, 112

Vladislav A. Vashchenko, Andrei Shibkov (Auth.) - ESD Design For Analog Circuits-Springer US (2010) PDF

Uploaded by

Copyright:

Available Formats

Vladislav A. Vashchenko, Andrei Shibkov (Auth.) - ESD Design For Analog Circuits-Springer US (2010) PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vladislav A. Vashchenko, Andrei Shibkov (Auth.) - ESD Design For Analog Circuits-Springer US (2010) PDF

Uploaded by

Copyright:

Available Formats

ESD Design for Analog Circuits

Vladislav A. Vashchenko · Andrei Shibkov

ESD Design for Analog

ISBN 978-1-4419-6564-6 e-ISBN 978-1-4419-6565-3

Library of Congress Control Number: 2010929991

© Springer Science+Business Media, LLC 2010

Printed on acid-free paper

Springer is part of Springer Science+Business Media (www.springer.com)

This Book and Simulation Software Bundle Project

Subject and Purpose of This Book

This hierarchy is established in order to enable both sequential and independent

The Book Structure

The material presented in Chapter 2 about conductivity modulation in the

Fig. 1 Roadmap for the book composition

2.5.1 Conductivity Modulation in a Floating Base

3.4.1 Integrated NPN BJT Devices . . . . . . . . . . . . . . . 114

5 ESD Network Design Principles . . . . . . . . . . . . . . . . . . . . 213

8.4.1 The Concept of Two-Stage SCR ESD Devices . . . . . . 422

1.1 Analog and Digital in Prism of ESD Design

V.A. Vashchenko, A. Shibkov, ESD Design for Analog Circuits, 1

1.2 Important Definitions

1.2.1 ESD Protection Network

on the self-protection capability of the internal components. The self-protection of

1.2.2 ESD Clamps

Conductivity modulation Typical lateral current

Avalanche breakdown Avalanche diodes; blocking 0.01–0.1

1.2.3 Absolute Maximum Limits and Pulsed SOA

1.2.4 ESD Pulse Specification

Peak current to pulse Rise time/pulse

Human body model 0.67 2–10/150

1.2.5 Breakdown and Instability

In avalanche breakdown of the reverse-biased p–n junction, current I increases

where the proportionality factor K depends on physical and geometrical parameters

COLLECTOR-EMITTER VOLTAGE (V)

Fig. 1.3 Simulation example: snapback I–V in NPN

The example of thermal instability is provided above to deliver a primary

DECIMMTM Simulation Examples for Introduction

Example 1.1 Simulation of the avalanche breakdown I–V in elementary quasi-1D

2.1 Important Definitions and Limitations

2.1.1 Basic Semiconductor Structures

V.A. Vashchenko, A. Shibkov, ESD Design for Analog Circuits, 15

(i) p–n structure operating at reverse bias in avalanche breakdown mode

2.1.2 Conductivity Modulation and Negative Differential

2.1.3 Spatial Current Instability, Filamentation, and Suppression

2.1.4 Snapback Operation

a high-conductivity or high-current state depending on the load characteristics in

PULSED CURRENT (A)

2.1.5 Notes to the Methodology of Material Presentation

(i) Silicon semiconductor material parameters

2.2 Avalanche Breakdown in Reverse-Biased p–n Structure

2.2.1 Analytical Description of the Avalanche Breakdown

From the solution of (1.1) and (1.2) at αn = αp = g, the current density is

2.2.2 Numerical Analysis of the Avalanche Breakdown

2.3 Double-Avalanche–Injection in p–i–n Structures

2.3.1 An Analytical Description of the Effect

E = EBR + δE (x, δI) . (1.10)

2.3.2 Numerical Analysis for the p–i–n Diode Structure

2.4 Avalanche–Injection in Si n+ –n–n+ Diode Structure

2.4.1 Analytical Approach