Advanced VLSI Design & Testability Issues 2021
Advanced VLSI Design & Testability Issues 2021
Edited by
Suman Lata Tripathi, Sobhit Saxena, and
Sushanta Kumar Mohapatra
MATLAB® is a trademark of The MathWorks, Inc. and is used with permission. The MathWorks
does not warrant the accuracy of the text or exercises in this book. This book’s use or discussion
of MATLAB® software or related products does not constitute endorsement or sponsorship by The
MathWorks of a particular pedagogical approach or particular use of the MATLAB® software
Reasonable efforts have been made to publish reliable data and information, but the author and
publisher cannot assume responsibility for the validity of all materials or the consequences of
their use. The authors and publishers have attempted to trace the copyright holders of all material
reproduced in this publication and apologize to copyright holders if permission to publish in this
form has not been obtained. If any copyright material has not been acknowledged please write and
let us know so we may rectify in any future reprint.
Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information
storage or retrieval system, without written permission from the publishers.
For permission to photocopy or use material electronically from this work, access www.copyright.com
or contact the Copyright Clearance Center, Inc. (CCC), 222 Rosewood Drive, Danvers, MA 01923, 978-
750-8400. For works that are not available on CCC please contact [email protected]
Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Typeset in Times
by codeMantra
Contents
Preface to the First Edition ....................................................................................... ix
Editors .................................................................................................................... xiii
Contributors ............................................................................................................. xv
v
vi Contents
CHAPTER ORGANIZATION
This book is organized in 20 chapters.
Chapter 1 covers various aspects of digital design with programmable devices,
e.g., programmable logic arrays, complex programmable logic device, and field-
programmable gate array (FPGA). Moreover, this chapter emphasizes on uses of
FPGA on various image-processing and biomedical applications.
Chapter 2 mainly focuses on the review of the basic concepts of digital design,
i.e., number system, logic families, combinational and sequential circuits, registers,
counters, and memories.
Chapter 3 deals with the basics of Verilog HDL and its special features toward
implementation of digital design as well as analog design. It starts with a background
on the emergence and importance of hardware description languages (HDLs) with a
brief introduction to VHDL and Verilog HDL.
Chapter 4 focuses on the basic understanding of digital design in perspective
of HDLs and its structure, which will enlighten the orientation of its usage with the
various scopes of HDL. Next queued up some advanced topics such as high-level
synthesis, HDL tool suites, and HDL-based design flows. Depending upon the basic
ideology of HDLs, the applications and various types of HDLs for analog circuits,
digital logic, and printed circuit boards (PCBs) are elaborated.
Chapter 5 introduces basics of various HDLs available for analog circuit design,
digital circuit design, and PCB design. The last two sections of the chapter are
devoted to the most popular HDLs for digital design, i.e., VHDL and Verilog.
Chapter 6 describes a general overview of the various technologies, materials,
and architectures; researchers are concentrating to continue the technology beyond
Moore’s law with low power consumption.
Chapter 7 presents the ebb and flow of current research related to the replace-
ment of silicon by new material or using new geometry for silicon-based FET (field-
effect transistor) design, considering various design parameters while providing
some desired solutions.
Chapter 8 begins with the introduction of the biosensor, followed by its classifica-
tion. Then a detailed study of the FET-based biosensor which includes the detection
principle and its types is presented. The detection principle is based on the variation
ix
x Preface to the First Edition
Chapter 17 deals with the digital beamforming that has many of the advantages
over its analog counterpart. In most cases, less power is needed to perform the beam
steering of the phased array antenna. This chapter will focus on the main compo-
nents involved in the implementation of digital beamforming such as analog-to-
digital converters, digital down converters, and so on.
Chapter 18 describes a cross-coupled pair of transistors that are used to gener-
ate negative resistance. Again, active inductors (AIs) are also used in the circuit
to make the voltage-controlled oscillator (VCO) more tunable over a larger band-
width as compared with the passive inductor–based VCO, where AIs are there
for coarse tuning and variable capacitors are for fine-tuning of the frequency of
oscillation.
Chapter 19 describes the basics of simulation and its categorization into logic
and fault simulation. The importance of fault simulation algorithms before actual
testing on hardware is explained. The evolution of algorithms for reducing test time
from the simplest serial to differential fault simulation algorithm is described by
explaining each algorithm. The concept of fault simulation is further practically
implemented using Verilog language in which parallel and concurrent algorithms
are implemented and simulation results in terms of waveforms are explained for bet-
ter understanding.
Chapter 20 deals with hardware obfuscation that basically means hiding the
IC’s structure and function, which makes it much more difficult to reverse-engi-
neer by the adversaries. Thereby, obfuscation provides a means of making the
circuit structurally and functionally difficult to comprehend, which increases the
cost and time required to reverse-engineer it, providing security.
Dr. Sobhit Saxena has a vast teaching experience of more than 10 years in various
colleges and universities. He has designed a new hybrid system of Li-ion battery and
supercapacitor for energy storage application. He worked as an SEM (scanning elec-
tron microscopy) operator for 4 years against MHRD fellowship. His area of exper-
tise includes nanomaterial synthesis and characterization, electrochemical analysis,
and modeling and simulation of CNT-based interconnects for VLSI circuits.
xiii
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
Contributors
Debiprasad Priyabrata Acharya Santosh Kumar Gupta
Department of Electronics and Department of Electronics and
Communication Engineering Communication Engineering
NIT Rourkela Motilal Nehru National Institute of
Rourkela, India Technology Allahabad
Prayagraj, India
Krutideepa Bhol
School of Electronics Engineering Aditya Kumar Hota
VIT-AP University Department of Electronics and
Amaravati, India Tele-communication Engineering
Veer Surendra Sai University of
Reena Chandel Technology
Department of Sciences Burla, India
Dr. Y.S Parmar University of
Horticulture and Forestry Biswajit Jena
Solan India Department of Electronic and
Communication Engineering
Rekha Chaudhary Koneru Lakshmaiah Education
School of Electronics and Electrical Foundation
Engineering Guntur, India
Lovely Professional University
Jalandhar, India S. Jena
School of Electrical Engineering
Ananya Dastidar Kalinga Institute of Industrial
Electronics and Communication Technology, KIIT Deemed to be
Engineering University
College of Engineering and Technology Bhubaneswar, India
Bhubaneswar, India
Himani Jerath
J. K. Das School of Electronics and Electrical
School of Electronics Engineering Engineering
Kalinga Institute of Industrial Lovely Professional University
Technology, KIIT Deemed to be Jalandhar, India
University
Bhubaneswar, India, Abhishek Kumar
Department of Electronics and
Pranab Kishore Dutta Communication Engineering
Department of Electronics and IIT Jodhpur
Communication Engineering Karwar, India
North Eastern Regional Institute of
Science and Technology
Itanagar, India
xv
xvi Contributors
R. L. Pradhan
HAL, SLARDC
CONTENTS
1.1 Introduction .................................................................................................... 1
1.2 Factory-Programmable Devices ..................................................................... 2
1.3 Read-Only Memory ........................................................................................ 2
1.4 Programmable Read-Only Memory ............................................................... 4
1.5 Erasable Programmable Read-Only Memory ................................................ 4
1.6 Electrically Erasable Programmable Read-Only Memory............................. 5
1.7 Field-Programmable Devices ......................................................................... 5
1.8 Programmable Array Logic ............................................................................ 6
1.9 Programmable Logic Array ............................................................................ 7
1.10 Generic Array Logic Devices ......................................................................... 8
1.11 Complex Programmable Array Logic ............................................................ 8
1.12 Field-Programmable Gate Array .................................................................. 10
1.12.1 Internal Architecture of Field-Programmable Gate Array ............. 10
1.12.1.1 Configurable Logic Blocks ............................................ 10
1.12.1.2 I/O Blocks ...................................................................... 12
1.12.1.3 Programmable Interconnects ......................................... 12
1.12.2 Design Flow of Field-Programmable Gate Array .......................... 12
1.12.3 Applications of Field-Programmable Gate Array in
Medical Imaging............................................................................. 14
1.13 Summary ...................................................................................................... 14
Bibliography ............................................................................................................ 15
1.1 INTRODUCTION
Programmable logic devices (PLDs) are integrated circuits, used for implementing
multiple functions that can be modified by programming and can be reconfigured by
the programmer according to the need of end user. These devices offer high perfor-
mance and flexibility to any digital system.
1
2 Advanced VLSI Design and Testability Issues
Digital designs mainly consist of logic functions made up of logic gates, flip-
flops, integrated circuits (ICs), and their combinations. Usually, this association
involves long interconnections, consumes large board space, and requires enough
design time for proper insertion and testing. Additionally, power consumption is one
of the major concerns in these designs. Apart from functional specifications such as
speed, power, and size, features such as programming yield, production complex-
ity, reprogramming, and debugging after manufacturing and software support are
also important concerns while evaluating chip performance. PLDs overcome these
difficulties and make the system more reliable and allow further modification at
later instant with reduced debugging effort. Desirable features such as reduced chip
size, faster execution, and low power consumption achieved by PLD attract users for
several applications. Available PLDs are based on either bipolar or complementary
metal oxide semiconductor (CMOS) technology.
Programmable devices are broadly classified into two types, i.e., factory program-
mable and field programmable, depending upon the flexibility of programmability.
One-time programmable devices that can only be programmed by the manufac-
turer at the factory and cannot be updated by the user at later instants are called
factory-programmable devices. After programming. i.e., after the metallization mask
step, these devices are tested at high temperature for stability and usually available
at wide frequency range. Examples of factory-programmable devices are read-only
memory (ROM) and mask-programmable gate array (MPGA). On the contrary,
field-programmable devices can be programmed a number of times and have much
programming flexibility with the user without involving semiconductor fab. These
devices require specialized equipment and software for programming. These devices
have comparatively small frequency range and are used for a wide variety of applica-
tions. Nowadays, PLDs are also available with security features and include security
bits in its standard. Figure 1.1 shows the hierarchical structure of PLDs.
PEEL
PROM PLA PAL GAL (Programmable
(Programmable (Programmable (Programmable (Generic Array Electrically
ROM) Logic Array) Array Logic) Logic) Erasable Logic
Devices)
.
Memory array
n inputs
. Decoder .
. 2n words ×m bits
. .
.
.
.
m outputs
memories. To have low power dissipation and better control over these memory orga-
nizations, row and column decoders are included in the memory architecture.
As the memory size is fixed by the manufacturer and cannot be updated by the
user, these devices are mainly used for high-density well-verified data such as firm-
ware updates.
• Programming flexibility
• Low-cost logic implementation
• Nonvolatile memory
• Better testing and debugging facility
However, EPROM devices are less cost-effective compared with PROM. Use of UV
ray for erasing requires special tool and programming skill. Again, reprogramming
Digital Design with PLDs 5
process increases the design time and slows down the overall system. EPROM is
used as bootstrap loader for storing computer bios and has also been used for micro-
controller for making video games.
F0 =∑ m(0,1,2,6) = A′B′ + BC ′
F = ∑ m(2,3,5,6,7) = AC + B
1
F = ∑ m(0,1,3,7) = A′B′ + BC
2
F = ∑ m(1,2,3,5,7) = C + A′B
3 (1.1)
Advantage of PAL structure over both PLA and PROM is that a register can be
incorporated with this device. These registers can be used for feeding back the out-
put signals toward input. Instead of using a number of min-terms multiple times to
form other functions, intermediate results are fed back to the input using the reg-
ister. This reduces the total number of min-terms required and simplifies the logic
implementation.
A B C
Fixed OR Array
A’B’
F0
BC’
AC’
F1
B
A’B’
BC F2
A’B F3
FIGURE 1.3 Internal structure for PAL. PAL, programmable array logic.
Digital Design with PLDs 7
In case of PAL and PROM devices, one of the arrays is programmable and other
is fixed, which limits its design flexibility and creates performance issue at high-
volume applications.
A B C
Programmable OR Array
A’B’
BC’
AC’
BC
A’B
F0 F1 F2 F3
• Array size grows as the implementation deals with complex logic functions.
This causes degradation in performance and programming density.
• These devices perform only combinational logic implementations.
speed of operation. This limits the uses of SPLD to SSI and MSI applications, deal-
ing with the small number of inputs and outputs. To increase the speed of operation
without increasing chip size exponentially, CPLDs are evolved.
CPLD structures consist of multiple SPLDs with the small number of program-
mable interconnection between them. Density of CPLDs is much larger compared
with that of the PAL devices. While PALs include hundreds of logic gates, CPLDs
contain thousand to ten thousand logic gates that make it suitable for medium com-
plex logic implementation. CPLDs also contain registers enabling them to be used
for sequential circuit design.
In general, a CPLD may be viewed to have a number of logic blocks, a dedi-
cated I/O block and a switch matrix interconnecting them. I/O block consists of I/O
elements that provide buffering for the input and output signals. The logic block
is formed by several logic elements known as macrocells. A macrocell comprises
AND and OR arrays, a dedicated flip-flop, and control signals for implementation of
the desired combinatorial or sequential functions. Fast routing between macrocells
leads to lower time delays within a logic block. Dedicated global clock lines lever-
age proper clock distribution among logic blocks and uniform timing properties.
The modern CPLDs contain 32–1700 macrocells enabling designers to implement
complex logic functions.
Major advantage of CPLD is that it starts operating as soon as the circuitry is
powered up. In-system programmability, JTAG for boundary scan and debugging,
interface to multiple logic levels (e.g., 1.8, 2.5, 3.3 V etc.), and nonvolatile nature of
programs add to the advantages of CPLDs.
There are several manufacturers of CPLDs such as Intel (Altera), Xilinx,
Lattice Semiconductor, Atmel, and Cypress Semiconductor. Architecture of Xilinx
CoolRunner-II CPLD shown in Figure 1.5 details the interconnection between the
main building blocks of the device. The Max II and Max V series of CPLDs from
Intel (Altera) and CoolRunner series of Xilinx are very popular in the industry.
PLA PLA
Advanced
I/O
I/O
Interconnect
Matrix
(AIM)
Figure 1.6 demonstrates the internal architecture of FPGA. The FPGA architecture
consists of three basic components, namely configurable logic blocks (CLBs), I/O
blocks (IOBs), and programmable interconnects. Apart from this, distributed block
memory, clock sources, and several intellectual property hardware elements (e.g.
embedded processors and peripheral drivers) are also available on-chip.
Interconnections
Horizontal
FIGURE 1.6 Architecture of Xilinx FPGA. FPGA, field-programmable gate array.
logic functions. Each CLB has look-up tables (LUTs), used for generation of logic
functions. Flip-flops are provided for sequential circuit implementation. A combina-
tion of LUTs and memory elements provides ROM functions. Based on the com-
plexity, CLBs also have carry structures, multiplexers, distributed RAM, and shift
registers. Distributed RAM can be used in single port/dual port mode with one clock
write and asynchronous read operations. The interconnection between LUT and reg-
isters in Xilinx Spartan 7 device is shown in Figure 1.7.
D Q
CE
CLK
S/R
Register
6 Input
LUT
Q6
D Q
Q5
CE
CLK
S/R
Register
FIGURE 1.7 Xilinx Spartan-7 FPGA LUT and register connections. FPGA, field-program-
mable gate array; LUT, look-up table.
12 Advanced VLSI Design and Testability Issues
TERMINATION OFF
Tristate
OE
Input
OBUF VCCO
IBUF OFF
PAD
Output
IBUF
IOB
For image and video processing applications, dedicated DSP slices are available
in specific devices containing multipliers and accumulators capable of operating at
frequencies in excess of 500 MHz. Preadder circuits greatly help in for symmetrical
filter (e.g., FIR filter) design.
The user specifies the design in textual (HDL) or graphical (schematic) form
using the design entry tool. Presynthesis simulation is carried out to verify the func-
tionality of the logic described in HDL. This step helps in amending design errors,
compatibility of parts, and specification issues.
The next step is compilation of the entered design comprising three substeps.
In the analysis phase, the entered design is translated into an intermediate format.
Different parts are linked together to generate the desired logic functions in the
synthesis phase. The tool places the various parts and routes them on the target
14 Advanced VLSI Design and Testability Issues
device. Also, the timing details are generated in the compilation phase. The tim-
ing analysis phase provides the clocking speed, intergate delay, critical path delay,
setup, and hold times. These data are used by the designers to define the speed of
the circuits. Postsynthesis simulation is run after synthesis of the design. It checks
timing problems, race, and hazard conditions and helps in determining suitable
clock frequency. A bit stream generated after place and route phase is used to
program the target FPGA using a programmer and programming cable (USB or
JTAG).
1.13 SUMMARY
In this chapter, we briefly discussed different PLDs and their uses for varying com-
plexity digital designs. These designs are categorized depending upon programming
flexibility and performance. Merits and demerits of different PLDs along with their
applications are outlined in each section. While factory-programmable logic devices
lack programming flexibility, their permanent storage capability suits them for most
of the firmware-based applications. On the contrary, field-programmable devices can
be reconfigured a million number of times and well suited for prototyping with very
less turnaround time. These devices save design time and production cost compared
to custom designs or ASIC designs. Application of FPGA in biomedical engineering
is enumerated for elaborated future study.
Digital Design with PLDs 15
BIBLIOGRAPHY
Intel Corporation, 2019. Accessed on 25 November 2019, https://www.intel.com/content/dam/
www/programmable/us/en/pdfs/literature/wp/wp-medical.pdf.
Intel Corporation, 2019. INTEL FPGAs. Accessed on 25 November 2019, https://www.intel.
in/content/www/in/en/products/programmable/fpga.html.
Z. Navabi, 2005. Digital Design and Implementation with Field Programmable Devices.
Boston, MA: Kluwer Academic Publishers.
V. A. Pedroni, 2004. Circuit Design with VHDL. Cambridge, MA: The MIT Press.
Rabaey, 2016. Digital Integrated Circuits: A Design Perspective. 2nd Ed. Pearson Education
India.
Sung-Mo-Kang, 2003. CMOS Digital Integrated Circuits, Analysis and Design. Tata McGraw-
Hill Education.
Xilinx, 2019. FPGAs & 3D ICs. Accessed on 25 November 2019, https://www.xilinx.com/
products/silicon-devices/fpga.html.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
2 Review of Digital
Electronics Design
Reena Chandel
Dr. Y.S Parmar University of Horticulture and Forestry
CONTENTS
2.1 Introduction to Digital Design ........................................................................ 18
2.1.1 Analog versus Digital Design ............................................................. 18
2.1.1.1 Analog Design ..................................................................... 18
2.1.1.2 Digital Design ...................................................................... 18
2.2 Number Systems ............................................................................................. 19
2.2.1 Binary ................................................................................................. 19
2.2.2 Octal and Hexadecimal ...................................................................... 19
2.3 Logic Families ................................................................................................ 19
2.3.1 Digital Integrated Circuit Characteristics........................................... 19
2.3.2 Resistor–Transistor Logic ................................................................... 20
2.3.3 Diode–Transistor Logic ...................................................................... 20
2.3.4 Emitter-Coupled Logic ....................................................................... 20
2.3.5 Transistor–Transistor Logic ................................................................ 22
2.3.6 Complementary Metal Oxide Semiconductor Logic .......................... 22
2.4 Combinational Logic ...................................................................................... 22
2.4.1 Boolean Equation................................................................................ 22
2.4.2 Introduction to Combinational Logic Circuits ................................... 24
2.4.3 Analysis and Design Procedure .......................................................... 24
2.4.4 Adder and Subtractor .......................................................................... 25
2.4.5 Decoder............................................................................................... 27
2.4.6 Encoder ............................................................................................... 27
2.4.7 Multiplexer and Demultiplexer ........................................................... 27
2.5 Sequential Circuits.......................................................................................... 29
2.5.1 Introduction to Sequential Circuits ..................................................... 29
2.5.2 Steps Involved to Design a Sequential Circuit.................................... 29
2.5.3 Types of Sequential Logic Circuits..................................................... 30
2.5.3.1 Comparison Table of Combinational and Sequential
Logic Circuits ...................................................................... 30
17
18 Advanced VLSI Design and Testability Issues
design, automation techniques are available, and thus, they are very simple to design
digital circuits [1]. Figure 2.2 illustrates the digital signal.
1001012 = [1 × 32 ] + [ 0 × 16 ] + [ 0 × 8 ] + [1 × 4 ] + [ 0 × 2 ] + [1 × 1]
1001012 = 3710
TABLE 2.1
Classification of Integrated Circuit
Level of Integration Number of Active Devices Per Chip
Smallscale integration (SSI) Less than 100
Medium-scale integration (MSI) 100–1,000
Large-scale integration (LSI) 1,000–100,000
Very large-scale integration (VLSI) More than 100,000
Ultralarge-scale integration (ULSI) More than 1 million
TABLE 2.2
Boolean Algebra Expression
Boolean
Expression Description Rule of Boolean Algebra
P+1=1 “P in concurrent with closed = CLOSED” Annulment
P+0=P “P in concurrent with open = P” Identity
P·1 = P “X in series with closed = P” Identity
P·0 = 0 “P in series with open = OPEN” Annulment
P+P=P “P in concurrent with P = P” Idempotent
P·P = P “P in succession with P = P” Idempotent
NOT P = P “NOT not P = P” Double negation
P+P=1 “P in concurrent with NOT P = CLOSED” Complement
P·P = 0 “P in succession with NOT P = OPEN” Complement
P+Q=Q+P “P in concurrent with Q = Q in concurrent with P” Commutative
P·Q = Q·P “P in succession with Q = Q in succession with P” Commutative
P + Q = P·Q “Invert and replace OR with AND” DeMorgan’s theorem
P·Q = P + Q “Invert and replace AND with OR” DeMorgan’s theorem
24 Advanced VLSI Design and Testability Issues
1. Commutative law:
The output of Boolean equation is same when the position of the input vari-
able is changed.
Example: Ps + Qs = Qs + Ps //(OR operator)
Ps * Qs = Qs * Ps //(AND operator)
2. Associative law:
The output of three-variable Boolean equation remains same while inter-
changing the brackets on both sides for addition and multiplication.
Example: Ps * (Qs * Rs) = (Ps * Qs) * Rs //(OR operator)
Ps + (Qs + Rs) = (Ps + Qs) + Rs //(AND operator)
3. Distributive law:
The output of addition of two products remains same as that of sum of
product of same variables.
Example: Ps + Qs Rs = (Ps + Qs) (Ps + Rs) //(OR operator)
Ps * (Qs + Rs) = (Ps * Qs) + (Ps * Rs) //(AND operator)
• The initial stage designer should clearly understand about the given prob-
lem statement.
• Determine the number of input line and output line required as per given
specification.
• Create a truth table in accordance with logic of given problem statement.
• Simplify the Boolean expression for each output by using K-map reduction
or any other reduction method.
• Sketch the digital logic circuit based on simplified Boolean expression, and
verify the output logic through truth table.
S = ( A XOR B )
C = ( A AND B )
TABLE 2.3
Truth Table
Input Output
Ain Bin Sout Cout
0 0 0 0
0 1 1 0
1 0 1 0
1 1 0 1
26 Advanced VLSI Design and Testability Issues
Similarly, for subtractor, AND gate is replaced with NAND gate, and truth table will
be changed as per the Boolean equation. Full adder is the logic circuit to add three
bits, namely two inputs and on carry of the previous output (Figures 2.10 and 2.11
and Table 2.4). Full adder is shown in Figures 2.10 and 2.11.
C = AB + BX + AX
Full adder circuit is used to achieve the adding of three inputs, including carry-in of
the previous state; the output of the circuit will be verified through predefined pos-
sible combination of inputs. The grouping of dual half adder can achieve the whole
TABLE 2.4
Truth Table of Full Adder
Input Output
Ain Bin Xin Sout Cout
0 0 0 0 0
0 0 1 1 0
0 1 0 1 0
0 1 1 0 1
1 0 0 1 0
1 0 1 0 1
1 1 0 0 1
1 1 1 1 1
Review of Digital Electronics Design 27
full adder circuit; the first half adder gets the addition of first two inputs, and then
the output of first half adder and remaining input get added and will generate the
summation and carry output of the full adder.
2.4.5 DECODER
In digital electronics, information is represented in binary codes, and decoder is
used to convert the coded information from one form to another. Decoder is a com-
binational circuit for converting n bit information to 2^n discrete outputs as shown
in Figure 2.12.
2.4.6 ENCODER
Encoder is a combination circuit used to convert the binary information in 2^n lines
to n output lines. Figure 2.13 shows the octal to binary encoder (Table 2.5).
TABLE 2.5
Truth Table of 8:3 Encoder
Inp7 Inp6 Inp5 Inp4 Inp3 Inp2 Inp1 Inp0 Xout1 Xout2 Xout3
0 0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 1 0 0 0 1
0 0 0 0 0 1 0 0 0 1 0
0 0 0 0 1 0 0 0 0 1 1
0 0 0 1 0 0 0 0 1 0 0
0 0 1 0 0 0 0 0 1 0 1
0 1 0 0 0 0 0 0 1 1 0
1 0 0 0 0 0 0 0 1 1 1
FIG
GUR
RE 2.15 1:4 Demulttiplexxer.
• Latches
• Flip-flops
Latchet are very basic storage elements that operate with signal levels rather than
signal transitions.
TABLE 2.6
Difference Between Combinational and Sequential Circuits
Combinational Circuits Sequential Circuits
Output is always depending on present input Output is always depending on present state and
at the instantaneous of time. previous input.
There is no need of memory unit. Memory unit is required to store the previous state output.
This circuit is faster. This circuit is slower.
It is easy to design this circuit. It is difficult to design this circuit.
Examples: half adder, full adder, magnitude Examples: flip-flop, register, counter, clocks, etc.
comparator, multiplexer, demultiplexer, etc.
Latches controlled by a clock transition are called flip-flops; thus the flip-flops are
clock edge sensitive. Different types of flip-flops are as follows:
2.6.1 SR FLIP-FLOP
SR flip-flop is set–reset the output of SR flip-flop and operates on positive- or
negative-edge clock transition. The circuit diagram of SR flip-flop is shown in
Figure 2.19 (Table 2.7).
2.6.2 D FLIP-FLOP
The D flip-flop is called delayed flip-flop that means one of the input will reach late
to gate as compared with another input (Figure 2.20 and Table 2.8).
2.6.3 JK FLIP-FLOP
JK flip-flop functions on positive or negative clock change. JK flip-flop is the better-
quality version of SR flip-flop taking care of unknown conditions in SR flip-flop
(Figure 2.21 and Table 2.9).
TABLE 2.7
Truth Table of SR Flip-Flop
Input Output
S R Q Qbar
0 0 (NC) (NC)
0 1 0 1
1 0 1 0
1 1 Unknown value Unknown value
TABLE 2.8
Truth Table of D Flip-Flop
Input Output
D Q
0 0
1 1
2.6.4 T FLIP-FLOP
T flip-flop is toggle flip-flop and is basic version of JK flip-flop. T flip-flop functions
on positive or negative clock change. The circuit diagram of T flip-flop is shown in
Figure 2.22 (Table 2.10).
Review of Digital Electronics Design 33
TABLE 2.9
Truth Table of JK Flip-Flop
Input Output
J K Q Qbar
0 0 Previous state (NC)
0 1 0 1
1 0 1 0
1 1 Toggle state
TABLE 2.10
Truth Table of T Flip-Flop
Input Output
T Q
0 1
1 0
2.7 COUNTERS
Digital circuit counters are used for the counting purpose. They count and store the
number of events occurred or the amount of period a specific occurrence has hap-
pened using the clock signal.
2.7.3 REGISTERS
Registers are the clocked sequential circuits designed using flip-flops and also con-
sist of combinational circuits. Flip-flops are used to store the 1-bit information and
combinational circuits to perform some simple tasks such as generating the signal for
resetting the register (Figure 2.25).
2.8 MEMORY
A memory is referred as any device that is capable of storing information that can be
retrieved as and when required [4]. Computer information is stored in digital format.
A memory is neither a sequential circuit as memories are not clocked nor a combina-
tional circuit as output of memory is dependent on previous input.
Review of Digital Electronics Design 35
• ROM
• Programmable read-only memory (PROM)
• Erasable programmable read-only memory (EPROM)
• Electrically erasable programmable read-only memory (EEPROM)
• Flash memory
36 Advanced VLSI Design and Testability Issues
• SRAM: Static RAM stores the bit information using six transistors, typi-
cally MOSFET, and is more expensive in comparison with the other RAM.
• DRAM: Dynamic RAM stores the bit information using transistors, typi-
cally MOSFET, and pair of capacitors. In this memory, transistor acts as
a switch, and information is stored as charge in capacitor. These types of
memories are less expensive.
Both SRAM and DRAM are volatile memories, which lose their information when
the power is removed.
2.8.3 FLASH
Flash memory is used for storage and transfer of information between computers. It
is a nonvolatile memory and can be electronically erased and reprogrammable. The
types of flash memory are NOR and NAND. NAND is better for serial data access.
Data read is faster in NOR, but data erase and write are faster in NAND.
TABLE 2.11
Optical and Magnetic Storage Devices
Optical Storage Devices Magnetic Storage Devices
1 The storage of this device has been done as The storage of this device has been done as a
a patterned image. magnetic form.
2 To read and write, LASER is required. To read and write, heads are required.
3 Storage will not be disturbed due to Storage will be disturbed due to magnetic field.
magnetic field.
4 As technologies are improved, these can be These are always readable and rewritable.
readable, writable, and rewritable.
5 Drivers are required. These contain built-in drivers.
6 These are easy to carry and safe. These are difficult to carry.
Review of Digital Electronics Design 37
REFERENCES
1. J. F. Warkerly, Digital Design: Principles and Practices, Pearson, USA, 2013.
2. T. L. Flyod, Digital Fundamentals, Pearson, USA, 2013.
3. S. Brown, Z. Vranesic, Fundamental of Digital Logic Design with VHDL, McGraw
Hill, USA, 2007.
4. M. Morris Mano, M. D. Ciletti, Digital Design: With An Introduction to Verilog HDL,
Pearson, USA, 2013.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
3 Verilog HDL for Digital
and Analog Design
Ananya Dastidar
College of Engineering and Technology
CONTENTS
Chapter Outline ........................................................................................................ 40
3.1 Introduction to Hardware Description Languages ......................................... 40
3.2 Verilog for Digital Design .............................................................................. 41
3.2.1 Verilog Language Basics .................................................................... 41
3.2.1.1 Keywords ............................................................................. 41
3.2.1.2 Comment Line ..................................................................... 41
3.2.1.3 Whitespaces ......................................................................... 42
3.2.1.4 Identifier............................................................................... 42
3.2.1.5 Variables .............................................................................. 42
3.2.1.6 Vector Data .......................................................................... 44
3.2.1.7 Numbers or Constant Values ............................................... 45
3.2.1.8 Parameter ............................................................................. 45
3.2.1.9 Sequential Statements .......................................................... 45
3.2.2 Additional Constructs ......................................................................... 49
3.2.2.1 time value (#) ....................................................................... 49
3.2.2.2 @ (sensitivity_list) ............................................................... 49
3.2.2.3 Generate ............................................................................... 50
3.2.2.4 Gate Primitives .................................................................... 50
3.2.2.5 Tristate Gates ....................................................................... 50
3.2.2.6 Switch-Level Primitives ....................................................... 51
3.2.3 Verilog Module Description ............................................................... 51
3.2.3.1 Ports ..................................................................................... 52
3.2.4 Operator Types ................................................................................... 52
3.3 Modeling Types .............................................................................................. 54
3.3.1 Behavioral or Algorithmic Model ...................................................... 54
3.3.1.1 Blocking and Nonblocking Statements ................................ 55
3.3.2 Dataflow or Register Transfer-Level Model ....................................... 56
3.3.3 Gate-Level or Structural Model.......................................................... 57
3.3.4 Switch-Level Model ............................................................................ 58
3.3.5 Mixed Model ...................................................................................... 59
3.4 User-Defined Primitives ................................................................................. 59
3.5 Test Bench....................................................................................................... 60
39
40 Advanced VLSI Design and Testability Issues
CHAPTER OUTLINE
This chapter deals with the basics of Verilog hardware description language (Verilog
HDL) and its special features toward implementation of digital and analog designs.
It starts with a background on the emergence and importance of HDLs with a brief
introduction to Verilog HDL. The features of Verilog for digital design include
data types, model types, and so on. The next section discusses some programming
concepts for analog electrical circuits using the features of Verilog-A. This chapter
ends with summarizing the limitations and applicability of the Verilog Family of
Languages.
The past few decades have seen the evolution of digital circuits from simple gates
to complex system-on-chips (SoCs). From tubes to transistors to integrated circuits
(ICs), design of digital circuits has come a long way. With the rise of very large-scale
integrated circuit (VLSI) technology, complicated digital implementation of an IC
not only became conceivable but also paved the way for automation in chip design.
Depending on the functionality of the IC, they may be termed as digital IC, analog
IC, or mixed-signal IC. Due to the complexity involved in IC design, verification of
the correctness of the design under different conditions becomes a challenge, and the
presence of millions of transistor on a single substrate makes detection of fabrica-
tion faults an additional challenge. So a software solution for design specification,
verification, and testing becomes essential to meet the requirements of the industry
in terms of economics and technical viability of a product. Chip designers have thus
turned to different tools such as electronic design automation (EDA) for both soft-
ware and hardware designs in order to bring down the time-to-design and in turn the
time-to-market for both circuit simulators and programming languages. These tools
have proved to be indispensible in high complexity circuit design by allowing the
transformation of the design specification to mask layout for transistor fabrication,
verification, and testing.
Shorter comment lines use double slash (//), whereas comments that extend to more
than one line use a pair of slash and asterisks (/*….*/).
3.2.1.3 Whitespaces
In order to increase the readability of the code, whitespaces are used. Whitespace
operators include space, tab, or newline as separators while writing Verilog codes.
3.2.1.4 Identifier
Names of variables and other elements in a Verilog code are represented by identi-
fiers where a valid identifier can be a combination of alphabet and numbers including
underscore and the special symbol $. It cannot begin with a numbers and must not
be a Verilog keyword.
3.2.1.5 Variables
Variables represent data types in Verilog as a net or register where net indicates the
interconnection of two of more points of a circuit and the registers are path compo-
nents that are capable of storing intermediate values.
3.2.1.5.1 Net
The net data type includes wire (most commonly used), wand (wired-AND), wor
(wired-OR), tri, supply0 (power supply connection), supply1 (power supply connec-
tion), etc. We must declare a net as a signal when it models ports of a design module
or when a net type variable is present on the left side of a continuous assignment
statement (assign). Default value of net is high impedence (z).
wire
The net data type wire represents the physical wire, and it models the
interconnection between gates and other logic modules. A wire cannot store
any value and so it may not be used inside a function or a block. An assign
statement and the output of a design module drive a wire. Apart from the
general wire, there are different wire types such as the wand (wired-AND)
that inserts a AND gate in the interconnection point and wor (wired-OR)
that inserts an OR gate in the interconnection points. The tri (three-state) is
used for multiple drivers.
Verilog HDL for Digital and Analog Design 43
Example
wire [3:0] W; // an 4 bit vector (4 wires make up a 4 bit bus)
wire A, B; // simple wire
wor R; //insert a or gate at the interconnection
assign R=A;
assign R=B;// R= A or B
tri [1:0] T;// T can hold the value 0, 1 or Z
Example
supply1 vdd;
supply0 gnd;
3.2.1.5.2 Register
The register on the other hand holds the value of a procedural assignment statement
until the next event triggers it. It is not a physical register but implies a variable-type
register, and it finds use only in functions and procedural blocks. There are different
register data types such as reg used for logic description, integer for loop calcula-
tions, real for system module design, and time for test benches value storage, of
which the reg is the most frequently used register data type.
reg
The default value of reg variables and unconnected registers has “x” as
initial value at the beginning of simulation, and it can be used to realize a
combinational as well as a sequential logic. Specifying the size of the reg at
the time of declaration is encouraged; otherwise, by default, it is 1-bit and
an unsigned number in arithmetic expressions.
Example
reg r; // a single bit register variable
reg [5:0] r; // a 6-bit vector;
reg [7:0] s, t; // two 8 bit variables.
integer
They are used in general-purpose variables especially as index values
in loops, constants, and parameters. Data are stored as signed numbers
by default, but the default size of integer is 32 bits. In case the integer
holds a constant, the minimum width needed is adjusted by the synthesis
tool.
44 Advanced VLSI Design and Testability Issues
Example
integer i; // by default 32-bit integer
assign i=55; // a is adjusted as a 6-bit variable
real
It is used to store floating-point numbers where rounding off to nearest
integer occurs when a real number is assigned to an integer variable.
Example
real eox, eps;
initial
begin
eox = 3.97;
eps = 8.854 e-14;
end
// now let us assign real value to an integer variable
integer int;
initial
int = eps; // int gets the value 8
time
It is used in test bench to hold the time of simulation. It is a nonsynthe-
sizable construct and used only for simulation purposes. The system func-
tion “$time” gives the current time of simulation.
Example
time present_tme;
initial
present_tme = $time;
All Verilog data types under the value set can have the following pos-
sible values: logic 0 or false conditions (0), logic 1 or true conditions (1),
unknown logic value (x), or the high impedance state (z). Time is a nonsyn-
thesizable Verilog construct.
Example
input [3:0] a;
// a is a 4 bit vector with a[3] as MSB and a[0] as LSB
output reg [0:4] x;
/* x is an output port register having data type reg which is
a 5 bit vector with x [0] as MSB and x [4] as LSB */
Verilog HDL for Digital and Analog Design 45
Syntax
<size> <radix> <value>
Example
2’b01 // 2 bit binary
8’hFF // 8 bit hexadecimal …1111 1111
3.2.1.8 Parameter
For customization of a design module during the instantiation process, parameter
construct can be used. The size of the parameter cannot be defined, but it is decided
from the value of the constant assigned to the parameter.
Example
parameter X=5, Y=10;
parameter HOT=2’hFF, WARM=2’hAA, TREPID=2’h55, COLD=2’h11;
Syntax
begin
sequential_stmt_s1;
sequential_stmt_s2;
…
sequential_stmt_sn;
end
The begin … end keywords are omitted when only a single procedural statement is
present inside a block.
46 Advanced VLSI Design and Testability Issues
TABLE 3.1
Variation of if … else Statement
Types of if … else Statement Examples
if (<expression1>) if (a > b)
sequential_stmt; z = 1’b1;
if (<expression1>) if (a > b)
sequential_ stmt; z = 1’b1;
else else
sequential_ stmt; z = 1’b0;
if (<expression1>) if (a > b)
sequential_ stmt; z = 2’b10;
else if (<expression2>) else if (b > a)
sequential_ stmt; z = 2’b01;
else if (<expression3>) else
sequential_ stmt; z = 2’b11;
else default_ stmt;
3.2.1.9.2 if … else
It is a sequential statement that can occur alone or as a group of statements inside
the “begin … end” block depending to implement conditional flow of data in a
design module. Depending on the type of conditional expression used, they may be
expressed as any one of the following ways as seen in Table 3.1. If there are multiple
sequential statements inside the if … else statement, then they must be enclosed
within a begin … end construct.
When using if … else statements to model combinational circuit, one must be
careful to define the outputs for all possible input conditions; otherwise, the circuit
will synthesize a latch.
3.2.1.9.3 case
We can have a single statement or a group of statements within the “begin … end”
construct inside a case statement. It can be used to replace a complex “if … else”.
Syntax
case (<case_expression>)
poss_expr1: sequential_ stmt;
poss_expr2: sequential_ stmt;
...
poss_exprn: sequential_ stmt;
default: default_ stmt;
endcase
Example
……..
parameter RED=1’d1, YELLOW=1’d2, GREEN=1’d3;
……..
case (colour)
RED: traffic=1’b0;
YELLOW: traffic=1’b0;
GREEN: traffic=1’b1;
default: traffic=1’b0; // default statement is used to cover
all unassigned conditions
endcase
………
Case has two variants—“casez” and “casex.” While all the “z” values are treated as
don’t cares in casez, both the “x” and “z” values are treated as don’t cares in casex
statement.
Example
reg [2:0] points;
integer score;
casex (points)
3’b1xx : score = 3;
3’bx1x : score = 2;
3’bxx1 : score = 1;
default : score = 0;
endcase
// If points is “3’bxz1”, the third expression will give match, and the score will be 1.
3.2.1.9.4 Loop
There are four types of loops used for implementing repetitive evaluation of expres-
sions in Verilog HDL—while, for, repeat, and forever. As long as the <expression>
is true, the while loop continues.
Syntax
while (<expression>)
sequential_ stmt;
Example
while (a<b)
a=a+1;
The “for” loop consists of three parts—an initial value (initial_expr), an expression
to see if the terminating condition is met (check_expr), and an update procedural
assignment statement that will change the value of the control variable (update_
expr). It executes as long as the expression check_expr is true.
48 Advanced VLSI Design and Testability Issues
Syntax
for (initial_expr; check_expr; update_expr)
sequential_statement;
Example
for (i=0; i<=50; i=i+1)
a[i] = 2’b11;
In order to execute a loop for a definite number of times, the construct “repeat” is
preferred.
Syntax
repeat (<rep_value>)
sequential_statement;
Example
repeat (20)
#10 clock = ~clock;
Syntax
forever
sequential_statement;
Syntax
module (port description);
………………..
Verilog HDL for Digital and Analog Design 49
//assign statement
assign x_out = a_in and b_in;
assign y_out = c_in + d_in;
assign z_out = f [5];
//procedural statements
………….
endmodule
In the above example, x_out, y_out, and z_out are of the type net, whereas a_in,
b_in, c_in, d_in, and f may be of the type net or reg. The assign statement imple-
ments both combinational and sequential logic as can be seen in the following:
Example
a=#5 ~b;// not of b is assigned 5 time unit later
#10 c = a;/* this statement is executed after 10 time units
elapses after the execution of the first statement */
3.2.2.2 @ (sensitivity_list)
An @ in Verilog indicates that the block will be executed when a trigger occurs on
any one of the sensitivity list parameters.
Example
input a_in, b_in, clock;
output c_out, d_out;
always @(posedge clock)
begin
c_out = a_in && b_in;
d_out=a_in || b_in;
end
Here, c _ out and d _ out get updated whenever a positive edge on the clock
is encountered, i.e., clock is the sensitivity list parameter. The @(*) eliminates the
problem of incomplete event expression list by adding all nets and variable that are
read by a statement inside the block. In the previous example, if always @(posedge
clock) is replaced with always @(*), the target variables c _ out and d _ out will
be updated when any one of the clock a _ in or b _ in triggers.
50 Advanced VLSI Design and Testability Issues
3.2.2.3 Generate
In order to dynamically generate the Verilog code prior to simulation, the generate
statement is used that allows us to create generic design, which can be updated as per
our requirement. For example, we can design an 8-bit parallel adder from an n-bit
parallel adder design code. The generate block is enclosed by the keywords generate
and endgenerate. The genvar is a special variable associated with the generate loop.
Example
module xnor_bitwise (z, a, b);
parameter N = 8;
input [N-1:0] a, b;
output [N-1:0] z;
genvar g;
generate for (t=0; t<N; t=t+1)
begin xnorgen
xnor XNG (z[t], a[t], b[t]);
end
endgenerate
endmodule
This code generates an 8-bit xnor capable of performing bitwise xnor operations on
a and b while providing the 8-bit result on z.
Example
module my_gates( XZ ,Nin, Ain, Oin, Xin)
input Nin, Ain, Oin, Xin;
output XZ;
wire NZ,AZ, OZ;
not N1 (NZ, Nin);
and A1(AZ, Nin, Ain);
or O1(OZ, AZ, Oin);
xor X1(XZ,OZ, Xin);
endmodule
the following tristate gate primitives—bufif1 (active high control tsb), bufif0 (active
low control tsb), notif1 (active high control tsi), and notif0 (active low control tsi).
Example
bufif1 buf1 (ZO, AIN, EN);
Example
Design Module
Input Output
(Functionality)
The functionality is internal to the design, and the designer can alter as per
requirement and for optimization. The fundamental building block of Verilog HDL
is the module (analogous to functions in C). We can have one or multiple modules in
a Verilog code (as we can have one or many functions in C), but a module descrip-
tion cannot contain other modules inside it, but one module may be instantiated by
another module (similar to functions in C).
Module Description
module <module_id> (<module_port_list>);
...
<module_functionality>
...
...
endmodule
The module description begins with the reserved word module followed by a mod-
ule_id, that is, any user-defined identifier. The module_port_list includes the input
and output ports. The module_ functionality includes the functionality of the design
module that determines how the output changes its states or values in response to any
change occurring at the input ports. The construct endmodule denotes the comple-
tion of the design specification process.
3.2.3.1 Ports
The ports are represented with a unique identifier, and the direction of the ports of a
design module is represented using the Verilog constructs input, output, and inout to
represent input signals, output signals, and bidirectional signals. While the input port
is of a variable of type net or reg, the output port is configured as net of the data type
wire, reg, wand, wor, or tri type, with the default being wire type. The inout port is
of a variable of type net.
Example
module first_eg (a,b,c,d);
input c;
input [1:0] d;
output reg a;// the output port is configured as a register
output b;// the output port b is configure as a wire by
default
// continuous assignment statements
…………….
// procedural assignment statement
………………….
endmodule
true or false (Boolean value), bitwise operators operate on bits and return a bit value.
Reduction operators on the other hand work on single-word operands and return a
single bit as output, and they operate on all the bits within the word. Table 3.2 shows
the various operators and their use in Verilog HDL.
TABLE 3.2
Operator Types in Verilog HDL
Operators Uses Examples
Arithmetic Operators
+ Unary (sign) positive +a
− Unary (sign) negative −b
+ Binary plus (addition) a+b
− Binary minus (subtraction) a−b
* Multiplication a*b
/ Division a/b
% Modulus a%b
** Exponentiation a**3
Logical Operators
! Logical negation !a
&& Logical AND a&&b
|| Logical OR a||b
Bitwise Operators
& Bitwise AND
| Bitwise OR
wire a, b, f1,f2;
~& Bitwise NAND
assign f1 = ~a|b;
~| Bitwise NOR
assign f2 = a^b;
^ Bitwise exclusive-OR
~^ Bitwise exclusive-NOR
Relational Operators
!= Not equal
== Equal a!=b
>= greater or equal a == b
<= less or equal a<b
> Greater a <= 0
< Less
Reduction Operators
& Bitwise AND assign a =3’b011;
| Bitwise OR assign b =3’b110;
~& Bitwise NAND assign f1 = ^a;
~| Bitwise NOR // f1 = 0
^ Bitwise exclusive-OR assign f2 = &(a^b);
~^ Bitwise exclusive-NOR // f2 = 0
(Continued)
54 Advanced VLSI Design and Testability Issues
infinite loop. The “initial” block consists of multiple statements grouped inside a “begin
… end” structure. Multiple initial blocks execute concurrently, and it specifies the stimu-
lus provided to the design and the outputs displayed for a design under test (DUT).
The “always” block also has a group of statements bounded by “begin … end,”
and it models an indefinite activity of a digital logic design, e.g., clock signals are
generated continuously. Like the initial block, multiple always blocks execute con-
currently. The always block is associated with the @ (event_ expression) for both
combinational and sequential designs. The event_expression may be associated with
any type of variable such as reg or wire, but within an initial and always blocks, the
type of variable to be assigned must be of the type reg. The reason is that in case of
sequential designs, the block triggers for event conditions and remains idle for other
conditions. In order to model this, the variable must be able to retain its past value
that is possible only by the use of reg-type variables.
Mixing of blocking and nonblocking statements within an always block is not a rec-
ommended coding practice. Moreover, one must be careful not to assign the same
variable as the target of both blocking and nonblocking assignments.
a = a + 1;
a <= b;
Using the above two expressions in the same always block is not permissible.
Model a 2-to-1 mux using behavioral model that has two inputs represented by a
2-bit vector I, a select line represented by S, and a 1-bit output represented by F. The
following code represents the 2-to-1 mux:
Example
module mux_2to1(F, S, I);
input [1:0] I;
input [1:0] S;
output reg F;
always @(S or I)// can be replaced with @(*)
begin
case (S)
1’b0: F=I[0];
1’b1: F=I[1];
default F=1’bx;
endcase
end
endmodule
assign F = I[S];
if (S) F = I[1];
else
F = I[0];
end
assign w = s? b: a;
Example
module mux_2to1(
input A, B;
input SEL;
output Z
);
wire t0;
wire t1, t2;
not N (t0, SEL);
and A0 (t1, SEL, A);
and A1 (t2, SEL, B);
or R (Z, t1, t2);
endmodule
A
SE t1
t0 Z
t2
B
In the aforementioned method, we have used one not gate, two and gates, and one or
gate. Designer has the knowledge of the internal functionality of the primitive prior
to instantiation in any circuit design. These primitives connect in a manner that it
realizes a multiplexer, and this type of modeling is referred to as structural modeling.
Example
Example
module mux_2to1 (F, S, I);
input S;
input [1:0]I;
output F;
Verilog HDL for Digital and Analog Design 59
wire SBAR;
not (SBAR, S);
cmos cmos_a (F, I[0], SBAR, S);
cmos cmos_b (F, I[1], S, SBAR);
endmodule
Example
module mixed_andor (input a, b, c, d, output y);
assign y = (a & b) | (c & d);
endmodule
module mux_mixed (input A, B, SEL, output Z);
wire t0;
not U1 (t0, SEL);
mixed_andor U2 (A, SEL, t0, B, Z);
endmodule
The UDP model does not model timing or process technology but models func-
tionality only. A functional block with a single output only can be modeled using
60 Advanced VLSI Design and Testability Issues
UDP. For unspecified cases, the output is set to “x,” whereas the don’t cares in inputs
are represented with a question mark (?).
Example
primitive udp_mux2to1 (F, A, B, S);
input S, A, B;
output F;
table
// S A B F
0 0 0 : 0;
0 0 1 : 0;
0 1 0 : 1;
0 1 1 : 1;
1 0 0 : 0;
1 0 1 : 1;
1 1 0 : 0;
1 1 1 : 1;
endtable
endprimitive
// S A B F
0 0 ? : 0;
0 1 ? : 1;
1 ? 0 : 0;
1 ? 1 : 1;
In sequential circuits, some output states are preserved for certain input conditions—
in case of SR flip-flops, when S=0 and R=0, then the next state Qn+1= Qn. A “-” is
used to represent this condition in the next state column in the lookup table of UDP.
Example
module tb_mux;
// Declaring Inputs
reg A;
reg B;
reg S;
// Declaring Outputs
wire F;
// Instantiate the Design Under Test (DUT)
mux2to1 dut (
.A(A), //explicit association
.B(B),
.S(S),
.F(F)
);
initial begin
// Apply Inputs
A = 0;
B = 0;
S = 0;
$monitor ($time, “A=%b, B=%b, S=%b, F=%b”, A, B, S, F);
// Wait 50 ns
#50;
//Similarly apply Inputs and wait for 50 ns
A = 0; B = 0; S = 1; #50;
A = 0; B = 1; S = 0; #50;
A = 0; B = 1; S = 1; #50;
A = 1; B = 0; S = 0; #50;
A = 1; B= 0; S = 1; #50;
A = 1; B = 1; S = 0; #50;
A = 1; B = 1; S = 1; #50;
end
endmodule
The input ports of the module have been declared as reg, as they appear on left-
hand side of the expressions appearing inside the initial block, whereas the output
ports are declared as wire. The $monitor ($time, “A=%b, B=%b, S=%b, F=%b”,
A, B, S, F) will generate the current simulation time, i.e., value of A, B, S, and
F in the simulation window in the following way (display format depends on the
simulator):
62 Advanced VLSI Design and Testability Issues
While $monitor will display the value of its variable, the moment any one of the
listed variables changes its value, $display will display the value of the listed vari-
able irrespective of any change occurring in the value of the variables. The initial
block in the test bench will execute only once, and the use of blocking statements
will help for proper implementation.
Writing a Verilog with proper indention with suitable comments is essential in
industry design. The following code uses behavioral style of modeling using the
procedural block always that in turn includes the sequential statement case:
Example
Since the previous example has defined res for all input conditions inside the case
statement, on synthesis, it will generate a combinational design of the ALU.
analog
V(out) <+ 5.0 + V(in);
// Flow signal-flow systems
analog
V(out) <+ 5.0 + V(in);
In order to model a resistor with node terminals n1 and n2 having node voltages
V(n1) and V(n2) and branch voltage V(b), the resistance is given as R Ω, whereas the
current flowing in the branch is I(b).
64 Advanced VLSI Design and Testability Issues
Example
//Resistor 1
The module description here also starts with a module id of the component (resistor)
followed by the nodes. The keyword electrical represents voltage to be the potential
and current to be the flow. The flow is positive if the flow is into the component.
Parameter R of the components refers to the resistance value in order to instantiate
the component. If the value of parameter is not specified, it is assigned the default
value, and being a constant, this value remains fixed. The branch named b refers that
a branch exists between n1 and n2. Here, the branch voltage V(b) is considered to be
equal to V(n1) − V(n2) (as per the orders of the terminals mentioned). If node voltage
of t1 is greater than that of t2, the branch voltage is positive. The current flowing in
the branch is I(b). In this case, it is positive as the current flows from n1 to n2. This
completes our topological description. In order to describe the branch flow relation,
we must specify a relation between the voltage and current of the branch. The analog
keyword denotes an analog process. Multiple statement would require a “begin …
end.” The behavior of a branch is specified by a contribution statement and a contri-
bution operator (<+).
It states that “the voltage on branch res must equal the product of current through
that branch and resistance R.” The resistor can also be modeled without defining a
branch explicitly.
//Resistor 2
Let us look at some simple circuit models that are implemented using Verilog-A HDL.
// Model a resistor with a dc source in series
Example
module port (n1, n2);
electrical n1, n2;
parameter real dc=0;
Verilog HDL for Digital and Analog Design 65
Example
analog begin
I(n1,n2) <+ V(n1,n2)/r;
I(n1,n2) <+ idt(V(n1,n2))/l;
I(n1,n2) <+ c*ddt(V(n1,n2));
end
Example
Verilog-A in collaboration with Spice and Verilog HDL has proved to be an excel-
lent synthesis and simulation tool for IC design engineers. Specification, simulation,
and verification of mixed-signal designs have become easier. They allow portability,
compactness, robustness, and rapid prototyping in hardware implementation before
actual tape-out.
66 Advanced VLSI Design and Testability Issues
3.8 SUMMARY
Verilog HDL provides many constructs, but many of them may not be synthesiz-
able, such as initial block, time, testbench, switch-level implementations, UDP, and
so on in specific synthesis tools. When modeling combinational designs, we must
keep in mind to give explicit output for every input conditions; otherwise, we may
get a sequential circuit. To produce synthesizable designs, we can use a netlist of
Verilog built-in gate primitives, continuous assignments, and behavioral statements.
We must avoid both feedback loops of any kind during combinational logic design
and mixing of blocking and nonblocking assignments. So when designing a synthe-
sizable logic, we must be careful to make use of only synthesizable constructs as
available in the simulator under use. Once a synthesizable design is ready, testing of
the design by giving signals to the synthesized design using a field-programmable
gate array (FPGA) or application-specific integrated circuit (ASIC) or complex pro-
grammable Logic Devices (CPLD) can be performed.
A VLSI design is based on design verification test, and the Verilog Family of
Languages provides a solution in handling each of these steps by minimizing the
complexity of each step. The Verilog Family of Languages including Verilog HDL,
Verilog-AMS, and Verilog-A has proved to be a popular EDA tool for both industry
and academics. An excellent choice for both digital and analog designs and their
features of simulation, synthesis, verification, and testability make them an excel-
lent choice for IC design engineers. These languages allow the designer to imple-
ment designs in various abstraction levels. Compactness, closeness to C, and wide
acceptance in industry are some major takeaways for Verilog HDL. Describing the
behavior of electrical circuits, development of analog models, and facilitating model
development using Verilog-A has proved to be quite useful. In order to handle the
interface between real environment signals and the advantages of digital signal pro-
cessing, the Verilog-AMS is a welcome addition for circuit designers.
REFERENCES
Ashenden, P. J. and Lewis, J. 2008. VHDL-2008 Just the New Stuff. Burlington, MA: Morgan
Kauffmann.
Navabi, Z. 1999. Verilog Digital System Design. New York: Mc Graw Hill.
Palnitkar, S. 2003. Verilog HDL: A Guide to Digital Design and Synthesis. Upper Saddle
River, NJ: Prentice Hall PTR.
Sutherland, S., Davidman, S. and Flake, P. 2004. System Verilog for Design. New York:
Springer.
4 Introduction to Hardware
Description Languages
Shasanka Sekhar Rout
GIET University
Salony Mahapatro
Asiczen Technologies
CONTENTS
4.1 Introduction .................................................................................................... 68
4.1.1 Basic Principles of Hardware Description Languages ..................... 68
4.1.2 Basic Concepts of Hardware Description Languages ...................... 69
4.1.2.1 Timing and Concurrency ................................................. 69
4.1.2.2 Hardware Simulation Process .......................................... 70
4.1.3 Hardware Description Language Design Tool Suites....................... 70
4.1.4 Types of Hardware Description Languages ..................................... 70
4.1.5 Design Using Hardware Description Language ............................... 71
4.1.6 Hardware Description Languages for Digital Design ...................... 71
4.1.7 Very High-Scale Integrated Circuits Hardware Description
Language .......................................................................................... 71
4.1.7.1 Entity Declaration ............................................................ 72
4.1.7.2 Architecture ..................................................................... 72
4.1.8 Verilog .............................................................................................. 74
4.1.8.1 Lexical Tokens ................................................................. 74
4.1.8.2 Data Types ....................................................................... 75
4.1.8.3 Timescale ......................................................................... 75
4.1.8.4 Continuous Assignments ................................................. 76
4.1.8.5 Procedural Assignment .................................................... 76
4.1.8.6 Procedures: Always and Initial Blocks ............................ 77
4.1.9 Test Bench in Verilog........................................................................ 77
4.1.10 System Verilog .................................................................................. 78
4.1.11 Test Bench Structure for System Verilog.......................................... 78
4.1.11.1 Design under Test............................................................. 79
4.1.11.2 Transaction ....................................................................... 79
4.1.11.3 Interface ........................................................................... 80
4.1.11.4 Generator ......................................................................... 80
4.1.11.5 Driver ............................................................................... 80
4.1.11.6 Monitor ............................................................................ 81
4.1.11.7 Scoreboard ....................................................................... 81
67
68 Advanced VLSI Design and Testability Issues
4.1 INTRODUCTION
Before getting started with learning about hardware description languages (HDLs),
one must begin with “Why?”. Answering that in simple terms, it is to deal with the
complexity in very large-scale integrated circuit (VLSI) design and its substantial
reduction with the introduction of register level and transfer level of abstraction.
Design methodologies using the traditional abstraction level, such as gate level
and so on, are no longer sufficient in the era of complex VLSI design. Hence, a
higher level of abstraction, i.e., register transfer logic (RTL) is introduced, which
is also known as computer language or HDL. By tradition, digital system was a
manual process of designing and capturing circuits with the use of schematic entry
tools. This process had many demerits and is quickly being replaced by innovative
methods.
In electronics, HDL is a particular computer language that helps to clarify the
structure and performance of electronic circuits, mainly digital circuits. The HDL
enables specific, formal explanation of electronic circuit that allows for automatic
investigation in circuits simulation. A net list (electronic components specification
with their connection) is allowed for synthesis of HDL description by HDL, which
can be positioned and routed to generate a set of masks for generating an integrated
circuit (IC). HDL looks like a C programming language. It is a textual explanation,
which consists of expressions and control structures statements [1]. Nevertheless,
HDLs are dissimilar from programming language as they openly consist of the per-
ception of time. HDL forms an essential division of electronic design automation
(EDA) systems particularly for composite circuits such as PLDs, ASICs, and so on.
HDL has evolved in excess of time, and the standard has altered. However, this chap-
ter focuses on essence of the core of the digital designs or design styles.
• Rise Delay: It is connected with a gate output transition from any value
to a 1.
• Fall Delay: It is related to a gate output transition from any value to a 0.
• Turnoff Delay: It is related to a gate output transition from any value to the
high impedance value (z).
If there is only one particular delay, then that value is implemented for all the transi-
tions. If there are two delays specified, then they belong to the value of rise and fall
delay, where turn-off delay is the smallest among two delays. If there are all three
delays given, they belong to rise, fall, and turnoff delay. If there are no delays given,
then zero is the default value.
Electronic circuits are active at all times, where timing related with each event
occurs. HDL is a language that explains the behaviors of electronic circuits, which
means it exactly models the time and concurrency of electronic circuits. However,
HDL must maintain the different delay times in the electronic circuits, such as inertial
and transport delay. Inertial delays can be implemented to model delays throughout
the capacitive networks. If the width of a pulse is smaller than the circuit’s switch-
ing time, it shall not come into view at the circuit’s output. The internal resistance
and capacitance of the logic gates create the gate delays, so called inertial. Transport
delays are implemented to model small length wires; hence, pulses are propagated
with respect to the width of the pulse. A reproduction of the pulse will come into
sight at the output after a particular delay time.
70 Advanced VLSI Design and Testability Issues
Event processing is the foundation of HDL simulation; hence, all HDL simulators
must be event-driven simulators. There are three important concepts to event-driven
simulation [3]:
i. Simulation time
ii. Delta time
iii. Event processing
Which pieces of hardware are sensitive to which events is determined throughout the
elaboration phase. This is simply known as a sensitivity list. For an event, one can
rapidly obtain a record of all hardware which is sensitive to it.
so on. Examples of HDLs that make to design digital circuits possible are Chisel,
Hydra, KARL, System C, System Verilog, Verilog, VHDL (very high-speed inte-
grated circuits HDL) [4], and so on. Examples of HDLs supporting for printed board
circuits (PCBs) are PHDL (PCB HDL), EDA Solver, SkiDL, and so on.
4.1.7.2 Architecture
Architecture is described by using structural, behavioral, dataflow, or even also
using mixed style. The entity name that is written for the architecture body should
be specified. The architecture statements should be inside the “begin” and “end”
keyword. Architecture declarative part contains variables, component declaration,
or constants.
• Operator
• The WHEN statements (WHEN/ELSE or WITH/SELECT/WHEN)
• The GENERATE statements
• The BLOCK statements
Hardware Description Languages 73
Example of VHDL code for logic gates and for 4-bit parity generator is presented
as follows:
entity 4bit_parity_gen is
port (b0, b1, b2, b3: in std_logic; odd, even: out
std_logic);
end 4bit_parity_gen;
architecture structural_arch of 4bit_parity_gen is
begin
process (b0, b1, b2, b3)
if (b0 ='0' and b1 ='0' and b2 ='0' and b3 =’0’)
then odd <= "0";
even <= "0";
else
odd <= (((b0 xor b1) xor b2) xor b3);
even <= not (((b0 xor b1) xor b2) xor b3);
end structural_arch;
Here, the first two lines “library ieee; and use ieee.std_logic_1164.all;” are essential
to include the standard packages and libraries, which contain all the declarations and
predefined functions. The ports are to be declared inside the entity block followed by
the architecture block containing the description of the hardware.
4.1.8 VERILOG
Verilog is the HDL used for describing the digital systems same as networking
switches or microprocessors or a memory unit. By using HDLs, the clumsy task of
explaining and representing digital hardware design at any level becomes very easy
and productive. Verilog supports designing hardware at various levels of abstraction.
The most widely used are behavioral level, RTL, and gate level [10].
a. Whitespaces include typeset for spaces or tabs or new lines. These are
mostly ignored apart from when they are used to split between the tokens.
These characters include blank space, tabs, new line, form feeds, and car-
riage returns.
b. Comments are represented in two forms: single-line comments (e.g., //Single
line comment) and multiline comments (e.g., /* Multiline comments */).
c. Identifiers are the names that are used to identify the object, such as a func-
tion, module, or register. Identifiers start with alphabetical or underscore
characters, e.g., V_erilog, v_hdl, _veriloghdl, etc. They are up to 1024 char-
acters long.
Hardware Description Languages 75
The employ of x and z is negligible in cases of synthesis. The most focused data
types are wire, register, input, output, inout, and integer.
4.1.8.3 Timescale
Verilog focuses on time units like a command to the software tool, by the use of a
“timescale compiler directive.” It has two parts: one is the time units and the other is
the time precision that is to be used. The precision component conveys the informa-
tion to the software tool how many decimal places of accuracy to be used by it. In the
following code, ‘timescale 1ns/10ps infers that the simulator tool is ordered to employ
time units of 1 ns and precision of 10 ps, which is 2 decimal places relative to 1 ns.
The ‘timescale directive can be declared in one or many Verilog source files. Such
directives having dissimilar values can be specified for unlike sections of a design.
When such happens, the simulator tool must be able to determine the differences
by the decision of the common denominator in all the specified time units and then
scaling every delays in every section of the design to the common denominator. The
‘timescale directive command is not bound to precise modules or to any exact file.
//////Verilog Snippet
‘timescale 1ns/1ps
module sample_snippet (q_0, q_1, out);
input q_0, q_1;
output out;
wire q_0, q_1;
reg out;
integer c;
assign out = 4;
initial begin
$display ("c = %d", c); //display values
end
endmodule
76 Advanced VLSI Design and Testability Issues
a. The blocking procedural assignments (=) are those statements that are exe-
cuted earlier than the execution of the statements that follow it in a sequen-
tial block. Such statements do not stop the execution of statements that are
following it in the parallel blocks.
b. The nonblocking procedural assignment (<=) statements permit the sched-
uling of assignments devoid of obstructing the procedural flow. The employ
of nonblocking procedural statements is done when it is needed to create
multiple register assignments in the similar time frame devoid of requiring
dependence upon each other.
FIGURE 4.1 Test bench architecture in Verilog. DUT, design under test.
78 Advanced VLSI Design and Testability Issues
input [6:0] a,
input [6:0] b,
output reg [6:0] c,
output reg [6:0] d
);
always @ (posedgeclk) begin
d <= b + a;
if ( b> a)
c <= b – a;
else if (b < a)
c <= a – b;
end
endmodule
module sample_snippet_test0 ( );
wire [6:0] c, d;
reg [6:0] a, b;
integer i;
initial begin
clk = 0;
for (i = 0; i< 99; i ++) begin
#5 clk = ~ clk;
end
end
sample_snippet_dut0 d0 (.a(a), .b(b), .c(c), .d(d)); //dut
instantiation
initial begin
a = 7’b111000; b = 7’b1010110;
# 20 a = 7’b0000111; b = 7’b1100101;
end
endmodule
TestBench-Top
test
environment
Transaction
Generator Driver
interface DUT
Transaction
Scoreboard Monitor
specific operation. The basic components of a system Verilog test bench architecture
are DUT, transaction, interface, generator, driver, monitor, scoreboard, test, environ-
ment, and top. The architecture is shown in Figure 4.2.
//dff_dut.sv
`timescale 1ns/1ps
module dff_dut (din, clk, rst, dout);
input clk,rst,din;
output logic dout;
always @(posedge clk) begin
if(rst) begin
dout <= 0;
end
else begin
dout <= din;
end
end
endmodule
4.1.11.2 Transaction
The transaction is a class that contains the elements or signals that are needed to be
driven for testing purpose.
//dff_transaction.sv
class transaction; // class declaration
rand bit din; // rand keyword for randomizing the bits
80 Advanced VLSI Design and Testability Issues
4.1.11.3 Interface
Interface is a component instantiated in the top module which contains the signals
that connect the verification environment with the DUT.
//dff_interface.sv
interface dff_intf(input bit clk);
logic din, dout;
bit rst;
modport drv(input clk, output din, rst);
modport mon(input clk, dout);
endinterface
4.1.11.4 Generator
Generator generates the stimulus by creating and randomizing the transaction class
and then sends it to the driver via mailbox.
//dff_generator.sv
`include "dff_transaction.sv"
class dff_gen;
dff_trans t1; // class handle declaration
mailbox gd_1;
function new (mailbox gd_2);
gd_1 = gd_2;
endfunction
task run();
t1 = new; // allocating memory to class handle
t1.randomize(); // calling in-built randomize function
gd_1.put(t1);
endtask
endclass
4.1.11.5 Driver
Driver class receives the stimulus or transactions from the generator class and then
drives the packet-level data sequences within the transaction into pin level to the DUT.
//dff_driver.sv
`include "dff_transaction.sv"
class dff_drv;
dff_trans t2;
mailbox gd_3;
mailbox ds_1;
virtual dff_intf vif_1;
function new (mailbox gd_4, mailbox ds_2, virtual
interface dff_intf vif_2);
Hardware Description Languages 81
gd_3 = gd_4;
ds_1 = ds_2;
vif_1 = vif_2;
endfunction
task run();
t2 = new;
gd_3.get(t2);
ds_1.put(t2);
vif_1.din = t2.din;
vif_1.rst = t2.rst;
endtask
endclass
4.1.11.6 Monitor
Monitor looks the pin-level signaling on the interface signals and also alters them
into packet level, which is to be sent to other components such as scoreboard,
checker, and so on.
//dff_monitor.sv
`include " dff_transaction.sv"
class dff_mon;
dff_trans t3;
mailbox ms_1;
virtual dff_intf vif_3;
function new(mailbox ms_2, virtual dff_intf vif_4);
ms_1 = ms_2;
vif_3 = vif_4;
endfunction
task run();
t3 = new;
ms_1.put(t3);
t3.dout = vif_3.dout;
endtask
endclass
4.1.11.7 Scoreboard
Scoreboard is a class that receives the data items from monitor and compares them
with the expected values.
//dff_scoreboard.sv
class dff_scb;
dff_trans t4, t5;
virtual dff_intf vif_5;
mailbox ds_3,ms_3;
function new (mailbox ds_4, mailbox ms_4, virtual dff_intf
vif_6);
ds_3 = ds_4;
ms_3 = ms_4;
82 Advanced VLSI Design and Testability Issues
vif_5 = vif_6;
endfunction
task run();
t4 = new;
t5 = new;
ms_3.get(t4);
ds_3.get(t5);
@(posedge vif_5.drv.clk) begin
if (t4.rst) begin
t4.dout <= 0;
end
else begin
t4.dout <= t4.din;
end
end
if( t4.dout = = t5.dout) begin
$display(" Scoreboard passed ");
end
else begin
$display(" Scoreboard isn't passed ");
end
endtask
endclass
4.1.11.8 Test
The test is liable for configuration of the test bench, initiation of the test bench com-
ponents construction process, and initiation of the stimulus that is to be driven.
// dff test
`include "dff_env.sv"
program dff_test (dff_intf i0);
initial begin
dff_env en = new(i0);
en.build();
en.run();
end
endprogram
4.1.11.9 Environment
The environment is the top-level container class that groups the higher-level
components.
//dff environment
`include "dff_trans.sv"
`include "dff_gen.sv"
`include "dff_drv.sv"
`include "dff_mon.sv"
Hardware Description Languages 83
`include "dff_scb.sv"
class dff_env;
dff_gen gen;
dff_drv drv;
dff_mon mon;
dff_scb scb;
virtual dff_intf vif;
mailbox ms, ds, gd;
function new (virtual dff_intf vif);
this.vif = vif;
endfunction
task build();
ms = new;
ds = new;
gd = new;
gen = new(gd);
drv = new(gd, ds, vif);
mon = new(ms, vif);
scb = new(ds, ms, vif);
endtask
task run();
gen.run();
drv.run();
mon.run();
scb.run();
#200 $finish;
endtask
endclass
4.1.11.10 Top
Top module is the topmost file that has the instances of the DUT and test bench and
consists of their connections declared in interface component.
//dff_top
`timescale 1ns/1ps
`include "dff_dut.sv"
`include "dff_intf.sv"
`include "dff_test.sv"
module dff_top;
bit clk;
dff_intf i1(clk);
dff_dut d0 (.din(i1.din), .clk(i1.clk), .rst(i1.rst),
.dout(i1.dout));
dff_test t0 (i1);
initial begin
i1.din=0;
#5 i1.din=1;
#25 i1.din=0;
#35 i1.din=1;
#50 i1.din=0;
84 Advanced VLSI Design and Testability Issues
#100 ;
end
initial begin
clk = 0;
for (int i = 0; i<256 ;i++)
begin
#5 clk = ~ clk;
end
end
endmodule
4.1.12 VERILOG-AMS
In analog and mixed-signal (A&MS) systems, Verilog-AMS plays as a behavioral lan-
guage. It is a derivative from IEEE standard 1364-2005 Verilog HDL. It also captures
the entire IEEE standard 1364-2005 Verilog HDL specification, analog equivalent for
explaining analog systems (Verilog-A), and extensions together for mentioning the full
Verilog-AMS. Verilog-AMS focuses on the designers of (A&MS) systems, IC creation,
and applying of modules for high-level behavioral and structural descriptions of sys-
tems. It is applicable to both electrical and nonelectrical systems explanation as well as
supportive for conservative and signal flow explanation.
It has the capacity to shorten design cycles and enlarge success of further mixed-
signal ICs [12].
The main features involve the following:
• Both analog and digital signals can be stated in the identical module
• Rising verification of analog presentation at the design’s top level
• Considerably diminishing the top-level simulation time
• Creating a suitable environment for architecture of any chip design
end
endmodule
The main project window in VIVADO is shown in Figure 4.3. In Sources panel, the
design file is clicked to get it in the coding area, and the programs are written for the
design as per preferred HDL. Figures 4.4 and 4.5 represent editor window of Xilinx
VIVADO and behavioral simulation result window, respectively.
WRITE YOUR
CODE HERE…
Step 3: Simulation
In the Flow Navigator panel at extreme left of the window, Run Simulation is
clicked to run the behavioral simulation of the design. There are different types of
simulation present, which are categorized majorly as behavioral simulation, func-
tional simulation, static timing analysis, gate-level simulation, switch-level simula-
tion, and transistor-level or circuit-level simulation.
Step 4: Synthesis, elaboration, and implementation
The next steps after simulation are synthesis, elaboration, and implementation
which provide a detailed information report regarding the area utilization, power
Hardware Description Languages 87
consumed, timing report, DRC reports, etc. Such information is highly crucial for
improvising a design. Placement and routing comes as next steps for dumping into
FPGA boards where appropriate pins are allotted for each I/O signals or interfacing
signals to interact with peer circuitry if available. The RTL-elaborated schematics at
each step are shown in Figures 4.6–4.8.
The reports generated from these steps can be used in debugging. Also, a designer
can write timing constraints and other constraint files scripted in “.tcl” to provide
more information to the HDL tool suite about the design specifications (Figures 4.7
and 4.8).
Step 2: Simulation
It is important to include “$dumpfile("dump.vcd"); $dumpvars;”in top module or
testbench.svin EDA Playground to run with EPWave enabled. Doing so pops up the
simulation wave window as shown in Figure 4.10.
Driver
Generator
Interface Transaction
FIGURE 4.11 Generator, driver, interface, and transaction classes for D flip-flop verification
environment.
90 Advanced VLSI Design and Testability Issues
Figure 4.12 shows scoreboard and environment classes; and Figure 4.13 presents
DUT and top module and test, top and monitor classes for D flip-flop verification
environment.
Step 2: Simulation
Next, the design file is compiled, and vsim.wlf is used to view the simulation waves
as shown in Figure 4.14.
Scoreboard
Environment
FIGURE 4.12 Scoreboard and environment classes for D flip-flop verification environment.
DUT
Test Monitor
Top
FIGURE 4.13 DUT and top module and test, top and monitor classes for D flip-flop verifica-
tion environment. DUT, design under test.
Hardware Description Languages 91
REFERENCES
1. IEEE Approved Draft. “Standard for VHDL Language Reference Manual, in IEEE
P1076/D13”, pp. 1–796, 2019.
2. https://en.wikipedia.org/wiki/Hardware_description_language.
3. R. W. Hartenstein, “Hardware Description Languages”, Elsevier Science Publisher,
Oxford, 1987.
4. J. P. Mermet, “Fundamentals and Standards in Hardware Description Languages”,
Series E, Springer Science, Barga, 1993.
5. C. D. Kloos, “Hardware Description Languages and their Applications: Specification,
Modelling, Verification and Synthesis of Microelectronic Systems”, Chapman & Hall,
Toledo, OH, 1997.
6. C. Stumm, C. Brugger and N. When, “White Paper - Investigate the Hardware
Description language Chisel”, Kluedo, Kaiserslautern University of Technology,
Germany, pp. 1–2, 2013.
7. S. Palnitkar, “Verilog HDL: A Guide to Digital Design and Synthesis”, 2nd Edition,
Prentice Hall PTR, Upper Saddle River, NJ, 2003.
8. D. E. Thomas and P. R. Moorby, “The Verilog® Hardware Description Language”, 5th
Edition, Kluwer Academic Publishers, New York, 2002.
9. IEEE Standard for Verilog Hardware Description Language, in IEEE Std 1364–2005
(Revision of IEEE Std 1364–2001), pp. 1–590, 2006. doi: 10.1109/IEEESTD.2006.99495.
10. S. Sutherland, S. Davidmann and P. Flake, “SystemVerilog for Design: A Guide to
Using System Verilog for Hardware Design and Modeling”, 2nd Ed., Springer, New
York, 2006.
11. IEEE Standard for SystemVerilog-Unified Hardware Design, Specification, and
Verification Language, in IEEE Std 1800–2017 (Revision of IEEE Std 1800–2012), pp.
1–1315, 2018. doi: 10.1109/IEEESTD.2018.8299595.
12. https://www.yumpu.com/en/document/view/11519359/verilog-ams-simulation-using
mentor-and-cadence-tools-prateek.
92 Advanced VLSI Design and Testability Issues
CONTENTS
5.1 Introduction .................................................................................................... 94
5.1.1 Motivation ........................................................................................... 94
5.1.2 Structure of Hardware Description Language.................................... 94
5.1.3 History ................................................................................................ 94
5.1.4 Hardware Description Language and Programming Languages ....... 95
5.1.5 Hardware Description Language Design Flow Expense .................... 96
5.2 Design Simulation, Debugging, and Verification with Hardware
Description Languages ................................................................................... 96
5.3 Introduction to VHDL .................................................................................... 96
5.3.1 Structure of Program .......................................................................... 97
5.3.2 VHDL Variables ................................................................................. 98
5.3.3 Functions, Libraries, and Packages .................................................... 98
5.3.4 Design Elements ................................................................................. 99
5.3.4.1 Structural .............................................................................99
5.3.4.2 Dataflow ............................................................................. 100
5.3.4.3 Behavioral .......................................................................... 101
5.3.5 Writing a Simple Code with VHDL ................................................. 101
5.3.6 Simulation and Synthesis .................................................................. 103
5.4 Introduction to Verilog ................................................................................. 104
5.4.1 Structure of Program ........................................................................ 104
5.4.2 Module Declarations......................................................................... 105
5.4.3 Verilog Variables, Operators, and Directives ................................... 105
5.4.3.1 Variable Data Types ........................................................... 105
5.4.3.2 Operators............................................................................ 105
5.4.4 Design Elements ............................................................................... 106
5.4.4.1 Structural ........................................................................... 106
5.4.4.2 Dataflow ............................................................................. 107
5.4.4.3 Behavioral .......................................................................... 107
5.4.4.4 Writing a Simple Code with Verilog.................................. 108
5.4.5 Simulation and Synthesis .................................................................. 109
References .............................................................................................................. 109
93
94 Advanced VLSI Design and Testability Issues
5.1 INTRODUCTION
Hardware description language (HDL) is a specialized programming language used
to label the electronic and digital logic modules. The building process and also plan
of the modules are made through HDL program. HDL enables the simulation output
and building blocks of digital logic circuit by using textual representation of input,
output, operation, and appearance. The compilation of HDL will convert textual
format into gate level, and the same has been on the field-programmable gate array
(FPGA) processor by using synthesis and implementation process. The language
assists to define any digital circuit in the system of physical. There are three HDLs
commonly used to define the behavior of the digital modules—Verilog, VHDL, and
SystemC. The VHDL and Verilog are the oldest languages as compare with SystemC
that is most and commonly used.
5.1.1 MOTIVATION
Because of the blast multilayered design of advanced electronic circuits since the
1970s, circuit creators required computerized rationale representations on the way
to remain made at a raised-up level without involvement of a particular electronic
component. HDLs are designed to realize the register transfer-level design.
5.1.3 HISTORY
The conception of an HDL as a medium for style capture was first introduced in
the 1950s; however, wide adoption by the look community did not begin till 1985.
Traditionally, the event of software package programming languages stirred the evo-
lution of HDLs. One example, among many, is that the artificial language APL that
was used as a kind of style entry for a logic automation system was developed at IBM
Hardware Description Languages 95
within the early 1960s. The notational conventions of APL were later utilized by
researchers at the University of Arizona to style AHPL (A Hardware Programming
Language).
Since its introduction within the early 1970s, AHPL has been hardly utilized in
nonacademic applications, however, served as a good teaching tool in schoolroom
environments. Within the three decades ranging from 1960, several HDLs such as
DDL, ISPS, and Zeus were introduced. However, the utilization of those languages
seldom exceeded analysis and tutorial applications. Part in reaction to the prolifera-
tion of HDLs and part because of its own desires, in 1980, the U.S. Department of
Defense initiated the event of VHDL as a part of its VHSIC program (very high-
speed integrated circuits).
The name VHDL comes from VHSIC hardware description language. The gen-
eral objective of this effort was to style one language that may enable the look, docu-
mentation, and analysis of hardware at varied levels of abstraction. Moreover the
intent was to create VHDL the typical HDL design projects and use of language as a
way of communication between departments.
In December 1987, VHDL was accepted by the IEEE (IEE88, IEE94) as a cus-
tomary description language used for modeling digital modules and systems. The
emergence of such business normal marked the start of widespread adoption of
HDLs by the planning community. In parallel with VHDL efforts, the Verilog alpha-
lipoprotein originated in 1983 at the entry-style automation and was introduced into
the market in 1985. Verilog was designed to deal with the necessities of circuit
designers and at a similar time to be intuitive and as easy as doable. As a result, the
language has fewer constructs than VHDL, its linguistics is not as advanced, and
development of Verilog-based tools is soon achievable. Verilog became associate
IEEE normal [IEE96] in 1995.
Today, Verilog and VHDL are used extensively and, maybe solely, by circuit
designers all over. Several simulation and synthesis tools are developed and mar-
keted that alter the analysis and realization of those lipoprotein models.
FIGURE 5.1 Design flow of HDL. HDL, hardware description language; RTL, register
transfer-level.
showed up. Shifted systems grant a trade-off between the predetermined precision
and, in this way, the estimation of a definitive algorithmic standard [1].
Entity is the explanation of the interface among a design and its external environ-
ment. It can similarly stipulate the assertions and reports that are part of the design
entity. A specified entity assertion might be collective via numerous design entities,
respectively of which consumes a dissimilar architecture.
Syntax of Entity
Entity test is
Generic declaration;
Port declaration;
End test;
Syntax of Architecture
Architecture architect_test of testis
Declerations
begin
Concurrent statements;
end architect_test
separates a capacity from a method is that it cannot contain Wait explanations. This
implies that capacities consistently devour zero reproduction time.
Syntax
function fun-name return type is
declarations;
begin
sequential statements;
end fun-name;
package pack_test is
constant declarations;
type declarations;
subprogram declarations;
end test_test;
entity andgate1 is
port(x, y: in bit;
rout: out bit);
end andgate1;
entity xorgate1 is
port(x, y: in bit;
rout: out bit);
end xorgate1;
entity h_adder1 is
port(e, f: in bit;
sum1, carry1: out bit);
end h_adder1;
component andgate1
port(p, q: in bit;
rout: out bit);
end component;
component xorgate1
port(p, q: in bit;
rout: out bit);
end component;
begin
u1 : andgate1 port map(p,q,carry1);
u2 : xorgate1 port map(p,rout,sum1);
end struct1;
5.3.4.2 Dataflow
The dataflow style and the behavior of the modules are expressed in the way data-
flow will take place in module by using concurrent signal. The parallel statements
are generated through WHEN and GENERATE. Apart from the assignment to con-
struct, the code can also use operators.
5.3.4.3 Behavioral
In this displaying style, the behavior of a substance as conventional of explanations
performed successively in the prearranged demand. Unbiased deliveries put confi-
dential a process function.
Procedures, FUNCTIONS, and PROCEDURES are the main areas of code that
are executed successively. In any case, overall, any of these squares is as yet simulta-
neous with some other articulations set outside it. One substantial portion of behav-
ior code is that it is not unnatural to consecutive rationale. Hypothesis consecutive
circuits are impartial as combinational circuits.
entity halfadd1 is
port( d: in std_logic_vector(0 to 1);
sum, carry: out std_logic);
end halfadd1;
case d is
when "00" => sum <= '0';
when "11" => sum <= '0';
when others => sum <= '1';
end case;
case d is
when "11" => carry <= '1';
when others => carry <= '0';
end case;
end process;
end behav;
library ieee;//Library
use ieee.std_logic_1164.all;//Library
entity Mux4_1 is
port( dir3: in std_logic_vector(2 downto 0);
dir2: in std_logic_vector(2 downto 0);
102 Advanced VLSI Design and Testability Issues
case muxsel is
when "00" => muxout <= I0;
when "01" => muxout <= I1;
when "10" => muxout <= I2;
when "11" => muxout <= I3;
when others => muxout <= "ZZZ";
end case;
end process;
end behav2;
entity DECODER2_4 is
port( testinp: in std_logic_vector(1 downto 0);
testout: out std_logic_vector(3 downto 0)
);
end DECODER2_4;
process (testinp)
begin
case testinp is
when "00" => testout <= "0001";
when "01" => testout <= "0010";
when "10" => testout <= "0100";
when "11" => testout <= "1000";
when others => testout <= "XXXX";
end case;
end process;
end behav3;
Hardware Description Languages 103
entity Comp is
generic(xtr: natural :=2);
port( p: in std_logic_vector(n-1 downto 0);
q: in std_logic_vector(n-1 downto 0);
lt: out std_logic;
eq: out std_logic;
gt: out std_logic
);
end Comp;
The sum of DUV and models that drive its information sources and procedure
and its yields is known as a test bench. It is prescribed to manufacture a test bench
that in any event comprises of the accompanying models:
The test system views a circuit as an assortment of sign and procedures. Sign can
change in esteem after some time under the effect of procedures. A sign change is
known as an exchange. In spite of the fact that equipment is parallel naturally, it is
commonly reenacted on a successive machine. In one way or different, forms that
are dynamic at the same time, just as sign that can change in esteem all the while,
must be managed so that the contrasts among reproduction and the genuine world are
as little as could reasonably be expected.
What the test system must do at a given minute is demonstrated through a run-
down of activities that is arranged by time. This is the occasion list. “Occasion” is the
assignment given to a sign change or a procedure actuation at a particular time. For
instance, if the procedure that is dynamic at minute t = t0 experiences the announce-
ment a <= “1” after 10 ns, at that point exchange, a <= “1” is put on the occasion list
at minute t = t0 + 10 ns. An exchange never produces results promptly, not regardless
of whether the code does not indicate any deferral (for instance, through a sign task
without the catchphrase after). All things are considered; the exchange is put on the
occasion list at minute t = t0 + Δ. Δ is equivalent to zero (or better: imperceptibly
little), yet it permits forms that occur all the while to be requested in time. This is
conceivable on the grounds that the accompanying applies: 0 < Δ < 2Δ …
The thought of an imperceptibly little deferral in reproduction is known as a delta
delay. The recreation begins with the development of the occasion list. All proce-
dures in the VHDL portrayal are put in the correct situation in the rundown. (Most
procedures start at time zero, applying the standard that a negligible time of Δ must
happen between two initiations.) During recreation, the occasion list is prepared in
the request for expanding time. New occasions that outcome from this are included
in the occasion list at the correct position. The reenactment is finished when the
occasion list gets vacant, when the reproduction is compelled to be ended by the
activity of the client or by a blunder [3].
Verilog describes the structure to be produced through design flow. Plans showed in
HDL are innovation free, simple to structure and troubleshoot, and normally more
meaningful than schematics, especially for huge circuits.
The different levels of abstraction in Verilog are as follows:
a. Based on algorithm.
b. Based on register transfer.
c. Based on gate.
d. Based on switch.
5.4.3.2 Operators
See Table 5.1.
106 Advanced VLSI Design and Testability Issues
TABLE 5.1
List of Operators
Type of Operator Operator Description
Arithmetic operators +, –, / Arithmetic
% Modulus
Logical operators ! NOT logic
&& AND logic
|| OR logic
Bitwise operators ~ Bitwise NOT
& Bitwise AND
| Bitwise OR
^ Bitwise XOR
^~ ~^ Bitwise XNOR
Reduction operators & Reduction AND
| Reduction OR
~& Reduction NAND
~| Reduction NOR
~^ ^~ Reduction XNOR
Equality operators == Logical equality
!= Logical inequality
Relation operators > Relational operators
>=
<
<+
Concatenation operator {} Concatenation
Shift operators << Left shift
>> Right shift
Conditional operator {}
5.4.4.2 Dataflow
Dataflow demonstration utilizes constant assignments and the watchword allocation.
A persistent task is an explanation that doles out an incentive to remaining. The data
type net list is working in V-HDL to express a bodily association among circuits and
components. The worth relegated to net lists are determined by an articulation that
utilizations operands.
For instance, expecting that factors remained declared, a 2:1 multiplexer through
information test inputs A and B, test select input S and the output test Y. the out-
put is represented through continues changes in output by executing this syntax
testY=(testA and testS) | (testB and testS).
Dataflow representation of 2_4_decoder is appeared in HDLs underneath. The
logic circuit is characterized with possibility of constant task explanations utilizing
Boolean articulations, one for each yield.
5.4.4.3 Behavioral
Social modeling behavioral display speaks to advanced circuits at a practical and algo-
rithm level. Behavioral model is utilized generally to represent consecutive modules. The
conduct displaying idea is to be introduced for combinational circuits. Behavior descrip-
tion utilizes the catchphrase consistently pursued by a rundown of practical task explana-
tions. The programmer need not know about gate level description of the module.
108 Advanced VLSI Design and Testability Issues
a. Behavioral
b. High-level
c. Register transfer-level
REFERENCES
1. Grewal, A. (2018). Review Report on VHDL (VHSIC Hardware description language).
Language, 27, 28.
2. LaMeres, B. J. (2019). Introduction to Logic Circuits & Logic Design with VHDL.
Springer, New York.
3. Gazi, O. (2019). A Tutorial Introduction to VHDL Programming. Springer, New York.
4. Cavanagh, J. (2017). Verilog HDL: Digital Design and Modeling. CRC Press, Boca
Raton, FL.
5. LaMeres, B. J. (2019). Introduction to Logic Circuits & Logic Design with VHDL.
Springer, New York.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
6 Emerging Trends
in Nanoscale
Semiconductor Devices
B. Vandana
KG Reddy College of Engineering and Technology
CONTENTS
6.1 Introduction................................................................................................... 112
6.2 Background.................................................................................................... 112
6.3 Nanotechnology Emerging Improvements.................................................... 115
6.3.1 Technology Dependency (Bulk to Silicon-on-Insulator Technology).... 115
6.3.2 Architectural Representation (Single- to MultiGate
Field-Effect Transistor)...................................................................... 116
6.3.2.1 Double-Gate Silicon-on-Insulator Metal Oxide
Semiconductor Field-Effect Transistors............................. 117
6.3.2.2 Triple-Gate Silicon-on-Insulator Metal Oxide
Semiconductor Field-Effect Transistors............................. 118
6.3.2.3 Surrounding-Gate Silicon-on-Insulator Metal Oxide
Semiconductor Field-Effect Transistors............................. 118
6.3.2.4 Other Multigate Metal Oxide Semiconductor
Field-Effect Transistors....................................................... 118
6.3.3 Material Technology.......................................................................... 118
6.3.3.1 Strained Silicon................................................................... 119
6.3.3.2 High-k Gate Dielectric and Metal Gate Electrodes............ 119
6.3.4 Existing Metal Oxide Semiconductor Field-Effect Transistor
Topologies at Nanoscale Regime....................................................... 119
6.3.4.1 Junctionless Metal Oxide Semiconductor Field-Effect
Transistor............................................................................ 119
111
112 Advanced VLSI Design and Testability Issues
6.1 INTRODUCTION
Scaling the complementary metal oxide semiconductor (CMOS) device dimensions
and the process technology has become more difficult for the semiconductor industry,
if this approach reaches to 3 nm physical lengths by 2020 (Ryckaert 2019). Beyond
the period of traditional CMOS scaling, various alternative electronic devices are
required to integrate on a silicon platform with no change in its functionality (Patro
and Vandana 2016; Deleonibus 2019). The first initiation of the technology was started
in the micrometer range, as due to Moore’s law prediction, the devices were force-
fully pushed to nanoscale. Before the era of nanoscale technology, microelectronics
innovates various technological benefits that are still compatible with the nanoscale
regime, and this plays a crucial role in microsystem designs.
The microelectronics industries have faced two conventional difficulties to
enhance the integrated circuit (IC) technology beyond the CMOS scaling. The
CMOS needs to be extending further with the consideration of scaling factors. The
challenges associated are high speed, dense, nonvolatile memory, and information
processing substantially. The references of alternate electronic devices include one-
dimensional structure (carbon nanotubes and compound semiconductor devices),
resonant tunneling diodes, single-electron transistors, and molecular and spin
devices. However, these concepts represent charge-based logic; restrictions of scal-
ing by thermodynamic will be minimum switching energy per binary operation.
Beyond this limit, new challenges need to be implemented that will compensate
electronic charge and extend the scaling of information processing technology. This
chapter extensively provides an outlook toward emerging new devices and associ-
ated technologies. These technologies may help microelectronics to serve as a link
between conventional CMOS and beyond CMOS scaling. This chapter discusses
new technology trends that exist for present-generation metal oxide semiconductor
field-effect transistor (MOSFET) at the nanoscale regime.
6.2 BACKGROUND
In the present era, the most leading industries in the commercial world are “semicon-
ductor industries.” This has interpreted a vast revolution in the growth of technology
for various applications. Due to its inherent features, this made human life easier
with the invention of portable devices for communication systems, medical sensors
instrument applications, military applications for coding/decoding exchange of infor-
mation, and many more that are able to access using Internet of things. Besides these
applications, a complex circuitry designed with nanoscale devices is incorporated on
Nanoscale Semiconductor Devices 113
a printed circuit board (PCB) to make the system size small. The nanoscale devices
play a vital role in emerging technology and trends in device innovations; scaling
semiconductor devices to nanoscale was started with Gordon Moore. Later came up
with Moore’s law known as “cramping number of transistors for every 18 months.”
Nanoelectronics is demarcated as nanotechnology and allows integration with
purely electronic devices, electronic chips, and circuits. As the years passed, Moore’s
law further extended to “More than Moore” (MtM) domain of development (Arden
et al. 2010). The diversification factor of Moore’s law inculcates scaling feature, and
the road to gigascale systems can be described as the “More Moore” domain of
development (“Nanoelectronics Applications Moving Simply at Nanoscale,” n.d.).
Scaling the feature size and packing density will still continue in following Moore’s
lore and enjoy the present innovations in social life. In accordance with this, a new
solution is important to face the challenges for nanodevices.
Basically, the effort made by the R&D for advance chip technology has been a
driving force within three to four decades, and this results in better performance and
high packing density. The manufacturing of ICs is processed through the top-down
approach, and the fabrication tools are improved with each technology generation.
The major intention is that if the size of the components is scaled to nanoscale, physi-
cal effects would stop the tools from working properly. The fabrication challenges of
the top-down silicon (Si) technology are considered with three fundamental design
limits: (i) transistor scalability, (ii) performance, and (iii) power dissipation (Kuo
and Lou, n.d.).
Out of three limitations, the most important design is transistor scalability. As
per the VLSI domain, CMOS is the major technology for integrated systems, the
technology especially suitable for ICs which operates with low power at low supply
voltage. The literature study makes us understand why scaling is necessary for the
semiconductor industry and the downscaling channel length (L CH) as the years pass;
it is reported that the L CH is about 180 nm in 2000 and the L CH is of 5 nm in 2020.
Due to the vast scaling of L CH, the considerations of the electric field (E-field) are
limited, and the power supply is also scaled down from 1.8 to 0.7 V. Therefore, the
low supply voltage is necessary for next-generation low-power CMOS circuits (Kuo
and Lou, n.d.).
The most straightforward path to meet the low-power systems is the MOS device
that lowers the power supply voltage. Due to continuous scaling, device geometry at
nanoscale leads to several leakage mechanisms. The challenges associated to reduce
short-channel effects (SCEs) are shown in Figure 6.1b. Nanoscale transistors usually
require low supply voltage to reduce the internal E-field and consume low power.
This intends to reduce threshold voltage (VTH) that substantially increases leakage
currents (IOFF). As the drain voltage increases, the depletion region of channel drain
widens and increases drain current (Adan et al. 1998). This causes to increase OFF
current, due to which the channel surface induces barrier lowering effect known
as drain-induced barrier lowering, or due to channel punch-through effects. All
the adverse effects that cause to reduce VTH in scaled devices are known as SCEs.
The SCEs are immune to channel length scaling and reduced oxide thickness. This
further increases E-field and the scaled oxide thickness, and high E-field destroys
the infinity input impedance, which affects the MOS transistor performance.
114 Advanced VLSI Design and Testability Issues
FIGURE 6.1 (a) Various leakage currents. (b) Challenges to reduce SCEs. SCEs, short-
channel effects.
The projections such as device physical dimensions, supply voltage, and power
consumption are taken according to the International Technology Roadmap for
Semiconductors (ITRS) (Moore and others 1998; “No Title,” n.d.). The various SCEs
show an overview of the intrinsic leakage currents of a MOSFET device as shown in
Figure 6.1a. The SCEs are source/drain current (ISD), source/drain extension current
(ISDE), or (I1). The subthreshold current (ISubVTH) or (I2) arises between source and
drain at zero gate voltage. The gate leakage (IGATE) or (I3, I4) flows through the oxide
Nanoscale Semiconductor Devices 115
• New materials with higher mobility for enhancing transport (SiGe, Ge,
GaAs, InGaAs, etc.)
• New gate stack to reduce tunneling and gate leakage (high-k dielectrics)
• New contact materials (metal and semiconductor)
• New S/D structures (e.g., raised S/D)
• Improved device architecture (SOI, UTB, DG, 3D FinFET, trigate to
improve SCEs)
FIGURE 6.2 Cross-section view of (a) bulk and (b) SOI MOSFET. SOI, silicon-on-insulator;
MOSFET, metal oxide semiconductor field-effect transistor.
radiation hardness, no latchup, and high device density. The oxide layer isolates the
body from the substrate and induces no reverse-biased junctions. The junction capaci-
tances are small where it reduces leakage currents and results in high-speed perfor-
mance. In biased condition, if the body terminal is not added, the device is referred as
floating-body SOI MOSFET. The floating body generates serious issues in the device,
which are suppressed through body control techniques.
In bulk MOSFET, the dependency of body effect occurs due to the change in the VTH
on the back gate bias. The back gate bias effect occurs due to the depletion width at the
oxide interface and the substrate. When the back gate bias becomes more negative, the
depletion region widens under the gate, which thereby requires high gate voltage (VGS)
to invert the channel, and hence this effect increases the VTH. SOI technology separates
the active region from the substrate through the isolation technique. Therefore, the VTH
is less dependent on the back gate bias as compared with bulk technology.
The formation of SOI MOSFET technology is achieved from partially depleted
(PD) SOI MOSFET with a single gate and fully depleted (FD) SOI MOSFET with
multigate. PDSOI MOSFETs are successful at early silicon-on-sapphire devices,
which are useful for the niche applications for radiation-hardened or high-temperature
electronics. This device has become the mainstream for the semiconductor manu-
facturer and has been started to use to fabricate high-performance microprocessors.
To operate a low-voltage PDSOI device, the contacts are created across the gate and
floating body; such a contact improves subthreshold slope (SS) and drives current of
the devices. The FDSOI MOSFET has a better electrostatic coupling between gate
and channel. This device is successful in various applications such as low-voltage,
low-power to RF ICs (Vandana, Das, et al. 2017).
FIGURE 6.4 I–V characteristics in log scale, and the inset figures represent n-type MOSFET
without and with junctions: (a) JLT and (b) IMT. MOSFET, metal oxide semiconductor field-
effect transistor; JLT, junctionless transistor.
(Majkusiak et al.1998) and the challenges facing (FD) SOI transistor are due to the
scaling dimensions of the active Si channel region.
In the case of single-gate FDSOI devices, in order to maintain full substrate deple-
tion under gate control, the TSi needs to be about a one-third to a half of the electrical
L G (Yan et al. 1992). The important concerns of multigate devices depend on dynamic
performance; the ID is multiplied by a factor of 2. For the same reason, the CGG is
twice to that of a single gate, and thus, the same intrinsic propagation delay (CGG.VDD/
ION) is obtained. The intrinsic power delay product (PDP) (CGG.VDD/2) of the device is
doubled if the devices are designed with the same width. Double-gate devices have to
operate at a VDD/2 time lower than that of a single gate, in order to have the same PDP
at a given transistor width. To reach the same CGG.VDD/ION, the drive current should be,
at VDD/2, two times the ID obtained with a single-gate device at VDD.
At the nanoscale regime, the emerging semiconductor technology possesses vari-
ous advance interpretations for CMOS applications. One among them is the junc-
tionless transistor (JLT) structure, which imposes a good candidate for scaling down
the SCEs and improvises the drive current at scaled L G. Figure 6.4 represents the
conventional MOSFET with inversion mode of operation and the MOS transistor
with no inversion mode of operation and no formation of junctions across the S/C/D
and hence is known as JLT from Figure 6.4a.
From inset figure, the S/D/C is highly doped for n-type MOSFET usually known
as a gated resistor, and the I–V characteristics are nearly the same as IMT, but the
principle of operation is different to that of IMT. At the subthreshold condition, the
channel seems to be fully depleted. As the applied gate bias, the work function dif-
ference between gate and channel makes the bands flat at VFB, and the channel is par-
tially depleted where the conduction takes place at channel (Colinge et al. 2010). The
formation of NPN at S/C/D from the inset Figure 6.4 b is an IMT structure. The I–V
characteristic in log scale represents full depletion at subthreshold condition, but due to
the VFB, small amount of current flows giving rise to leakage currents. With the applied
gate bias, the electrons are accumulated at the top surface of the gate, and the channel
slowly turns to be inverted from p-type to n-type forming a channel region from source
to drain (Vandana et al. 2018; Vandana, Mohapatra, et al. 2017; Vandana et al. 2019).
Nanoscale Semiconductor Devices 121
6.3.4.3 Spintronics
Spintronics is coined from the words “spin” and “electronics.” This is a branch of
physics with the study of the spin of the electrons and its magnetic moment along
with its fundamental charge. Electrons spin is represented by two types, i.e., up and
down. This is a good alternative against the present transistors, which requires a
controlled flow of electrons for turning on and off the transistors. Unlike transis-
tors, changing the spin requires less energy and hence less power consumption by
the devices. The spin of the electrons is measured as it generates tiny electric fields.
FIGURE 6.5 Schematic representation of (a) Spin-FET and (b) Spin-MOSFET; the arrows
indicate the direction of magnetization of the S and D contacts. MOSFET, metal oxide semi-
conductor field-effect transistor.
122 Advanced VLSI Design and Testability Issues
Also, these spintronics-based devices can be used with the help of commonly avail-
able metals such as iron, aluminum, and so on. The independent behavior of the spin
of the electrons from the energy helps the spintronics device to be nonvolatile in
nature, and hence, data storage can be done even after the loss of power.
There are many devices developed to perform similar functions as that of transis-
tors. The spin valve is one of them. It was first invented by Dieny et al. in 1991. This
revolutionizes the industries that deal with storage devices and magnetic sensors. In
a similar way, giant magnetoresistance (GMR)–based and tunneling magnetoresis-
tance (TMR)–based devices are working. IBM, Honeywell, Freescale, SanDisk, and
so on are some of the companies that are consistently marketing and improvising
the technology associated with these types of devices. Magnetic tunnel junction and
ferroelectric tunnel junction are also some of the most important spintronic devices
that are used for logic and memory-based applications.
Figure 6.5 shows the proposed Spin-FET and Spin-MOSFET (Joshi 2016). The
best example of today’s equipment in the market is organic light-emitting diode that
shows good visual effects and consumes very little power as compared with other
televisions. So, it can be inferred that spintronics-based devices have a very good
future.
Here, VG is the gate voltage. When VG is zero, the injected spins that are transmit-
ted through the 2-DEG layer reach at the collector with the same polarization. When
VG >> 0, the precession of the electrons is controlled with electric field; here S and D
regions have opposite direction of magnetization and do not require spin precession
of spin-polarized electrons in the channel (Datta and Das 1990; Zhang et al. 2014;
Joshi 2016).
6.3.4.4 Memristor
The word “memristor” is coined from two words, i.e., memory and resistor. This is
categorized as the fourth fundamental electrical component after resistor, capacitor,
and inductor. The theoretical concept was first proposed by L. Chua in 1971.
In Figure 6.6, a two-terminal memristor-based device is shown. Depending upon
the voltage applied, the flow of electrons is controlled by the thickness of the mate-
rial and hence can work as a transistor in switching-based applications. Owing to its
nonvolatile nature and no leakage power, it is considered to be a better replacement
against a transistor. Applications such as memory storage and designing logic gates
are some of the areas where memristors can work in a robust manner because of its
high-speed switching.
Memristor can be classified into two sections on the basis of the way it works:
firstly, ionic thin film and molecular memristor, which works on the basis of its
material composition and properties of the lattices, and secondly, magnetic and spin-
based memristor, which depends on the orientation or spin of the electrons of the
material used for functionalities in various memory storage-based applications. This
device is in a developing stage, and many industries are focusing on developing cost-
effective and long-lasting memristor-based systems.
FIGURE 6.9 Schematic cross section of an HEMT. HEMT, high-electron mobility transistor.
6.4 CONCLUSIONS
This chapter focuses on the various technologies available, which will be of major
attention in the coming years. This is due to the fact that the transistor size minia-
turization has come to its saturation point. Also, these scaled channel transistors
are facing lots of problems. The devices that are discussed have lots of advantages
such as they are operating at very low voltages and hence consume very low power.
Nanoscale Semiconductor Devices 125
Also, these can work in systems where high-speed computation is required. Some
of the devices do not depend upon the flow of electrons for switching on and off the
device but the spin of the electrons. These behaviors make these devices nonvolatile
in nature and hence can store the data for a longer period of time.
REFERENCES
Adan, A O, T Naka, A Kagisawa, and H Shimizu. 1998. “SOI as a Mainstream IC Technology.”
In 1998 IEEE International SOI Conference Proceedings, 9–12.
Andrieu, F, C Dupré, F Rochette, O Faynot, L Tosti, C Buj, E Rouchouze, et al. 2006. “25nm
Short and Narrow Strained FDSOI with TiN/HfO2 Gate Stack.” In Proceedings of
Symposium on VLSI Technology, 2006. Digest of Technical Papers, 134–135.
Appenzeller, J, Y-M Lin, J Knoch, and Ph Avouris. 2004. “Band-to-Band Tunneling in Carbon
Nanotube Field-Effect Transistors.” Physical Review Letters 93 (19). APS: 196805.
Arden, W, M Brillouët, P Cogez, M Graef, B Huizing, and R Mahnkopf. 2010. “More-than-
Moore White Paper.” Version 2: 14.
Chang, L L, and L Esaki. 1977. “Tunnel Triode-a Tunneling Base Transistor.” Applied Physics
Letters 31 (10). AIP: 687–89.
Colinge, J-P. 2007. “Multi-Gate SOI MOSFETs.” Microelectronic Engineering 84 (9–10).
Elsevier: 2071–2076.
Colinge, J-P, C-W Lee, A Afzalian, N D Akhavan, R Yan, I Ferain, P Razavi, et al. 2010.
“Nanowire Transistors Without Junctions.” Nature Nanotechnology 5 (3). Nature
Publishing Group: 225–229. doi:10.1038/nnano.2010.15.
Cooper, D R, B D’Anjou, N Ghattamaneni, B Harack, M Hilke, A Horth, N Majlis, et al.
2012. “Experimental Review of Graphene.” ISRN Condensed Matter Physics 2012.
Hindawi Publishing Corporation.
Datta, S, and B Das. 1990. “Electronic Analog of the Electro-Optic Modulator.” Applied
Physics Letters 56 (7). AIP: 665–667.
Deleonibus, S. 2019. Electronic Devices Architectures for the NANO-CMOS Era. CRC Press.
Fiegna, C, H Iwai, T Wada, T Saito, E Sangiorgi, and B Ricco. 1993. “A New Scaling
Methodology for the 0.1-0.025 m MOSFET.” In Proceedings of Symposium on VLSI
Technology, 33.
Hisamoto, D, T Kaga, Y Kawamoto, and E Takeda. 1989. “A Fully Depleted Lean-Channel
Transistor (DELTA)-a Novel Vertical Ultra Thin SOI MOSFET.” In Proceedings of
Technical Digest - International Electron Devices Meeting, 1989. IEDM’89, 833–836.
http://www.Seas.Ucla.Edu/Cmoslab/People.Html. n.d.
Huang, L-J, J O Chu, S Goma, C P D’Emic, S J Koester, D F Canaperi, P M Mooney, et al.
2001. “Carrier Mobility Enhancement in Strained Si-on-Insulator Fabricated by Wafer
Bonding.” In Proceedings of Symposium on VLSI Technology, 2001. Digest of Technical
Papers, 57–58.
Joshi, V K. 2016. “Spintronics: A Contemporary Review of Emerging Electronics Devices.”
Engineering Science and Technology, an International Journal 19 (3). Elsevier:
1503–1513.
Krivokapic, Z, C Tabery, W Maszara, Q Xiang, and M R Lin. 2003. “High Performance
45nm CMOS Technology with 20nm Multi-Gate Devices.” Solid State Devices and
Materials 1998 Business Center for Academic Societies: 760–761.
Kuo, J B, and J-H Lou. n.d. “Low-Voltage CMOS VLSI Circuits.”
Lee, S-Y, S-M Kim, E-J Yoon, C W Oh, I Chung, D Park, and K Kim. 2004. “Three-
Dimensional MBCFET as an Ultimate Transistor.” IEEE Electron Device Letters 25
(4). IEEE: 217–19.
126 Advanced VLSI Design and Testability Issues
Majkusiak, B, T Janik, and J Walczak. 1998. “Semiconductor Thickness Effects in the Double-
Gate SOI MOSFET.” IEEE Transactions on Electron Devices 45 (5). IEEE: 1127–1134.
Mathew, L, M Sadd, S Kalpat, M Zavala, T Stephens, R Mora, S Bagchi, C Parker, J Vasek,
and D Sing. 2005. “Inverted T Channel FET (ITFET)-Fabrication and Characteristics
of Vertical-Horizontal, Thin Body, Multi-Gate, Multi-Orientation Devices, ITFET
SRAM Bit-Cell Operation. A Novel Technology for 45nm and beyond CMOS.” In
Proceedings on Technical Digest IEEE International Electron Devices Meeting,
IEDM, 713–716.
Miyano, S, M Hirose, and F Masuoka. 1992. “Numerical Analysis of a Cylindrical Thin-
Pillar Transistor (CYNTHIA).” IEEE Transactions on Electron Devices 39 (8). IEEE:
1876–1881.
Mizuno, T, N Sugiyama, T Tezuka, T Numata, and S Takagi. 2002. “High Performance CMOS
Operation of Strained-SOI MOSFETs Using Thin Film SiGe-on-Insulator Substrate.”
In Proceedings of Symposium on VLSI Technology, 2002. Digest of Technical Papers,
106–107.
Moore, G E, et al. 1998. “Cramming More Components onto Integrated Circuits.” Proceedings
of the IEEE 86(1): 82–85.
Park, D. 2006. “3 Dimensional GAA Transitors: Twin Silicon Nanowire MOSFET and Multi-
Bridge-Channel MOSFET.” In Proceedings of International SOI Conference, 2006
IEEE, 131–134.
Patro, B S, and B Vandana. 2016. “Low Power Strategies for beyond Moore’s Law Era: Low Power
Device Technologies.” Design and Modeling of Low Power VLSI Systems. IGI Global, 27.
Rim, K, J L Hoyt, and J F Gibbons. 1998. “Transconductance Enhancement in Deep Submicron
Strained Si N-MOSFETs.” In Proceedings of Technical Digest - International Electron
Devices Meeting, 1998. IEDM’98, 707–710.
Roy, K., S Mukhopadhyay, and Hamid Mahmoodi-Meimand. 2003. “Leakage Current
Mechanisms and Leakage Reduction Techniques in Deep-Submicrometer CMOS
Circuits.” Proceedings of the IEEE 91 (2). IEEE: 305–327.
Ryckaert, J. 2019. “Scaling below 3nm Node: The 3D CMOS Integration Paradigm
(Conference Presentation).” In Advanced Etch Technology for Nanopatterning VIII,
10963:109630O.
Sekigawa, T, and Y Hayashi. 1984. “Calculated Threshold-Voltage Characteristics of an
XMOS Transistor Having an Additional Bottom Gate.” Solid-State Electronics 27
(8–9). Pergamon: 827–828.
Semiconductor Industry Association. 2015. “The International Technology Roadmap for
Semiconductors.”
Suzuki, K, T Tanaka, Y Tosaka, H Horie, and Y Arimoto. 1993. “Scaling Theory for Double-
Gate SOI MOSFET’s.” IEEE Transactions on Electron Devices 40 (12). IEEE: 2326–2329.
Tiwari, S, M V Fischetti, P M Mooney, and J J Welser. 1997. “Hole Mobility Improvement in
Silicon-on-Insulator and Bulk Silicon Transistors Using Local Strain.” In Proceedings
of Technical Digest - International Electron Devices Meeting, 1997. IEDM’97, 939–941.
Vandana, B, J K Das, B S Patro, and S K Mohapatra. 2017. “Exploration towards Electrostatic
Integrity for SiGe on Insulator (SG-OI) on Junctionless Channel Transistor (JLCT).”
Facta Universitatis, Series: Electronics and Energetics 30(3): 383–390.
Vandana, B, S K Mohapatra, J K Das, and B S Patro. 2017. “Prospects of 2D Junctionless
Channel Transistor (JLCT) Towards Analog and RF Metrics Using Si and SiGe in
Device Layer.” Journal of Low Power Electronics 13 (3). American Scientific
Publishers: 536–544.
Vandana, B, Prashant Parashar, B S Patro, K P Pradhan, S K Mohapatra, and J K Das.
2019. “Mole Fraction Dependency Electrical Performances of Extremely Thin SiGe
on Insulator Junctionless Channel Transistor (SG-OI JLCT).” Advances in Signal
Processing and Communication. Springer: 573–581.
Nanoscale Semiconductor Devices 127
Abhishek Kumar
IIT Jodhpur
CONTENTS
7.1 Introduction .................................................................................................. 129
7.2 Moore’s Law and the International Technology Roadmap for
Semiconductors............................................................................................. 131
7.3 CMOS Scaling Challenges and Solutions with New FET Geometries ........ 132
7.4 NanoDevices Beyond Complementary Metal Oxide Semiconductor .......... 136
7.5 Technical Challenges and Solutions ............................................................. 142
7.6 Conclusion .................................................................................................... 144
Acknowledgements ................................................................................................ 144
References .............................................................................................................. 144
7.1 INTRODUCTION
Recently, the microelectronic industry has got massive benefits from the metal oxide
semiconductor field-effect transistor (MOSFET) scaling down. The shrinkage of
transistors to measurements beneath 100 nm empowers a vast number of transistors
to be located on a solitary chip. Transistor scaling gave the advantages of low-cost
manufacturing, high-speed data transfer, and multitasking ability.
Moore’s “law” and the International Technology Roadmap for Semiconductors
(ITRS) are complimenting each other since the early 1990s. Moore’s “law” is primar-
ily founded on the integration of the number of transistors into a microchip. The ITRS,
on the other hand, empowers the semiconductor industry to translate this reflection
into the real world. So carefulness is needed while explaining “Moore’s law,” as a
physical or numerical law. This initiates to discuss the role of this on the growth of
the semiconductor industry. So, in this chapter, Moore’s law is hashed out concisely
in conjunction with the ITRS primarily from the MOSFET scaling viewpoint [1–3].
The majority of the integrated circuits (IC) employ complementary metal oxide
semiconductor (CMOS) technology. Scaling down of MOSFET from μm to nm
129
130 Advanced VLSI Design and Testability Issues
has the advantages of cheaper circuits, smaller capacitances, higher IC speeds, less
power consumption, etc. But scaling down results in short gate length, which in turn
leads to the problems as follows:
Section 7.5 highlights the technical challenges faced by nanodevices and possible
solutions. Finally, concluding remarks are given.
Speed of operation is the key to measure the performance factor. Maximum input
is the off-chip frequency, and the signal frequency is the output to high-performance
devices. The on-chip local frequency is not raising quicker than off-chip frequency
according to the recent edition of the ITRS.
The frequency of operation is highly dependent on switching speed of CMOS logic
components that are constituted by transistors. The frequency hence is inversely pro-
portional to the delay time that is taken to propagate the signal through the CMOS
logic circuit. In a technology node, the delay time of an inverter can be used as an
approximation of delay time for that technology, which empirically is calculated as
CMOS delay time = gate capacitance (drain voltage/drain saturation current)
While scaling the device, since CMOS delay time is inversely proportional to the
factor by which it is scaled, it allows faster circuit operations. The packing density of
transistors is inversely proportional to the total chip area, and hence, the density will
increase by a factor of κ 2, where κ ≈ 2 is the scaling constant [1–5].
instead of the oxy-nitride. For scaling around 45 nm, the rule is to enclose
metal gates and substitute SiO2 by dielectric materials of high-k. Zirconium
dioxide (ZrO2), hafnium oxide (HfO2), and aluminum oxide (Al2O3) have
higher relative dielectric constants than SiO2. In case of high-k gate dielec-
tric, there exists always smaller gate leakage current for huge energy barrier
among silicon and the dielectric.
D. Problems beyond 45 nm:
Transit frequency ft is generally used as a speed performance parameter
of transistors. This is the frequency where the generalized small-signal cur-
rent gain is unity. At high electrical fields, due to short channel and satura-
tion effects, the gate current (ig)–based scaling of ft has put down to ~ ig − 1
all over the 45-nm node. Consequently, the scaling factor is getting doubled
during the scaling from 90 to 45 nm. So a decrease in ig by a factor of 2
builds the ft by a factor of 4. Albeit, due to very high drain–source small-
signal output conductance (gds) in case of aggressively scaled CMOS cir-
cuits and many “beyond CMOS” accesses, irrespective of the load gL, no
small-signal voltage gain can be attained any longer. The higher intrinsic
voltage gain gm /gds restricts the maximum attainable voltage gain in a circuit,
where gm is the transconductance. As a result, there is an incompatibility
with high-speed architectures of circuits having voltage gain–dependent
functionality. So it is nonmeaningful of utilizing the benchmark parameter
as only ft. Hence, in addition to ft, at least the intrinsic gain gm /gds must be
considered as a parameter. Another efficient parameter for performance can
be the maximum frequency of oscillation fmax, because without utilizing any
impractical terminations, it admits entire transistor parasitics. Next, perfor-
mance parameters can be the noise figure and linearity. At 10-nm node, the
strong effect of scaling on ft can be less sounded out with ~ ig-x and 0 < x < 1,
to a greater degree or extent. To void unreasonable electrical fields in the
channel, drain–source voltage (Vds) has to be reduced with ig, leading to a
Vds around 0.5 V for the 10-nm node. Furthermore, due to disproportion-
ate scaling of the threshold voltage Vth, the value of Vds − Vth is importantly
reduced, and this, in turn, challenges the driving of the succeeding stage. At
45-nm node, the value of gm /gds is being decreased from 15 to 5 because of an
increase in gds. Ongoing of this phenomenon, calculation at 10 nm will lead
to nearly unity intrinsic gain. The possible solution is to use gate-all-around
structures and multiple gates in a FET. The main aim of both the structures
is to maximize gate-to-channel capacitance and minimize drain-to-channel
capacitance. Figure 7.2 illustrates the scaling disputes in planar devices.
E. Scaling around 22 nm:
At 22 nm, for the structures, gate length scaling can be achieved beyond
the boundary of mainstream MOSFET. With the lessen oxide thickness, the
gate acquires exact command on the channel, just merely right at the Si sur-
face. Reduction in channel length leads to shortened drain-to-channel dis-
tance. More control of the drain than the gate is experienced along some
distant paths below the Si surface where the control of the gate is weak since
it is far away. As Vgs < Vt, an N-channel MOSFET acquires the off state.
134 Advanced VLSI Design and Testability Issues
FIGURE 7.2 Scaling disputes in planar devices. SCE, short-channel effect; NMOS, n-type
metal oxide semiconductor.
Albeit, there will be a leakage current streaming among the drain and source.
This also makes an issue of serious power consumption in standby operation.
Hence, there is a requirement of maximizing the gate-to-channel capacitance
while minimizing the drain-to-channel capacitance. As a result, there will be
a reduction in drain leakage current. Drain-to-source and the drain-to chan-
nel distance gets reduced when the channel length is reduced. This, in turn,
increases the influence of the drain. The leakage paths are still under the
surface, although oxide thicknesses are thin. This is because the gate has a
perfect control on the Si surface, and deep in the channel, the drain has more
control. For eradicating deeply submerged leakage, the gate should have more
control over the channel. This is attained by giving gate control from several
sides of the channel. Due to the very thin Si film, there is less probability of
the existence of leakage path far from one of the gates. More than one gate
structure is called multigate MOSFET or sometimes FinFET.
• FinFET
Considering any technology node, the FinFET provides more advan-
tages over its planar counterpart as stated in the following:
– It provides better electrostatic control over the channel along with
easy “choking off” of channel. FinFETs show nearly ideal sub-
threshold behavior, which is not easily achieved in planar technol-
ogy without significant design endeavor.
– It provides reduced short-channel effects, which is too complicated
in planar technology and gives rise to an enormous impact on gate
length variations and therefore on electrical performance.
– It has high integration density or 3D structure due to its vertical
channel orientation, which delivers more performance per linear
CMOS-Based FET 135
“W” than planar even after the isolation dead area among the fins
is taken into consideration.
– It has a minimal variability, mainly variability arising from random
dopant fluctuation mainly due to doping-free channels. Also, vari-
ability associated with line-edge roughness (LER), the random dif-
ference of gate line edges from the intended ideal shape, that results
in nonuniform channel lengths is lower in FinFETs.
– FinFET requires much lower dopant concentrations in the channel
region.
– Since the gate is defined from the top of the fin, the predominant part
of the gate is defined by etching processes, which have very low LER.
– Furthermore, it provides the ability to operate at much lower sup-
ply voltages and extends voltage scaling, which was leveling off in
CMOS devices, and allows further static and dynamic power savings.
– FinFET device technology is the most assuring device technol-
ogy for broadening Moore’s law all the way down to 10 nm or even
toward 7 nm. It is fully compatible with CMOS in both bulk and
silicon-on-insulator varieties. Also, FinFET technology is fully
compatible with the CMOS back-end design processes, reducing
the need for new FinFET-specific developments in that area.
Albeit, no new technology is completely free of risk or challenges.
The challenges faced by FinFET technology are as follows:
– The modeling of FinFET is a significantly more complicated pro-
cess. Both precise FinFET parasitic extraction and generation of
good yet compact SPICE models are challenging to a greater extent
with FinFETs than with planar devices. For most design activities,
the mentioned complexities are transparent to the designer. Albeit,
there remain optimization challenges for the circuit designer who
wishes to utilize FinFET technology.
– Analog, as well as digital, design optimization becomes more com-
plicated due to limited selection in channel length range and finite
granularity of the fin width. Although the desired fin width can be
generated by ganging together many fins, the length and width can-
not be selected freely. This is because of the 3D structure of FinFETs,
which does not allow controlling variability for the high-aspect-ratio
processes with nonuniform pitches or locally varying pitches. Thus,
there are a significant number of restricted design rules for FinFETs.
– FinFET structure has better gate control and lower threshold volt-
age with less leakage. But, at lower technology node, say, below
10-nm node, the issue of leakage begins again. This contributes to
many other issues such as threshold flattening, an increase in power
density, and thermal dissipation.
A. NW Transistors:
It is a type of transistor where the channel is replaced or created with
very thin (few atomic length) wire made up of material different than a
silicon substrate. It has been extensively researched, and many material
alternatives and different geometries have been proposed. NW FETs, also
known as gate-all-around or surrounding-gate FETs, offer better scaling
opportunities as their nonplanar geometry offers better electrostatic con-
trol of channel than in conventional FETs. Researchers and industrialists
are garnering attention in NW research and development while banking
on several key factors, including their high-yield reproducible electronic
properties, higher carrier mobility, smooth surfaces, and producing radial
and axial NW heterostructures. Apart from these, the better scaling fac-
tor (below 10 nm) and cost-effective bottom-up fabrication have been the
primary focus in industrial acceptability. However, smaller NW diameters
create quantum confinement, resulting in the change of inversion charge
from the surface to bulk inversion. These fabrication imperfections cause
variations in actual NW dimensions, leading to perturbations in charge car-
rier potential and scattering, which causes degradation in charge transport
CMOS-Based FET 137
TABLE 7.1
Comparison of Graphene, CNT, and Nanowire FETs [10]
Properties Graphene CNT Nanowire
Transport Two-dimensional One-dimensional,
Quasi one-dimensional for
Quantum effectsextremely small diameters,
otherwise three-dimensional
Potential for high Device width of few Yes, provides Provides if device diameter is
linearity nanometers is required very small
Ambipolar properties Possible Possible Possible
On/off ratio Poor since no or small Currently poor due Good
bandgap to metallic tubes
The flexibility and versatility of OFETs have enabled them for a wide
range of new applications, for instance, in the field of wearable systems,
bioelectronics, or as driving elements in a flat panel display. Impressive
progress has been achieved in the design, synthesis, and processing of
organic semiconductors in the past few years. Organic semiconductors can
be more attractive owing to comparable functioning as traditional amor-
phous inorganic semiconductor materials and their near-infinite tunability.
Recently, more attention is given toward the solution-processable, air-
stable HP organic n-type semiconductor. Meanwhile, some critical issues
still need further investigation such as operational stability, low-cost and
large-area fabrication process, device integration, as well as functionaliza-
tion in sensor fields. The study of the electronic defect structure of organic
semiconductors is the main area of research investigation. Along with
device optimization, producing novel organic semiconductor materials and
employing thin-film alignment techniques are other directions to achieve
HP devices. It is anticipated that by combining correct organic semiconduc-
tor materials and earmark fabrication techniques, HP devices for various
applications could be obtained. Recent research has highlighted that vertical
organic transistors give assurance of increasing the performance of flexible
organic transistors while keeping their low-cost advantage. Albeit vertical
organic transistors have not been adopted for inorganic semiconductors, they
may show some key advantages for organic semiconductors [13,14,20].
E. Spin Transistors:
A spin transistor or spintronic transistor is a magnetically sensitive tran-
sistor. This is currently still being developed. It is also named for spintronics.
Spin electronics caters to the analysis of the intrinsic spin of the electron and
its related magnetic moment, in addition to its primal electronic charge, in
solid-state devices. In spintronics, along with charge state, electron spins are
put upon as a further degree of freedom with implications in the efficiency of
data storage and transfer. Spintronic systems are regularly acknowledged in
dilute magnetic semiconductors and Heusler alloys. They are quite compel-
ling in the field of quantum computing and neuromorphic computing.
One advantage over regular transistors is that these spin states can be
modified without the application of an electric current. This permits for the
detection of hardware that is much smaller and sensitive. The raised sensi-
tivity of spin transistors is also being studied in producing more sensitive
automotive sensors. A second advantage is that due to the semipermanent
spin of an electron, it can be used as a means of creating cost-effective
nonvolatile solid-state storage that does not need the constant application of
current to sustain. It is the technologies that are being explored for magnetic
random access memory. Because of its high potential for practical use in the
computer world, spin transistors are currently being researched in various
firms throughout the world, such as in England and in Sweden [15].
F. SETs:
Due to their tiny size and reasonable low-power dissipation rate, SETs
are desirable devices for future large-scale integration. The SET consists
CMOS-Based FET 141
of three terminals, for example, drain, gate, source, and the second gate
is optional. SET has a tiny conductive island coupled to a gate electrode
with gate capacitance CG. The operation of SET is based on islands that
are small conducting particles surrounded by insulating material. An island
can be charged only by a small number of electrons. Due to the insulating
layer that circumvents the island, the transport of electrons to and from the
island is possible only by tunneling effect. Thus, the voltage of each island
is only in the form of quantized values. Source and drain electrodes are
joined to the island through a tunnel barrier (junction). The tunnel bar-
rier that controls the motion of every single electron has two conductors
separated by thin layer, and it is modelled as tunneling resistances R DS and
junction capacitances CDS. The increased gate bias attracts electrons to the
island only through either drain or source tunnel barrier, and the number
of electrons in the island only has a fixed integer. Therefore, the increased
gate bias makes electrons flow one by one when a small voltage is applied
between the source and drain electrodes by means of the “Coulomb block-
ade” phenomenon. Figure 7.4 shows the schematic of a SET.
Its operation is alike to that of MOSFET except that electron conduction
takes place in one electron at a time, whereas in MOSFET, many electrons
simultaneously take part in the conduction. With proper gate voltage appli-
cation, the potential energy of the conduction island is made low enough to
allow one electron from the source to tunnel to the conduction island. With
the potential energy of the drain lower than that of the conduction island,
the electron then tunnels to the other side to reach the drain. With the con-
duction island empty and the potential lower again, the process repeats. The
SET fabrication process in silicon is CMOS compatible. The use of SETs
TABLE 7.2
Applications and Limitations of Nanoelectronics Devices
Device Applications Limitations
Nanowire transistors In faster and smaller more dense • Scattering of electrons through
processors wire lateral surface when
propagation length is smaller
than average mean free path
for electrons in given material,
resulting in smaller than
expected current.
• Nonuniformity in nanowire
geometry is the available
industrial process technology,
making it difficult for mass
production
Graphene • Appeared as an alternate to • Control of geometry
nanoribbon transistors silicon-channels FETs with high • Process control is still not
sensitivity achieved
• Finds usefulness in certain • So it is ineligible for mass
applications including bio- and production
chemical sensing applications
Carbon nanotube • Attractive as future interconnects • Tighter gate control through
field-effect transistors • Very attractive to Si-based ultrathin high-k gate dielectrics
semiconductor industry is demanded
• The starting material must be
purified
Single electron • Useful for implementing binary Implementation is still in theoretical
transistors logic circuits and experimental phases
• Useful for implementing
multivalued (MV) logic circuits
and binary MV mixed logic
circuits
• Useful for scientific
instrumentation and metrology
Organic field-effect • Very useful in biodegradable • Large in size
transistors devices • More power required for
• Achieving good applications in operation
sensors and LEDs and flexible
transistors
Spin transistors Very good for memory devices Implementation is still in theoretical
and experimental phases. Few
working chips have been developed
144 Advanced VLSI Design and Testability Issues
7.6 CONCLUSION
In this chapter, we have reviewed various scaling down aspects of MOSFET, pos-
sible new FET geometries, need for nanoelectronic devices, and challenges to the
design of nanoelectronic devices with some solutions. Although substantial transis-
tor challenges arise for technologies past 45 nm, various possible solutions are being
explored to drive Moore’s law forward. Although nanoelectronics is the solution for
technologies past 22 nm, it is essential to build up a novel knowledge and reliability
prototype for nanoelectronics to enable industries for predicting, optimizing, and
designing direct reliability and performance of nanoelectronics.
ACKNOWLEDGEMENTS
The authors acknowledge Mohan Krishna (fourth-year ECE student) and Goutham
(second-year ECE student) of MITS (Andhra Pradesh) for their support in chunking
of materials.
REFERENCES
1. Hiroshi, I, May, 2016. End of the scaling theory and Moore’s law. IEEE (IWJT) pp: 1–4.
2. Zeitzoff, P.M., and Chung, J.E., 2005. A perspective from the 2003 ITRS: MOSFET
scaling trends, challenges, and potential solutions. IEEE Circuits and Devices
Magazine, 21(1), pp: 4–15.
3. Meindl, J.D., 2003. Beyond Moore’s law: The interconnect era. Computing in Science
& Engineering, 5(1), pp: 20–24.
4. Kuhn, K.J., May, 2009. Moore’s Law past 32nm: Future challenges in device scaling.
13th International Workshop on Computational Electronics (IEEE), pp: 1–6.
5. ITRS Reports online: http://www.itrs2.net/itrs-reports.html.
6. M.T. Abuelma’atti, 2012. MOSFET scaling crisis and the evolution of nanoelec-
tronic devices: The need for paradigm shift in electronics engineering education. 6th
International Forum on Engineering Education (ELSEVIER), pp: 432–437.
7. Y-B Kim, 2010. Review paper: Challenges for nanoscale MOSFETs and emerging nano-
electronics. Transactions on Electrical and Electronic Materials, 11(3), pp: 93–105.
8. Razavieh, A., Zeitzoff, P., and Nowak, E.J., 2019. Challenges and limitations of CMOS
scaling for FinFET and beyond architectures. IEEE Transactions on Nanotechnology,
18, pp: 999–1004.
9. Horowitz, M., Alon, E., Patil, D., Naffziger, S., Kumar, R., and Bernstein, K., 2005.
Scaling, power, and the future of CMOS. IEEE International Electron Devices Meeting
(IEDM Technical Digest), pp: 7–14.
10. Ellinger, F., Claus, M., Schröter, M., and Carta, C., 2011. Review of advanced and beyond
CMOS FET technologies for radio frequency circuit design. IMOC (IEEE), pp: 347–351.
CMOS-Based FET 145
11. Vora, P.H., and Lad, R., 2017. A Review Paper on CMOS, SOI and FinFET Technology,
Industry article.
12. Geim, A.K., and Novoselov, K.S., 2007. The rise of grapheme. Nature Materials, 6,
pp: 183–191.
13. Chang, J., Lin, Z., Zhang C., and Hao Y., 2017. Organic field-effect transistor: Device
physics, materials, and process. INTECH, 125–145.
14. Ostroverkhova, O., 2018. Handbook of Organic Materials for Electronic and Photonic
Devices. Woodhead Publishing Series in Electronic and Optical Materials. Elsevier,
Sawston, Cambridge, UK, pp: 875–891.
15. Joshi, V.K., 2016. Spintronics: A Contemporary Review of Emerging Electronic
Devices. Elsevier, Sawston, Cambridge, UK, pp: 1503–1513.
16. Haron, N.Z., and Hamdioui, S., December 2008. Why is CMOS scaling coming to an
END? 3rd International Design and Test Workshop (IEEE), pp: 98–103.
17. Stillmaker, A., and Baas, B., 2017. Scaling equations for the accurate prediction of CMOS
device performance from 180 nm to 7 nm. Integration (ELSEVIER) 58, pp: 74–81.
18. Bohr, N., 1935. Can quantum-mechanical description of physical reality be considered
complete? Physical Review, 48(8), pp: 696–702.
19. Jacob, A.P., Xie, R., Sung, M.G., Liebmann, L., Lee, R.T., and Taylor, B., 2017.
Scaling challenges for advanced CMOS devices. International Journal of High Speed
Electronics and Systems, 26(01n02), pp: 1–76.
20. Varshney, G., Gotra, S., Pandey, V.S., and Yaduvanshi, R.S., 2019. A proximity coupled
two-port MIMO graphene antenna with pattern diversity for THz applications. Nano
Communication Networks, 21, pp: 456–463.
21. Liu, C., Zhang, X., Zhang, J., Muruganathan, M., and Mizuta, H. 2020. Origin of non-
linear current-voltage curves for suspended zigzag edge graphene nanoribbon. Carbon.
https://doi.org/10.1016/j.carbon.2020.05.010
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
8 Analytical Design of
FET-Based Biosensors
Khuraijam Nelson Singh and Pranab Kishore Dutta
NERIST
CONTENTS
8.1 Introduction .................................................................................................. 147
8.2 Types of Biosensors ...................................................................................... 149
8.2.1 Electrochemical Biosensor ............................................................... 149
8.2.2 Optical Biosensor.............................................................................. 150
8.2.3 Piezoelectric Biosensor..................................................................... 150
8.2.4 Calorimetric Biosensor ..................................................................... 150
8.3 Field-Effect Transistor–Based Biosensors .................................................... 150
8.3.1 Working of Field-Effect Transistor–Based Biosensor ...................... 151
8.3.2 Some Common Types of Field-Effect Transistor–Based
Biosensors ......................................................................................... 152
8.3.2.1 Ion-Sensitive Field-Effect Transistor Biosensor ................ 152
8.3.2.2 Nanowire Field-Effect Transistor Biosensor ...................... 153
8.3.2.3 Carbon Nanotube Biosensor .............................................. 153
8.3.2.4 Dielectrically Modulated Field-Effect
Transistor Biosensor........................................................... 154
8.3.2.5 Tunnel Field-Effect Transistor Biosensor .......................... 155
8.3.2.6 Junctionless Field-Effect Transistor Biosensor .................. 155
8.4 Modeling of Field-Effect Transistor–Based Biosensors ............................... 156
8.4.1 Modeling of Dielectrically Modulated Field-Effect
Transistor–Based Biosensors ............................................................ 158
8.4.1.1 Surface Potential ................................................................ 158
8.4.1.2 Electric Field ...................................................................... 160
8.4.1.3 Threshold Voltage .............................................................. 162
8.4.1.4 Sensitivity........................................................................... 163
8.5 Summary ...................................................................................................... 165
References .............................................................................................................. 165
8.1 INTRODUCTION
Research on biosensors has seized the interested researchers over the past few
decades due to their various advantages and applications. They are used in the dis-
covery of drugs, monitoring of diseases, agriculture, food quality control, industrial
wastage monitoring, military, etc. [1]. The sensing analyte is the main element that
differentiates a biosensor from the other physical/chemical sensors. In general, the
147
148 Advanced VLSI Design and Testability Issues
The invention of the first biosensor by Clark and Lyons in 1962 marked the start of
a new era of biomolecules detection [3]. The biosensor was used for the detection
of glucose. Diabetes is a widespread chronic disease that not only affects adults but
children too. In 2016, the World Health Organization (WHO) estimated that 1.6 mil-
lion people died due to diabetes [4]. Such severeness has led to the rise in the demand
for blood glucose detector globally. The increasing demand, in turn, increases the
research of glucose detection using biosensors.
Similarly, the demand for a faster, cheaper, and easier method for detection of bio-
molecules in various fields has led to the advancement of biosensor research. With
the recent achievements, detection of a disease-specific biomarker in vitro as well
as in vivo has achieved high accuracy [5]. The biosensors are proven to be highly
sensitive in the detection of disease markers such as lactate, cytokines, proteins, and
antibodies. These biosensors can also detect the presence of the target biomolecules
even when the bonded amount of biomolecules is meagre.
Some of the advantages of using biosensors are as follows [6]:
i. They can be used for the detection of both ionic and nonionic biomolecules.
ii. They can be used to detect a minimal amount of target biomolecules. Thus,
their sensitivity is very high.
iii. They can be used to monitor any specific biomolecules continuously.
Analytical Design of FET-Based Biosensors 149
i. Electrochemical biosensor
ii. Optical biosensor
iii. Piezoelectric biosensor
iv. Calorimetric biosensor
In a nucleic acid biosensor, DNA is bonded onto the surface of the electrode. The
interaction of the DNA and the working electrode causes a hybridization reaction,
resulting in a change of conductivity of the device. In a label electrochemical biosen-
sor, the analyte is bonded in between the label and the working electrode. A label is a
substance that on interaction with the analyte produces a change, which can be used to
detect the presence of the analyte. Although label biosensor provides accurate results,
it needs a solid binding of the labeling substance as the detection process depends on
the specific reaction that takes place between the label and the biomolecule of inter-
est, whereas a label-free biosensor detects the changes when the target biomolecules
interact with the sensing surface of the biosensor scrapping the sandwiching process
of biomolecules in between the label and the working electrode. Label-free detec-
tion process provides faster analysis, lower reagent cost, and continuous monitoring
capability in real time. It also allows detection of the analyte without any alteration
to its natural form. Popular examples of the commercially available electrochemical
biosensor are pregnancy test strips and urine analysis strips.
150 Advanced VLSI Design and Testability Issues
The interaction of the electrochemical biosensor and the analytes produces vari-
ous types of signals. Based on these signals, the electrochemical biosensor can be
further classified as amperometric, potentiometric, conductometric, and FET-based
biosensors [2].
the detection of analytes without the labeling process. Thus, the detection process
becomes easier. The advancement in the metal oxide semiconductor FET (MOSFET)
study also provides high miniaturization capability, which allows in vivo detection of
the analytes. Its compatibility to integrate with various IC technologies, fast processing
speed, ability to sense in parallel, and lower cost further expand its scope of application.
FIGURE 8.2 Schematic diagram of a nanogap FET biosensor. FET, field-effect transistor.
152 Advanced VLSI Design and Testability Issues
site with the biomolecules changes the dielectric, and charge of the region changes,
which changes the sensing parameter of the biosensor (threshold voltage) as shown
in Figure 8.3. Due to the maturity of FET technology, it gives the advantage of the
relative ease of theoretical study and fabrication.
voltage changes [24]. The threshold voltage also depends on the charge and dielectric
constant of the bonded biomolecules.
FIGURE 8.7 Schematic diagram of TFET biosensor. TFET, tunnel field-effect transistor.
156 Advanced VLSI Design and Testability Issues
FIGURE 8.8 Schematic diagram of junctionless FET biosensor. FET, field-effect transistor.
When using JLT as a biosensor, a part of the gate oxide is etched in which the bio-
molecules are bonded, as shown in Figure 8.8 [31]. Upon interaction of the biomol-
ecules and the gate oxide of the sensor, the effective dielectric of the nanogap region
changes, resulting in the shift of the gate capacitance. These changes are captured as
a change in the electrical properties of the biosensor, such as drain current, threshold
voltage, and so on.
FIGURE 8.9 Schematic structure of underlap FET biosensor. FET, field-effect transistor.
the device channel. Equation 8.1 represents the Poisson’s equation, upon solving of
which the electric potential of the device is determined [13].
∂2 φ j ( x , y ) ∂2 φ j ( x , y ) qNa
+ = (8.1)
∂x 2 ∂ y2 cSi
where j = 1 or 2 and represents the two regions of the device channel. ϕj (x, y) is the
potential in the device channel, q is the electronic charge, Na is the channel doping
concentration, and 𝜖 Si is the silicon dielectric value.
The potential in the channel is then appoximated by using a parabolic approxima-
tion given by
φ j ( x , y ) = K j 0 ( x ) + K j 1 ( x ) y + K j 2 ( x ) y 2 (8.2)
where Kj0 (x), Kj1 (x), and Kj2 (x) are all constants that depend only on x-coordinate,
and their values are determined by solving the equations that exist in the boundary
of the channel.
The Poisson’s equation is then simplified into a 1D nonhomogeneous partial dif-
ferential equation:
∂ 2 φ fs , j ( x )
− A jφ fs , j ( x ) = B j (8.3)
∂ x2
The solutions of the simplified Poisson’s equation give the electric potential profile
in the channel:
φ fs1 ( x ) = Pe
1
λ1x
+ P2e− λ 1 x + γ 1 , for R-I (8.4)
The electric potential can be used to find the threshold voltage of the device. The
value of the gate voltage where the potential becomes lowest gives the threshold
voltage of the device.
∂ 2 φ j ( x , y ) ∂ 2 φ j ( x , y ) qNa
+ = (8.6)
∂ x2 ∂ y2 cSi
FIGURE 8.10 Schematic diagram of UDG DMFET. UDG DMFET, underlap double-gate
dielectrically modulated field-effect transistor.
Analytical Design of FET-Based Biosensors 159
where j = 1 or 2 and it corresponds to R-I or R-II of the device as shown in Figure 8.5,
q is the electronic charge, Na is channel doping concentration, and 𝜖 Si is dielectric
value of silicon.
Since the potential in the channel has parabolic nature, it can be approximated as [32]
φ j ( x , y ) = K j0 ( x ) + K j1 ( x ) y + K j2 ( x ) y 2 (8.7)
where the x-dependent contants Kj0, Kj1, and Kj2 are solved by using the boundary
conditions shown in the following:
∂φ j ( x , y ) C
= j ⎡⎣φ fsj ( x ) − Vgsj
'
⎤⎦ (8.10)
∂y y= 0
c Si
∂φ j ( x , y ) C
= − j ⎣⎡φbsj ( x ) − Vg'sj ⎤⎦ (8.11)
∂y y = tch
cSi
where ϕfsj(x) and ϕbsj(x) are the front and back surface potential of the jth region. The
binding of the biomolecules causes the capacitance in R-I (C1) to change, whereas the
capacitance in R-II (C2) remains the same which values are given in the following:
Cox 2Cbio c 2c ⎡ ⎛ t +t ⎞⎤
C1 = , Cox 2 = ox , Cbio = bio sinh ⎢ cosh −1 ⎜ 1 M ⎟ ⎥
Cox 2 + Cbio t2 nπ L1 ⎣ ⎝ t1 ⎠ ⎦
Cox 2Chigh c
C2 = , Chigh = high− K
Cox 2 + Chigh t1
where 𝜖ox, 𝜖high, and 𝜖bio are dielectric constants of SiO2, high-k, and biomolecule,
respectively; Cbio is the fringing capacitance due to binding of biomolecules [33], and
qQbio
VFB1 = VFB 2 − , VFB2 = φ M − φ Si
C1
where Qbio, ϕM, and ϕ Si are charge density of biomolecules, work functions of gate
electrode, and work function of silicon, respectively.
Now, substituting the values of Kj0, Kj1, and Kj2 in Equation (8.7), the Poisson’s
equation can be simplified as
160 Advanced VLSI Design and Testability Issues
∂ 2 φ fsj ( x )
− A jφ fsj ( x ) = B j (8.12)
∂ x2
where
2C j qN a 2C j '
Aj = φ fsj ( x ) , B j = − Vgsj
cSitSi cSi cSit Si
φ fs1 ( x ) = Pe
1
λ1x
+ P2e− λ 1 x + γ 1 , for R-I (8.13)
where λj = A j , and γ j = − B j . The values of P1, P2, P3, and P4 can be solved by
Aj
using the boundary conditions given in the following:
φ fs1 ( L1 ) = φ fs 2 ( L1 ) (8.16)
∂ φ 1 ( x , y ) ∂ φ 2 ( x , y )
= (8.18)
∂x x = L1
∂x x = L1
where Vbi is the built-in potential and Vds is the drain terminal voltage.
Figure 8.11 shows the effect of the binding of anti-AI and DNA on the sur-
face potential profile of UDG DMFET. It shows that binding of the biomole-
cules shifts the surface potential plot from its original position before the binding
happens. And since DNA has a larger value of dielectrics and charge value, the
change it brings is more in comparison with that of anti-AI. If the sensing site
length increases, the area for loading biomolecules increases. The increase in
biomolecules binding causes a greater shift in the surface potential, as shown in
Figure 8.12.
dφ fs1 ( x )
E1 = = λ 1 ( Pe
1
λ 1x
− P2e− λ 1 x ) , for R-I (8.19)
dx
Analytical Design of FET-Based Biosensors 161
FIGURE 8.11 Variation of surface potential along UDG DMFET channel when L1 : L2 = 1:1.
UDG DMFET, underlap double-gate dielectrically modulated field-effect transistor.
FIGURE 8.12 Variation of surface potential along UDG DMFET channel when L1 : L2 = 2:1.
UDG DMFET, underlap double-gate dielectrically modulated field-effect transistor.
dφ fs1 ( x )
E2 =
dx
( )
= λ 1 P3eλ 2 ( x − L1 ) − P4e− λ 2 ( x − L1 ) , for R-II (8.20)
Figure 8.13 shows the variation in the electric field of the UDG DMFET along its
channel. It is observed that binding of the biomolecules does not change the electric
field in the drain side of the UDG DMFET. Thus, no hot electron effect happens due
162 Advanced VLSI Design and Testability Issues
FIGURE 8.13 Effect on the electric field along the channel of UDG DMFET when biomol-
ecules bind when L1 : L2 = 1:1. UDG DMFET, underlap double-gate dielectrically modulated
field-effect transistor.
to the binding of the biomolecules on the device. The figure also shows that binding
of the biomolecules changes the electric field in the R–I of the device. The high elec-
tric field in the source side also causes the injection of the electrons from the source
to channel to increase. Thus, the mobility of the device also increases. The effect of
increasing the sensing site to the electric field is shown in Figure 8.14. As more bio-
molecules are bonded, a greater shift in the electric field is produced. But increasing
the sensing site still does not cause any hot electron effect.
where xmin and ψF are the minimum potential location and fermi potential.
In a MOSFET with a multiple-gate electrode, the higher work function deter-
mines the location of the minimum potential [39,40]. But, in this study, there is a
variation in the electrical parameters due to the biomolecules binding. Thus, thresh-
old voltage has been a model for the two regions.
The minimum potential location is determined by solving the following equation:
dφ fs , i ( x )
=0 (8.22)
dx x = xmin
Analytical Design of FET-Based Biosensors 163
FIGURE 8.14 Effect on the electric field along the channel of UDG DMFET when biomol-
ecules bind when L1 : L2 = 2:1. UDG DMFET, underlap double-gate dielectrically modulated
field-effect transistor.
Substituting xmin value in Equation (8.21) and then solving the equation, Vth is found as
C2 + C22 − 4C1C3
Vth = , for R-I (8.23)
2C1
C5 + C52 − 4C4C6
Vth = , for R-II , for R-II (8.24)
2C4
Figure 8.15 shows the effect of binding the anti-AI and DNA on the threshold voltage of
the UDG DMFET. It is observed that binding of biomolecules anti-AI and DNA changes
the threshold voltage. Both anti-AI and DNA have negative charges. The negative charge
of the biomolecules hinders the charge inversion in the channel; thus the threshold volt-
age increases. Their dielectric values are higher than unity (air) which helps the mol-
ecules to oppose the external electric filed, which tries to align the molecules. Thus, the
higher the negative charge and dielectric value, the larger the threshold voltage.
8.4.1.4 Sensitivity
The sensitivity of FET-based biosensor is defined as the change in the threshold volt-
age of the device due to the biomolecules binding [41]. It is represented as
Figure 8.16 presents the sensitivity of the UDG DMFET as a biosensor. UDG
DMFET has a sensitivity of 200 mV when bonded with anti-AI and of 440 mV when
164 Advanced VLSI Design and Testability Issues
FIGURE 8.15 Threshold voltage of UDG DMFET with and without binding of biomol-
ecules. UDG DMFET, underlap double-gate dielectrically modulated field-effect transistor.
FIGURE 8.16 Sensitivity of UDG DMFET in detecting anti-AI and DNA. UDG DMFET,
underlap double-gate dielectrically modulated field-effect transistor; anti-AI, antibody of
avian influenza.
bonded with DNA. Thus, enough shift in threshold voltage is produced when bio-
molecules are bonded to the UDG DMFET. It shows that UDG DMFET is a very
option to choose for the detection of such biomolecules. On the top of having high
sensitivity or massive threshold voltage shift, binding of biomolecules produces extra
short-channel effect (SCE). Thus, such FET-based biosensors provide outstanding
performance for detection of biomolecules, which provide high sensitivity, lower
cost, label-free detection, and faster processing.
Analytical Design of FET-Based Biosensors 165
8.5 SUMMARY
The application of the FET device as a biosensor is presented in this chapter.
Different types of FET-based biosensor are reviewed to ingrain with the basic and
the available structures. This chapter focuses on UDG DMFET, as it has a simple
structure, good electrical characteristics, and high structural stability. The study has
been made based on the modeling of the device. All the model data have been veri-
fied by the simulation results of ATLAS (SILVACO). Electrical parameters such as
surface potential, electric field, threshold voltage, and sensitivity are discussed in
detail to understand the effect of biomolecules binding on the electrical character-
istics of the device. The higher the value of dielectric and charge of the device, the
higher the change. The binding of the biomolecules causes no hot electron effect, as
the electric field toward the drain is not affected by the binding. Thus, UDG DMFET
is an excellent choice in the detection process of biomolecules, as it has high sensitiv-
ity as well as low SCEs. The low supply voltage and size is a factor that contributes
immensely in lowering the power consumption of the device.
REFERENCES
1. Koyun, A., Ahlatcolu, E., Koca, Y., & Kara, S. (2012). Biosensors and their principles. A
Roadmap of Biomedical Engineers and Milestones, IntechOpen, London, pp. 117–142.
2. Thévenot, D. R., Toth, K., Durst, R. A., & Wilson, G. S. (2001). Electrochemical biosen-
sors: recommended definitions and classification. Analytical Letters, 34(5), 635–659.
3. Clark Jr, L. C., & Lyons, C. (1962). Electrode systems for continuous monitoring in
cardiovascular surgery. Annals of the New York Academy of Sciences, 102(1), 29–45.
4. Roglic, G., & World Health Organization (Eds.). (2016). Global Report on Diabetes.
Geneva: World Health Organization.
5. Malima, A., Siavoshi, S., Musacchio, T., Upponi, J., Yilmaz, C., Somu, S., & Busnaina,
A. (2012). Highly sensitive microscale in vivo sensor enabled by electrophoretic
assembly of nanoparticles for multiple biomarker detection. Lab on a Chip, 12(22),
4748–4754.
6. Koyun, A., Ahlatcolu, E., Koca, Y., & Kara, S. (2012). Biosensors and their principles. A
Roadmap of Biomedical Engineers and Milestones, IntechOpen, London, pp. 117–142.
7. Patel, S., Nanda, R., Sahoo, S., & Mohapatra, E. (2016). Biosensors in health care: the
milestones achieved in their development towards lab-on-chip-analysis. Biochemistry
Research International, 1, 1–12.
8. Pohanka, M. (2018). Overview of piezoelectric biosensors, immunosensors and DNA
sensors and their applications. Materials, 11(3), 448.
9. Ivnitski, D., Abdel-Hamid, I., Atanasov, P., & Wilkins, E. (1999). Biosensors for detec-
tion of pathogenic bacteria. Biosensors and Bioelectronics, 14(7), 599–624.
10. Park, S. C., Cho, E. J., Moon, S. Y., Yoon, S. I., Kim, Y. J., Kim, D. H., & Suh, J. S.
(2007). A calorimetric biosensor and its application for detecting a cancer cell with
optical imaging. In World Congress on Medical Physics and Biomedical Engineering
2006 (pp. 637–640). Springer, Heidelberg.
11. Hundeck, H. G., Weiss, M., Scheper, T., & Schubert, F. (1993). Calorimetric biosensor
for the detection and determination of enantiomeric excesses in aqueous and organic
phases. Biosensors and Bioelectronics, 8(3–4), 205–208.
12. Choi, J. M., Han, J. W., Choi, S. J., & Choi, Y. K. (2010). Analytical modeling of a
nanogap-embedded FET for application as a biosensor. IEEE Transactions on Electron
Devices, 57(12), 3477–3484.
166 Advanced VLSI Design and Testability Issues
13. Chakraborty, A., & Sarkar, A. (2017). Analytical modeling and sensitivity analysis of
dielectric-modulated junctionless gate stack surrounding gate MOSFET (JLGSSRG)
for application as biosensor. Journal of Computational Electronics, 16(3), 556–567.
14. Im, H., Huang, X. J., Gu, B., & Choi, Y. K. (2007). A dielectric-modulated field-effect
transistor for biosensing. Nature Nanotechnology, 2(7), 430–434.
15. Kimura, J., Ito, N., Kuriyama, T., Kikuchi, M., Arai, T., Negishi, N., & Tomita, Y.
(1989). A novel blood glucose monitoring method an ISFET biosensor applied to trans-
cutaneous effusion fluid. Journal of the Electrochemical Society, 136(6), 1744–1747.
16. Palan, B., Santos, F. V., Karam, J. M., Courtois, B., & Husak, M. (1999). New ISFET
sensor interface circuit for biomedical applications. Sensors and Actuators B: Chemical,
57(1–3), 63–68.
17. Bergveld, P. (1986). The development and application of FET-based biosensors.
Biosensors, 2(1), 15–33.
18. Ahn, J. H., Choi, S. J., Han, J. W., Park, T. J., Lee, S. Y., & Choi, Y. K. (2010). Double-
gate nanowire field effect transistor for a biosensor. Nano letters, 10(8), 2934–2938.
19. Zhang, G. J., & Ning, Y. (2012). Silicon nanowire biosensor and its applications in dis-
ease diagnostics: a review. Analytica Chimica Acta, 749, 1–15.
20. Iijima, S. (1991). Helical microtubules of graphitic carbon. Nature, 354(6348), 56.
21. Yang, N., Chen, X., Ren, T., Zhang, P., & Yang, D. (2015). Carbon nanotube based
biosensors. Sensors and Actuators B: Chemical, 207, 690–715.
22. Gupta, S., Murthy, C. N., & Prabha, C. R. (2018). Recent advances in carbon nanotube
based electrochemical biosensors. International Journal of Biological Macromolecules,
108, 687–703.
23. Tran, T. T., & Mulchandani, A. (2016). Carbon nanotubes and graphene nano field-
effect transistor-based biosensors. TrAC Trends in Analytical Chemistry, 79, 222–232.
24. Gu. B., Park, T. J., Ahn, J. H., Huang, X. J., Lee, S. Y., & Choi, Y. K. (2009). Nanogap
field‐effect transistor biosensors for electrical detection of avian influenza. Small, 5(21),
2407–2412.
25. Baba, T. (1992). Proposal for surface tunnel transistors. Japanese Journal of Applied
Physics, 31(4B), L455.
26. Sarkar, D., & Banerjee, K. (2012). Proposal for tunnel-field-effect-transistor as ultra-
sensitive and label-free biosensors. Applied Physics Letters, 100(14), 143108.
27. Narang, R., Saxena, M., & Gupta, M. (2015). Comparative analysis of dielectric-
modulated FET and TFET-based biosensor. IEEE Transactions on Nanotechnology,
14(3), 427–435.
28. Xu, H. F., Dai, Y. H., Gui Guan, B., & Zhang, Y. F. (2017). Two-dimensional analytical
model for asymmetric dual-gate tunnel FETs. Japanese Journal of Applied Physics,
56(1), 014301.
29. Dwivedi, P., & Kranti, A. (2016). Applicability of transconductance-to current ratio
(gm/Ids) as a sensing metric for tunnel FET biosensors. IEEE Sensors Journal, 17(4),
1030–1036.
30. Lee, C. W., Afzalian, A., Akhavan, N. D., Yan, R., Ferain, I., & Colinge, J. P. (2009).
Junctionless multigate field-effect transistor. Applied Physics Letters, 94(5), 053511.
31. Ajay, Narang, R., Saxena, M., & Gupta, M. (2015). Investigation of dielectric modu-
lated (DM) double gate (DG) junctionless MOSFETs for application as a biosensors.
Superlattices and Microstructures, 85, 557–572.
32. Wagaj, S. C., Patil, S., & Chavan, Y. V. (2018). Performance analysis of shielded chan-
nel double-gate junctionless and junction MOS transistor. International Journal of
Electronics Letters, 6(2), 192–203.
33. Tripathi, S. L., Mishra, R., & Mishra, R. A. (2012). Multi-gate MOSFET structures
with high-k dielectric materials. Journal of Electron Devices, 16, 1388–1394.
Analytical Design of FET-Based Biosensors 167
34. Zhang, X. Y., Hsu, C. H., Cho, Y. S., Lien, S. Y., Zhu, W. Z., Chen, S. Y., ... & Huang,
S. X. (2017). Simulation and Fabrication of HfO2 Thin Films Passivating Si from a
Numerical Computer and Remote Plasma ALD. Applied Sciences, 7(12), 1244.
35. Ahn, J. H., Choi, S. J., Im, M., Kim, S., Kim, C. H., Kim, J. Y.,& Choi, Y. K. (2017).
Charge and dielectric effects of biomolecules on electrical characteristics of nanowire
FET biosensors. Applied Physics Letters, 111(11), 113701.
36. Cuervo, A., Dans, P. D., Carrascosa, J. L., Orozco, M., Gomila, G., & Fumagalli,
L. (2014). Direct measurement of the dielectric polarization properties of DNA.
Proceedings of the National Academy of Sciences, 111(35), E3624–E3630.
37. Narang, R., Saxena, M., Gupta, R. S., & Gupta, M. (2011). Dielectric modulated tun-
nel field-effect transistor—A biomolecule sensor. IEEE Electron Device Letters, 33(2),
266–268.
38. Tiwari, P. K., Dubey, S., Singh, M., & Jit, S. (2010). A two-dimensional analytical
model for threshold voltage of short-channel triple-material double-gate metal-oxide-
semiconductor field-effect transistors. Journal of Applied Physics, 108(7), 074508.
39. Kumar, M. J., & Chaudhry, A. (2004). Two-dimensional analytical modeling of fully
depleted DMG SOI MOSFET and evidence for diminished SCEs. IEEE Transactions
on Electron Devices, 51(4), 569–574.
40. Goel, E., Kumar, S., Singh, K., Singh, B., Kumar, M., & Jit, S. (2016). 2-D analytical
modeling of threshold voltage for graded-channel dual-material double-gate MOSFETs.
IEEE Transactions on Electron Devices, 63(3), 966–973.
41. Singh, K. N., & Dutta, P. K. (2019, March). Comparative analysis of underlapped sili-
con on insulator and underlapped silicon on nothing dielectric and charge modulated
FET based biosensors. In 2019 Devices for Integrated Circuit (DevIC), Kalyani, India
(pp. 231–235). IEEE Conference Proceedings.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
9 Low-Power FET-
Based Biosensors
Prasantha R. Mudimela and Rekha Chaudhary
Lovely Professional University
CONTENTS
9.1 Introduction .................................................................................................. 169
9.2 Principle of Operation .................................................................................. 170
9.3 Silicon Nanowire Biosensor ......................................................................... 173
9.4 Organic Field-Effect Transistor .................................................................... 174
9.5 Classification and Advances in Bio-FETs .................................................... 176
9.6 ImmunoFET ................................................................................................. 179
9.7 Cell-Based Bio-FET ..................................................................................... 180
9.8 Conclusions................................................................................................... 183
References .............................................................................................................. 184
9.1 INTRODUCTION
A biosensor, defined by the International Union of Pure and Applied Chemistry
(IUPAC), is a device that uses specific biochemical reactions mediated by isolated
enzymes, immune systems, tissues, organelles, or whole cells to detect chemical
compounds by using electrical, thermal, or optical signals [1]. FET (field-effect
transistor)-based biosensor (Bio-FET) is an electrically and chemically insulat-
ing layer that separates the analyte solution from the semiconducting device. The
first Bio-FET developed by Piet Bergveld was the ion-sensitive field-effect transis-
tor (ISFET) used for electrochemical and biological applications in 1970. The Bio-
FET is a field-effect transistor (metal oxide semiconductor field-effect transistor
[MOSFET] based) that is gated by variations in the surface potential induced by
molecules binding. When charged molecules, such as biomolecules, bind to the FET
gate, which is usually a dielectric material, they can change the charge distribution
of the underlying semiconductor material resulting in a change in conductance of the
FET channel. As shown in Figure 9.1, Bio-FET consists of two main compartments:
(i) the biological recognition element and (ii) the FET. Its construction is basically
centered on the ISFET, a type of MOSFET where the metal gate is substituted by a
membrane, solution of electrolyte, and reference electrode.
In other words, biosensors are defined as an operative tool that is used to study
the biomolecular interactions such as DNA hybridization, antibody–antigen interac-
tions, protein–protein interactions, receptor–ligand binding, DNA–protein binding,
169
170 Advanced VLSI Design and Testability Issues
and other types of interaction [2–6]. A basic biosensor structure consists of three
segments: biorecognition element, a transducer, and a signal-processing unit. The
biorecognition element senses the biological interactions; the transducer converts the
response into electrical signal; and the signal-processing unit processes and ampli-
fies the signals. Clark and Lyon reported the very first biochemical-based biosen-
sors in 1962 [7]. After that, a lot of different biosensors have been reported. As
the biosensor’s development has emerged, an interesting sensing approach has been
proposed, i.e., Bio-FETs. Due the fast development in the solid-state technologies,
Bio-FETs have come into the picture and have drawn the attention of researchers
in sensing mechanism. Bio-FETs are most suitable candidates for applications that
demand very high sensitivity, very fast response time, and mass production with low
cost.
has same structure as MOSFET, but in ISFET, the metal gate is replaced
by dielectric layer, electrolyte, and reference electrode. The dielectric layer
acts as a sensing membrane. The schematic diagram of ISFET is shown in
Figure 9.3. The capacitance of dielectric defines the sensitivity of sensor.
The dangling bonding sites available on dielectric surface determine the
charge concentration on the surface. The use of high-k dielectric has been
done to study the sensing mechanism [12, 13]. The high-k material improves
sensitivity and prevents drifting. The voltage is applied at reference elec-
trode. The potential at reference electrode will break the electrolyte/pH into
ions (H+ or OH−). The dangling bonds or hydroxyl groups present at dielec-
tric will attract the ions due to which charge is formed at dielectric surface
and current conduction occurs in the channel. The hydroxyl groups can
donate a proton or accept a proton. The schematic of site-binding model of
SiO2 is shown in Figure 9.4. The surface hydrolysis bond of Si–OH differs
in various solutions because of different pH values. Nernst limit is defined
as the maximum sensitivity of ISFET, i.e., 59.2 mV/pH.
Dual-Gate ISFET (DG-ISFET): DG-ISFET consists of have a back gate
on the other side with respect to active channel. The literature shows that
the sensitivity of DG-ISFET can go beyond Nernst limit [14–18]. The
Low-Power FET-Based Biosensors 173
shows the accumulation of positive charges, the last state shows the accumulation of
negative charges. The sensitivity of nanowire FET sensor is high because of channel
confinement effect [39]. If the nanowire diameter decreases in nanoscale, the S/V
ratio increases rapidly. The sensitivity of device is dominated by high S/V ratio.
Fabrication Process: Two different methods are used for the fabrication of
SiNW biosensor, i.e., bottom-up and top-down. In bottom-up method [29],
chemical vapor deposition was used to grow SiNW on substrate. But in top-
down method [40], SiNW is formed by lithography process. For nanopat-
terning and reactive ion etching, e-beam lithography is used to create a
three-dimensional wire structure. In the created wire structure, light dop-
ing is done in the sensing region, and heavy doping is done in the source/
drain contact. Top-down method is mostly used due to its compatibility
with standard CMOS process.
Applications: This CMOS SiNW has been used to detect hepatitis B virus
DNA [41], cardiac troponin I [42], and NT-proBNP [43] in clinical sample
successfully.
organic semiconductor layer is used as a receptor in sensing as shown in Figure 9.7. The
receptor layer has its own electronics and mechanical properties. The material selec-
tion categorizes it into p-type or n-type. OFET works in accumulation mode [44]. The
channel current is controlled by gate voltage. The mobility of OFET (10 −1 to 10 −2 cm2/
Vs) is lesser when compared with the mobility of silicon. But the organic semiconduc-
tor is vastly well suited with flexible substrate. OFET has been used for various sensing
applications such as chemical sensing, biological sensing, and humidity sensing. On
the basis of structure of the device, OFET can be categorized as follows:
TABLE 9.1
Major Benchmarks in Bio-FET Device
Year Concept Given/Structure Proposed
1970 ISFET
1976 Bio-FET
1980 EnFET
1980 ImmunoFET
1981 Coupling of cells with MOS
1991 CPFET
1997 Beetle/chip Bio-FET
1997 DNA hybridization
9.6 IMMUNOFET
The enhanced proclivity to utilize the high specificity and sensitivity of biomol-
ecules and living biological systems for sensor function has been the major force for
the increased use of the biosensor in numerous real-life situations. The collaboration
of antibody–antigen results in the ability to recognize the biological molecules. The
antibody is an intricate biomolecule containing hundreds of individual amino acids.
Structure of these amino acids is well organized. The antigen could be a macromol-
ecule structure which is generated in response with the antibody. This generation
of antigen against antibody structure happens due to a defensive mechanism of the
180 Advanced VLSI Design and Testability Issues
organism. Proteins with molecular weights larger than 5000 Daltons are normally
immunogenic that are susceptible to being recognized. The minute change in the
chemical modification of the molecular structure of the antigen leads to a tremen-
dous reduction in its affinity toward the original antibody. So immune sensors are
useful for measuring the wellness of the human immune system that subsequently
leads to a superb diagnostic tool in the field of medicine.
The increased possibility for monitoring the direct relationship between the antibody
and antigen leads to the pursuits toward the modeling of biosensors, as the actual rec-
ognition takes place at the surface of the sensor in the layer of antibodies. The concept
of the direct immunizing ISFET was developed by Schenck in 1978. In immunoFET,
the gate is modified by immobilizing the antibodies or antigens mostly in the structured
membrane as shown in Figure 9.12. The formation of the antibody and antigen pattern on
the gate of the ISFET results in the detectable change in the charge distribution. This con-
sequently leads to the direct modulation of the direct current of the ISFET. The unique
possibility of the direct detection of immune logical reactions has been realized through
the application of the immunoFET. Under the ideal conditions, the immunoFET is at
least theoretically capable to measure the concentration of immune molecules with a
very low detection limit. The ideal conditions required here are a truly capacitive inter-
face that leads to immobilization of immunological binding sites, complete coverage of
the antibody, highly charged antigen, and comparatively low ionic strength.
The measure problem observed in the immunoFET is to transmit or transduce
the recognizable molecular action between the antibody and antigen into a signal
that can be measured. It has been observed that potential charge distribution plays a
significant role in the transfer of immunological signals to the ISFET.
FIGURE 9.12 Schematic structure of an immunoFET with immobilized antibody (Ab) mol-
ecules. Ag, antigen molecules.
Low-Power FET-Based Biosensors 181
gate insulator of FET. Figure 9.13 shows the side view of a cell -transistor. Cells
are actually the living microstructures that consist of chemicals, ions, enzymes,
different kinds of proteins, and organic molecules in high concentration. Cells
are capable of processing different received signals via activation in a paral-
lel manner of various signal pathways. Although there are some problems with
the life span of cells, the motivation behind developing a Bio-FET is that living
components can directly respond to received information [80, 81]. Such informa-
tion along with some additional information is very useful in clinical diagno-
sis and many other fields. Using cell-based Bio-FETs, one can easily study the
effects of different compounds, pollutants, toxic materials, and so on on cellular
metabolism.
Working: Two different methods are used to find out the state of cells. The
first method uses the energy of cell metabolism. The ISFET-based chemi-
cal sensors are used to measure the variation in extracellular pH, ion con-
centration, and so on. In cell-based Bio-FETs, Si3N4 or Al2O3 is used as
a gate insulator for extracellular acidification measurements [82–84]. In
the measurement of extracellular acidification, the working of cell-based
Bio-FET is explained as follows: In a chamber having an array of cell-
based Bio-FETs, measurements of rate of acidification are done generally.
A proton amount of 108 protons/s is produced by a single cell in steady-
state conditions. This quantity can be increased to high value if external
receptor stimulation is done. But that also depends upon type of cell and
the receptor used. When the pump gets stopped, the protons that are gener-
ated get accumulated and the extracellular medium gets acidified. The rate
of change of extracellular acidification is detected by the ISFET. The high
sensitivity can be achieved if the volume of chamber is small. But the con-
finement of produced protons to small volume–sized chamber is still chal-
lenging. Different signaling pathways are parallelly activated. The integral
of acidification rate is not enough to get the information about the effects
of different parameters on cells. So a method has been reported in which
both acidification and respiration in a cell has been done by a cell-based
Bio-FET. In this, a noble metal electrode that performs reduction of oxy-
gen molecules has been used [85]. The ISFET can be integrated with other
microsensors on a single chip.
182 Advanced VLSI Design and Testability Issues
Other method involves the features of different cells such as muscle cells
or cell network. It involves the potential measurement of extracellular and
intracellular voltages. To record the extracellular voltage of muscle tissues,
ISFET-based sensor was developed in 1970. Such a sensor was used for neu-
rophysiological measurements [18]. The point contact model is the common
method that has been used to demonstrate the signal transfer from cells to FET
structure and further explains the characteristics of output signal [81]. Some
other methods are also reported [86]. But all the methods do not include the
effects of ion-sensitive property or pH, but these can play an important role
in changing the interfacial potential over the gate surface. Another impor-
tant parameter is “adhesion.” A strong adhesion between neuron and trans-
ducer surface is desired. So the specific resistance in the adhesion region was
measured [87]. The advancement in the area depends on getting the positive
results, repeatability, and making a disposable device structure.
Applications: Cell-based Bio-FETs are used for the detection of cell metabo-
lism and extracellular potential measurement [88, 89].
Beetle/Chip FET: The idea of bioelectronic sensor was proposed by Rechnitz
in 1986 in which the complete sensory organ of an insect is taken [90].
A beetle/chip sensor is a kind of biosensor in which an insect antenna is
directly coupled to a FET [75]. This approach was investigated and then
optimized [91–93]. In this method, when the antenna detects a particular
odor, then a voltage is generated in the antenna. Because of the potential
developed, the drain current is changed in the transistor. The insect antenna
is made up of different parts such as small hairs called sensilla. The sensilla
contains neurons in which the recognition of odor is processed. Because of
the sensed odor, a potential is developed in the insect antenna. Thus, drain
current flows in FET [94–96].
A bioelectronic interface is set up between insect antenna and FET so
that the signal generated by the antenna can be measured. There are two
methods to couple the antenna and FET [93]. In the first method, the whole
beetle is fixed in the cell, and antenna tip is dipped on the electrolyte. So
in this way, the antenna is in direct contact with insulator of FET through
electrolyte as shown in Figure 9.14. In between the head and neck, the refer-
ence electrode like platinum wire is placed.
In the second method, the antenna is separated from the insect and then
placed on the electrolyte as shown in Figure 9.15. When the odor is sensed by
the antenna, the air current over antenna depolarizes the electrolyte, which
in turn changes the conductance of FET between source and drain. The
change in conductance changes the drain current in the FET. Practically, a
beetle/chip-based FET was developed in which antenna of Colorado potato
beetle was taken [94]. The more intense the odor, the stronger the signal
was obtained. The changes in drain current define the sensor signal.
Application: Beetle/chip FET is used to find out the right time for pesticide
application by the detection of plant damage in potato field [65]; smoldering
fires such as burning coal, paper, and wood can be detected by means of
monitoring fire-specific odors [95–98].
Low-Power FET-Based Biosensors 183
9.8 CONCLUSIONS
In this chapter, the operating principle and application of low-power FET-based bio-
sensor is explained. The working of the biosensor is discussed in detail. In future,
the Bio-FETs can be developed in areas such as follows: (i) Bio-FETs use biological
receptor to bind with particular drug, which leads to change in protein structure; (ii)
an array of ISFET can be used to analyze the nucleic acids and can be used in various
applications such as clinical diagnosis, environment monitoring, and food industry;
the DNA hybridization can be done using ISFET structure; (iii) the formation of large
neural network for drug detection using conductive polymer field-effect transistors
(CPFET) structure; (iv) the bioanalytical microsystems integrate the Bio-FETs with
184 Advanced VLSI Design and Testability Issues
TABLE 9.2
Types of Bio-FETs
Type of Bio-FET Application
EnFET EnFETs can be used in applications such as determination of
glucose in blood serum, urine, urea, and so on
GenFET Biorecognition of DNA and RNA
ImmunoFET Detection of antibody–antigen binding, recognition of
immunological molecules
Cell based Bio-FET Detection of cell metabolism and extracellular potential
measurement
Beetle/chip FET To find out the right time for pesticide application by the
detection of plant damage in potato field, smoldering fires
such as burning coal, paper, and wood can be detected by
means of monitoring fire-specific odors
some other kinds of sensors; such products are cost-effective; and (v) antennas of
insect are very useful when combined with the FET structure. It can be used for appli-
cations such as finding out the leakage of gas in a laboratory or smoke. Although a lot
of research has been done in low-power-based Bio-FETs, still there is a huge scope for
different applications that are still untouched. Remarkable results are obtained with
FET-based miniaturized structures when combined with sensor.
REFERENCES
1. PAC, IUPAC Recommendations 64, 143 (1992).
2. F. Patolsky, G. Zheng, O. Hayden, M. Lakadamyali, X. Zhuang, and C. Lieber, Proc.
Natl. Acad. Sci. USA 101, 14017 (2004).
3. D. Kim, Y. Jeong, H. Park, J. Shin, P. Choi, J. Lee, and G. Lim, Biosens. Bioelectron.
20, 69 (2004).
4. E. Stern, J. F. Klemic, D. A. Routenberg, P. N. Wyrembak, D. B. Turner-Evans,
A. D. Hamilton, D. A. LaVan, T. M. Fahmy, and M. A. Reed, Nature 445, 519 (2007).
5. E. Stern, E. Steenblock, M. Reed, and T. Fahmy, Nano Lett. 8, 3310 (2008).
6. S. D. Caras, D. Petelenz, and J. Janata, Anal. Chem. 57, 1920 (1985).
7. L. C. Clark Jr. and C. Lyons, Ann. N. Y. Acad. Sci. 102, 29 (1962).
8. A. Sedra and K. Smith, Microelectronic Circuits, Oxford University Press, USA (2004).
9. G. Zheng, F. Patolsky, Y. Cui, W. U Wang, and C. M. Lieber, Nat. Biotechnol. 23, 1294
(2005).
10. A. J. Bard and L. R. Faulkner, Electrochemical Methods, Fundamentals and
Applications, 2nd ed., John Wiley & Sons, Inc., New York (2001).
11. P. Bergveld, IEEE Trans. Biomed. Eng., 17, 70 (1970).
12. S. Chen, J. G. Bomer, E. T. Carlen, and A. van den Berg, Nano Lett., 11, 2334 (2011).
13. I.-K. Lee, M. Jeun, H.-J. Jang, W.-J. Chob, and K. H. Lee, Nanoscale, 7, 16789 (2015).
14. O. Knopfmacher, A. Tarasov, W. Fu, M. Wipf, B. Niesen, M. Calame, and C.
Schonenb
¨ erger, Nano Lett., 10, 2268 (2010).
Low-Power FET-Based Biosensors 185
41. C.-W. Huang, H.-T. Hsueh, Y.-J. Huang, H.-H. Liao, H.-H. Tsai, Y.-Z. Juang, T.-H. Lin,
S.-S. Lu, and C.-T. Lin, Sens. & Actu. B, 181, 867 (2013).
42. P.-W. Yen, C.-W. Huang, Y.-J. Huang, M.-C. Chen, H.-H. Liao, S.-S. Lu, and C.-T. Lin,
Biosens. Bioelectron., 61, 112 (2014).
43. J.-K. Lee, I.-S. Wang, C.-H. Huang, Y.-F. Chen, N.-T. Huang, and C.-T. Lin, Sensors, 17,
2733 (2017).
44. L. Torsi, M. Magliulo, K. Manoli, and G. Palazzo, Chem. Soc. Rev., 42, 8612 (2013).
45. S. H. Jeong, J. Y. Lee, B. Lim, J. Lee, Y. Y. Noh, Dyes Pigm., 140 244–249 (2017).
46. D. Khim, G. Ryu, W. Park, H. Kim, M. Lee, and Y. Noh, Adv. Mater., 28 2752–2759
(2016).
47. S. S. Prasad and D. R. Nair, Int. J. Sci. Res., 6 1258–1261 (2017).
48. D. K. Nair and K. S. Kumar, Development of gamma radiation dosimeter using copper
phthalocyanine & zinc phthalocyanine based OFET, IEEE International Conference
on Circuits and Systems (ICCS), 17662327, (2017).
49. B. Cai, S. Wang, Y. Ning, L. Huang, Z. Zhang, and G.-J. Zhang, ACS Nano, 8, 2632
(2014).
50. Z. Wang and Y. Jia, Carbon, 130, 758 (2018).
51. X. Dong, Y. Shi, W. Huang, P. Chen, and L.-J. Li, Adv. Mater., 22, 1649 (2010).
52. B. Cai, L. Huang, H. Zhang, Z. Sun, Z. Zhang, and G.-J. Zhang, Biosens. Bioelectron.,
74, 329 (2015).
53. T.-Y. Chen, P. T. K. Loan, C.-L. Hsu, Y.-H. Lee, J. T.-W. Wang, K.-H. Wei, C.-T. Lin,
and L.-J. Li, Biosens. Bioelectron., 41, 103 (2013).
54. D. J. Kim, I. Y. Sohn, J. H. Jung, O. J. Yoon, N. E. Lee, and J. S. Park, Biosens.
Bioelectron., 41, 621 (2013).
55. D.-J. Kim, H.-C. Park, Il Y. Sohn, J.-H. Jung, O. J. Yoon, J.-S. Park, M.-Y. Yoon, and
N.-E. Lee, Small, 9, 3352 (2013).
56. S. Mao, G. Lu, K. Yu, Z. Bo, and J. Chen, Adv. Mater., 22, 3521 (2010).
57. Y. X. Huang, X. C. Dong, Y. M. Shi, C. M. Li, L. J. Li, and P. Chen, Nanoscale, 2, 1485
(2010).
58. P. K. Ang, A. Li, M. Jaiswal, Y. Wang, H. W. Hou, J. T. L. Thong, C. T. Lim, and K. P.
Loh, Nano Lett., 11, 5240 (2011).
59. S. Caras and J. Janata, Anal. Chem., 1980, 52, (1935–1937).
60. D. R. Thevenot, K. Toth, R. A. Durst and G. S. Wilson, Biosens. Bioelectron., 16,
121–131 (2001).
61. E. A. H. Hall, Biosensors, Open University Press, Milton Keynes, 1990.
62. R. Ulber and T. Scheper, Disposable Bioprocessing Systems, in Enzyme and Microbial
Biosensors, ed. A. Mulchandani and K. R. Rogers, Humana Press, Totowa, NJ, 1998,
35–50.
63. K. Wan, J. M. Chovelon, N. Jaffrezic–Renault and A. P. Soldatkin, Sens. Actuators, B,
58, 399–408 (1999).
64. J. Hu, Sens. Mater., 8, 477–484 (1996).
65. M. J. Schöning, and A. Poghossian, Analyst, 127(9), 1137–1151 (2002).
66. V. Volotovsky, A. P. Soldatkin, A. A. Shul’ga, V. K. Rossokhaty, V. I. Strikha and A. V.
El’skaya, Anal. Chim. Acta, 322, 77–81 (1996).
67. D. V. Gorchkov, A. P. Soldatkin, S. Poyard, N. Jaffrezic-Renault and C. Martelet,
Mater. Sci. Eng., C 5, 23–28 (1997).
68. A. P. Soldatkin, D. V. Gorchkov, C. Martelet and N. Jaffrezic–Renault, Mater. Sci.
Eng., C 5, 35–40 (1997).
69. B. H. van der Schoot, H. Voorthuyzen and P. Bergveld, Sens. Actuators, B, 1, 546–549
(1990).
70. H. I. Seo, C. S. Kim, B. K. Sohn, T. Yeow, M. T. Son and M. Haskard, Sens. Actuators,
B, 40, 1–5 (1997).
Low-Power FET-Based Biosensors 187
71. K. Y. Park, S. B. Choi, M. Lee, B. K. Sohn and S. Y. Choi, Sens. Actuators, B, 83,
90–97 (2002).
72. V. Volotovsky and N. Kim, Sens. Actuators, B, 49, 253–257 (1998).
73. A. B. Kharitonov, M. Zayats, A. Lichtenstein, E. Katz and I. Willner, Sens. Actuators,
B, 70, 222–231 (2000).
74. E. Tobias-Katona and M. Pecs, Sens. Actuators, B, 28, 17–20 (1995).
75. S. V. Dzyadevich, Y. I. Korpan, V. M. Arkhipova, M. Yu. Alesina, C. Martelet, A. V.
El’skaya and A. P. Soldatkin, Biosens. Bioelectron., 14, 283–287 (1999).
76. A. Poghossian, Sens. Actuators, B, 44, 361–364 (1997).
77. D. G. Pijanowska and W. Torbicz, Sens. Actuators, B, 44, 370–376 (1997).
78. J. P. Cloarec, N. Deligianis, J. R. Martin, I. Lawrence, E. Souteyrand, C. Polychronakos
and M. F. Lawrence, Biosens. Bioelectron., 17, 405–412 (2002).
79. A. B. Kharitonov, J. Wasserman, E. Katz and I. Willmer, J. Phys. Chem. B, 105, 4205–
4213 (2001).
80. L. Bousse, Sens. Actuators, B, 34, 270–275 (1996).
81. A. Offenhäusser and W. Knoll, Trends Biotechnol., 19, 62–66 (2001).
82. A. Fanigliulo, P. Accossato, M. Adami, M. Lanzi, S. Martinoia, M. Grattarola and C.
Nicolini, Sens. Actuators, B, 32, 41–48 (1996).
83. M. Lehmann, W. Baumann, M. Brischwein, R. Ehret, M. Kraus, A. Schwinde, M.
Bitzenhofer, I. Freund and B. Wolf, Biosens. Bioelectron., 15, 117–124 (2000).
84. S. Martinoia, N. Rosso, M. Grattarola, L. Lorenzelli, B. Margesin and M. Zen, Biosens.
Bioelectron., 16, 1043–1050 (2001).
85. M. Lehmann, W. Baumann, M. Brischwein, H. J. Gahle, I. Freund, R. Ehret, S.
Drechsler, H. Palzer, M. Kleintges, U. Sieben and B. Wolf, Biosens. Bioelectron., 16,
195–203 (2001).
86. C. Sprössler, M. Denyer, S. Britland, W. Knoll and A. Offenhäusser, Phys. Rev. E, 60,
2171–2176 (1999).
87. V. Kiessling, B. Müller and P. Fromherz, Langmuir, 16, 3517–3521 (2000).
88. B. Wolf, M. Brischwein, W. Baumann, R. Ehret and M. Kraus, Biosens. Bioelectron.,
13, 501–509 (1998).
89. W. H. Baumann, M. Lehmann, A. Schwinde, R. Ehret, M. Brischwein and B. Wolf,
Sens. Actuators, B, 55, 77–89. (1999).
90. S. Belli and G. Rechnitz, Anal. Lett., 19, 403–405 (1986).
91. M. J. Schöning, S. Schütz, P. Schroth, B. Weissbecker, A. Steffen, P. Kordos, H. E.
Hummel and H. Lüth, Sens. Actuators, B, 47, 235–238 (1998).
92. P. Schroth, M. J. Schöning, S. Schütz, Ü. Malkoc, A. Steffen, M. Marso, H. E. Hummel,
P. Kordos and H. Lüth, Electrochim. Acta, 44, 3821–3826 (1999).
93. P. Schroth, M. J. Schöning, P. Kordos, H. Lüth, S. Schütz, B. Weissbecker and H. E.
Hummel, Biosens. Bioelectron., 14, 303–308 (1999).
94. M. J. Schöning, P. Schroth and S. Schütz, Electroanalysis, 12, 645–652 (2000).
95. P. Schroth, H. Lüth, H. E. Hummel, S. Schütz and M. J. Schöning, Electrochim. Acta,
47, 293–297 (2001).
96. M. J. Huotari, Sens. Actuators, B, 71, 212–222 (2000).
97. M. J. Schöning, P. Schroth and S. Schütz, Electroanalysis, 12, 645–652 (2000).
98. P. Schroth, M. J. Schöning, H. Lüth, B. Weissbecker, H. E. Hummel and S. Schütz,
Sens. Actuators, B, 78, 1–5 (2001).
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
10 Nanowire Array–
Based Gate-All-
Around MOSFET
for Next-Generation
Memory Devices
Krutideepa Bhol
VIT-AP University
Biswajit Jena
Koneru Lakshmaiah Education Foundation
Umakanta Nanda
VIT-AP University
CONTENTS
10.1 Introduction ................................................................................................ 189
10.2 Brief Review on Sentaurus TCAD ............................................................. 193
10.2.1 Sentaurus TCAD Codes ............................................................... 194
10.3 Device Design and Simulation ................................................................... 195
10.4 Results and Discussions .............................................................................. 195
10.5 Conclusion .................................................................................................. 200
References .............................................................................................................. 200
10.1 INTRODUCTION
Planar transistors have been considered as the heart of integrated circuits for more
than a few decades; during this period, the size of the individual transistors has regu-
larly decreased. As the size decreases, planar transistors progressively suffer from
the abominable short-channel effect (SCE), especially “off-state” leakage current,
which increases the idle power required by the device.
Several gates on multiple surface surround the channel in case of a multigate
device. It thus provides a better electrical control over the channel, allowing more
effective suppression of “off-state” leakage current. Multiple gates also allow
189
190 Advanced VLSI Design and Testability Issues
enhanced current in the “on” state, also known as drive current. Among different
multigate devices, cylindrical surrounding-gate MOSFET (metal oxide semicon-
ductor field-effect transistor) provides better gate control over the channel. Also
saturated device dimensions produce high gate leakage current and several SCEs
such as threshold voltage roll-off, drain-induced barrier lowering, corner effect,
and so on. [1,2]. However, many device geometries have already been developed
and investigated in detail to improve the performance. Cylindrical gate-all-around
(GAA) MOSFET provides superior gate controllability compared with single-gate
and other multigate structures [3]. It is considered as one of the better models that
has the capability to overcome the physical scaling limit of conventional CMOS
(complementary metal oxide semiconductor) technology. The GAA MOSFET in the
fully depleted regime shows improved robustness against SCEs and also reduces the
threshold voltage and subthreshold swing (SS) [4]. The SCEs in the small-dimension
devices result in high off-state leakage current (Ioff ) and high SS. So Ion/Ioff ratio of the
device decreases significantly, which affects the circuit speed and dissipation power.
However, the impact of SCEs can be reduced by scaling the gate oxide thickness and
increasing the channel doping concentration for channel length beyond 100 nm [5].
The objective of device scaling is to shrink transistor dimension so that more
number of transistors can be placed on a chip. Typically, the scaling factor “k” is √2,
so that the area is reduced to one-half and the number per area increases by a factor
of 2. Integration of billions of transistors on a single chip has been feasible due to
the possibility to pattern every minor feature on silicon through optical lithography.
As optical lithography enters the subwavelength regime, light diffraction and inter-
ference from subwavelength pattern feature causes image disorder. Therefore, pat-
terning becomes difficult without adopting resolution enhancement techniques. As
the traditional planar MOSFET design slowed down due to scaling problems, new
material or gate engineering was necessary in order to match current requirement of
semiconductor industries. Evolution of MOSFET technology started and achieved
milestone both in device design and performance. Gate engineering technology
changed traditional MOSFET to GAA MOSFET as shown in Figure 10.3. In fully
depleted double-gate (DG) and cylindrical surrounding-gate MOSFET, the SCEs
are governed by the electrostatic potential, and in these structures, the channel is
confined by the gate metal. So gate provides superior controllability to reduce these
types of effects (Figure 10.1).
Just after the invention of transistor, researchers realize that by reducing the size
of the transistor, its performance can be improved. After that, MOSFET scaling
played an important role in device architecture. Basically, there are various meth-
ods of scaling, but constant electric field scaling proposed by Dennard et al. played
a vital role [7]. Since the early 1990s, semiconductor manufacturing companies
and researchers have come to the front in order to define the future semiconductor
industry. This initiative gave birth to ITRS (International Technology Roadmaps for
Semiconductors) [6]. Every year, ITRS publishes an issue that helps to set a bench-
mark for future device improvement. The CMOS technology is one of the frontline
developments achieved by semiconductor industries from a long time. And the basic
component behind the CMOS technology is the MOS transistor. At the end of the
1990s, a new device structure was developed which was efficient to increase the
Nanowire Array–Based Gate-All-Around MOSFET 191
FIGURE 10.1 Evolution of the number of transistors per chip predicted [6]. DRAM,
dynamic random access memory.
performance of conventional MOSFET. And the new device was named as silicon-
on-insulator (SOI) in which the transistors are made in a thin silicon layer sitting on
the top of a SiO2 layer. SOI device has the ability to reduce parasitic capacitances.
But these devices were not that much efficient to reduce SCEs. SCE occurs when
the dimensions of a transistor shrunk down and the closeness between source and
drain hampers the controllability of the gate to control the potential distribution and
flow of current through the channel. So for future device purpose, it was impossible
to reduce the length anymore. Research to develop this device came to an end with
multigate device structure.
With the continuous development for improved drain current and SCE, a classical
planar device is transformed into multigate device structures such as double-/triple-/
quadruple-gate devices. The first article on the double-gate MOSFET was given by
T. Sekigawa and Y. Hayashi in 1984 [8]. The triple-gate MOSFET is an improved
model of the double-gate MOSFET where the gates are surrounded to three sides
of the channel. Π-gate and Ω-gates are some developed models of triple gate for
improved electrostatic characteristics. The structure that theoretically offers the
best possible control of the channel region with the help of gate is surrounding-gate
MOSFET. Surrounding-gate device with square or circular cross section was given
by N. Singh et al. in 2006 [9]. But the analytical model for cylindrical surrounding-
gate MOSFET was developed by several researchers [10–26]. Some GAA structures
are given in Figures 10.2–10.4.
Continuous downscaling in device dimension to fulfill the demands of present
semiconductor technology by introducing high-speed and low-power devices is
pushing the CMOS technology to the ultimate nanoscale dimension. Out of differ-
ent nanoscale structures, cylindrical GAA MOSFET is acting as a workhorse for
the scaled device family. Due to cylindrical geometry, the device has the ability to
preserve the device ability toward SCEs. Apart from this, a tight capacitive coupling
with higher electrostatic controllability is obtained in case of a cylindrical GAA
192 Advanced VLSI Design and Testability Issues
FIGURE 10.3 GAA structure with cubical gate and oxide but cylindrical channel. GAA,
gate-all-around.
…………………………..
(sde:set-default-material "Metal/Oxide/Silicon")
(sdegeo:create-cylinder (position 0 0 0.002) (position 0 0
0.088) 0.023 "Metal/Oxide/Silicon" "Gate1/Dielectric/Channel")
…………………………..
For multiple nanowires, which is called nanowire array, different silicon nanow-
ires should be taken for different coordinates. Similarly, all other regions associated
with MOSFET are defined with proper coordinates and accurate materials. In order
to provide the contact points for the device, the device code can be written as follows:
I. The grid (or geometry) of the device contains a description of the various
regions, that is, boundaries, material types, and the locations of any electri-
cal contacts.
II. The data fields contain the properties of the device, such as the doping pro-
files, in the form of data associated with the discrete nodes.
File
{
* input files:
* output files:
Nanowire Array–Based Gate-All-Around MOSFET 195
}
Electrode {
{ Name="source/drain/gate" Voltage=0.0/0.0/0.0 }
}
Physics {
}
Plot {
}
FIGURE 10.6 Three-dimensional device structure of the proposed model with nanowire
array.
FIGURE 10.7 (a) and (b) The electrostatic potential distribution of the proposed nanowire
array–based MOSFET model along the channel for two different gate biases. MOSFET,
metal oxide semiconductor field-effect transistor.
With a lower gate bias of 0.2 V, the corresponding potential distribution in each
nanowire is shown in Figure 10.7a. From the figure, it can be observed that the poten-
tial is lower at the source side and comparatively higher at the drain side. The source
side has a potential nearly equal to the built-in potential, whereas for drain side, the
built-in potential is added up with the applied drain voltage. This is the reason that
we always get asymmetric potential curve. However, for a gate bias of nearly 0.7 V,
the potential is high throughout the channel, but the starting potential at the source
side and end potential at the drain side remain similar to the lower gate bias. Again
Nanowire Array–Based Gate-All-Around MOSFET 197
it can be observed that, for each and every nanowire, the potential distribution is not
uniform. This is because of the fact that the arrangement of nanowires to form the
array and number of nanowires plays a vital role in potential distribution. Apart from
this, due to the channel and drain junction with different drain bias, the potential
curve is affected a little. Each nanowire has a different potential profile and hence
a different potential minima. The average of these potential minima is used to cal-
culate overall threshold voltage of the device. It is also observed that the nanowires
placed at the center have uniform potential curve with lower potential minima, as a
result of which their contribution in deciding threshold voltage of the device is more
compared with other peripheral nanowires.
The analysis is further extended to examine the effect of higher gate bias on the
potential distribution as well as electric field distribution of individual nanowires
as shown in Figure 10.8a and b, respectively. The nanowires are individually num-
bered from 1 to 13 as shown in the figure. Except pillar number 1, 2, 3, and 4, almost
all the pillars exhibit similar potential distribution throughout the pillar. But there
is a little deviation in the peripheral nanowires due to their positioning. However,
the electric field distribution for the silicon pillars is different for each pillar due to
the different potential distribution and depletion lines due to drain bias. The results
can be well observed from Figure 10.8b. The analysis is further extended toward
the switching ratio calculation of the device by illustrating the on current and off
current.
Figure 10.9 shows the gate voltage versus drain current analysis of the proposed
device for different drain current. An inset is also given for proper analysis of on and
off current in logarithmic form. From the figure, the dependency of drain voltage on
drain current is observed. With the increase in drain voltage, the corresponding on
current is increasing by keeping the off current constant. An acceptable range of
switching ratio is observed for the proposed device, and the dependency of drain
current on drain voltage clearly indicates toward the stability of the device in high
electric field.
FIGURE 10.8 (a) Higher gate bias on the potential distribution. (b) Electric field distribu-
tion of individual nanowires.
198 Advanced VLSI Design and Testability Issues
FIGURE 10.9 Gate voltage versus drain current analysis of the proposed device for different
drain voltages.
FIGURE 10.10 Transfer characteristic analysis of the proposed device in linear and loga-
rithm scales.
Figure 10.10 shows the transfer characteristics curve of the proposed model for a
gate voltage of 1 V with a drain voltage of 0.5 V. The gate voltage versus drain current
curve is represented in both linear and logarithmic forms in order to ensure effective
IOn/IOff ratio. As this ratio plays a vital role during memory design, so with an improved
switching ratio, this device seems to be a suitable candidate in logic applications. The
threshold voltage of the proposed device is extracted by taking the derivative of trans-
conductance as shown in Figure 10.11. A lower value of threshold voltage for the pro-
posed model indicates toward better memory device with fast switching also.
The variation in switching ratio with different threshold voltage is shown in
Figure 10.12. From the figure, it can be observed that the device with higher threshold
Nanowire Array–Based Gate-All-Around MOSFET 199
FIGURE 10.11 Threshold voltage extraction from the transconductance of the proposed
device.
FIGURE 10.12 Threshold voltage versus IOn /IOff ratio of the proposed device.
voltage is having lower switching ratio and vice versa. However, the proposed device
exhibits a threshold voltage of 0.1 V with a switching ratio about 40 mA, which indi-
cates toward next-generation memory device.
The variation in drain current with respect to the drain voltage for different gate
voltage is illustrated in Figure 10.13. From the figure, the improved output character-
istic of the proposed model is observed, and the variation with different gate voltage
is also observed. Higher drain current is observed due to higher electron mobility in
200 Advanced VLSI Design and Testability Issues
FIGURE 10.13 Gate voltage versus drain current analysis of the proposed device for
different drain voltage.
the silicon pillars. Array of nanowires that act as channels are responsible for high
mobility of the electrons.
10.5 CONCLUSION
Introducing multiple nanowires in place of single silicon pillar increased the drain
current with reduced threshold voltage. The potential distribution in individual
nanowire results in improved electron mobility by enhancing the electrostatic con-
trollability. However, arrangement of nanowires to form the channel plays a vital role
during device performance metrics calculation. Distribution of nanowires in circular
form exhibits better performance compared with other geometrical arrangements of
the nanowires. With improved characteristics, the proposed device can be a fruitful
member in GAA family for next-generation memory–based devices.
REFERENCES
1. K. Suzuki, T. Tanaka, Y. Tosaka, H. Horie, and Y. Arimoto, “Scaling theory for double-
gate SOI MOSFETs”, IEEE Tran. Electron Dev., vol. 40, (1993) 2326.
2. A. Chaudhry and M. J. Kumar, “Controlling short-channel effects in deep submicron
SOI MOSFETs for improved reliability: A review”, IEEE Trans. Dev. Mater. Rel.,
vol. 4, (2004) 99.
3. A. A Orouji, M. Jagadesh Kumar, “A new symmetrical double gate nano scale MOSFET
with asymmetrical side gates for electrically induced source/drain”, Microelectron.
Eng., vol. 83, (2006) 409.
4. T. K. Chiang, “A scaling theory for fully-depleted, surrounding-gate MOSFET’s:
Including effective conducting path effect”, Microelectron. Eng. vol. 77, (2005) 175.
5. J. P. Colinge, “An SOI voltage-controlled bipolar-MOS device”, IEEE Trans. Electron
Devices, vol. 34 (1987) 845.
Nanowire Array–Based Gate-All-Around MOSFET 201
6. B. Davari, H. Dennard, AND G. G. Shahidi, “CMOS Scaling for high performance and
low power-the next ten years”, Proc. IEEE, vol. 83, (1995) 595.
7. B. Jena, S. Dash, S. Routray, G. P Mishra, “Inner-gate-engineered GAA MOSFET to
enhance the electrostatic integrity”, NANO: Brief Rep. Rev., vol. 14, (2019), 1950128.
8. T. Sekigawa and Y. Hayashi, “Calculated threshold-voltage characteristics of an XMOS
transistor having an additional bottom gate”, Solid-State Electron., vol. 27, (1984) 827.
9. N. Singh, A. Agarwal, L. K. Bera, T. Y. Liow, R. Yang, S. C. Rustagi, C. H. Tung,
R. Kumar, G. Q. Lo, N. Balasubramanian, D. L. Kwong, “High- performance fully
depleted silicon nanowire (diameter<5 nm) gate-all- around CMOS devices”, IEEE
Electron Dev. Lett., vol. 27,(2006) 383.
10. C. P. Auth, J. D. Plummer, “Scaling theory for cylindrical fully-depleted surrounding
gate MOSFETs”, IEEE Electron Dev. Lett., vol. 18, (1997) 74.
11. S. L. Jang, S. S Liu, “An analytical surrounding gate MOSFET model”, Solid State
Electron., vol. 42, (1998) 721.
12. A. Kranti, S. Haldar, R.S. Gupta, “Analytical model for threshold voltage and I-V char-
acteristics of fully depleted short channel cylindrical/surrounding gate MOSFET”,
Microelectron Eng., vol. 56, (2001) 241.
13. S. Hyun Oh, D. Monroe, J. M. Hergenrother, “Analytic description of short channel
effects in fully depleted double gate and cylindrical surrounding gate MOSFET”, IEEE
Electron Dev. Lett, vol. 21, (2000) 445.
14. D. Jimenez et al., “Continuous analytic I-V model for surrounding gate MOSFETs”,
IEEE Electron Dev. Lett., vol. 25, (2004) 571.
15. T. K. Chiang, “A new two dimensional analytical model for threshold voltage in
undoped surrounding gate MOSFETs”, Solid-State Int. Circ. Technol., (2006) 1234.
16. L. Zhang, J. He, J. Zhang, J. Feng, “A continuous yet Explicit In 2006, a carrier based
core model for the long channel undoped cylindrical surrounding gate MOSFETs”,
NSTI-Nanotech., vol. 3, (2008) 590.
17. J. He, Y. Tao, F. Liu, J. Feng, “Analytical channel potential solution to the undoped sur-
rounding gate MOSFETs”, Solid State Electron., vol. 51, (2007) 802.
18. H. Kaur, S. Kabra, S. Bindra, S. Haldar, R. S. Gupta, “Impact of graded channel (GC)
design in fully depletedcylindrical/surrounding gate MOSFET (FD CGT/SGT) forim-
proved short channel immunity and hot carrier reliability”, Solid-State Electron.,
vol. 51, (2007) 398.
19. F. Liu, J. He, L. Zhang, J. Zhang, J. Hu, C. Ma, and M. Chan, “A charge-based model for
long-channel cylindrical surrounding-gate MOSFETs from intrinsic channel to heavily
doped body”, IEEE Trans. Electron Dev., vol. 55, (2008) 2187.
20. T. K. Chiang, “A new compact subthreshold behavior model for dual-material sur-
rounding gate (DMSG) MOSFETs”, Solid-State Electron., vol. 53, (2009) 490.
21. C. Li, Y. Zhuang, R. Han, “Cylindrical surrounding-gate MOSFETs with electrically
induced source/drain extension”, Microelectron. J., vol. 42, (2011) 341.
22. V. M. Srivastava, K. S. Yadav, G. Singh, “Design and performance analysis of cylindri-
cal surrounding double-gate MOSFET for RF switch”, Microelectron. J., vol. 42, (2011)
1124.
23. P. Ghosh, S. Haldar, R. S. Gupta, M. Gupta, “An analytical drain current model for dual
material engineered cylindrical/surrounded gate MOSFET”, Microelectron. J., vol. 43,
(2012) 17.
24. B. Jena, S. Dash, K. P Pradhan, S. K Mohapatra, P. K Sahu, G. P Mishra, “Performance
analysis of undoped cylindrical gate all around (GAA) MOSFET at subthreshold
regime”, Adv. IN Nat. Sci.: Nanosci. Nanotechnol., vol. 6, (2015), 035010.
25. B. Jena, S. Dash, G. P Mishra, “Effect of underlap length variation on DC/RF perfor-
mance of dual material cylindrical MOS”, Int. J. Num. Model: Electron Network, Dev.
Fields, vol. 30, (2016), e2175.
202 Advanced VLSI Design and Testability Issues
26. L. Zhang, C. Ma, J. He, X. Lin, M. Chan, “Analytical solution of subthreshold channel
potential of gate underlap cylindrical gate-all-around MOSFET”, Solid-State Electron.,
vol. 54, (2010) 806.
27. Guide, Sentaurus Device User, and N. Version. “Synopsys TCAD Sentaurus.” San Jose,
CA, USA (2017).
11 Design of 7T SRAM Cell
Using FinFET Technology
T. Santosh Kumar and Suman Lata Tripathi
Lovely Professional University
CONTENTS
11.1 Introduction ................................................................................................ 203
11.2 SRAM Cell Architectures Based on CMOS Technology .......................... 204
11.2.1 6T SRAM Cell .............................................................................. 204
11.2.2 7T SRAM Cell .............................................................................. 205
11.2.3 8T SRAM Cell .............................................................................. 205
11.2.4 10T SRAM Cell ............................................................................ 206
11.2.5 12T SRAM Cell ............................................................................ 206
11.3 SRAM Cell Architectures Based on FinFET Technology ......................... 206
11.3.1 Proposed Design 7T SRAM Cell ................................................. 207
11.3.2 Operations of 7T SRAM .............................................................. 208
11.4 Result Analysis ........................................................................................... 209
11.5 Different Types of Leakage Current in SRAM .......................................... 210
11.5.1 Subthreshold Leakage Current ..................................................... 210
11.5.2 Gate Leakage ................................................................................ 210
11.5.3 Junction Tunneling Leakage ......................................................... 211
11.5.4 Different Leakage Reduction Techniques .................................... 211
11.5.4.1 Self-controllable Voltage Level Technique .................. 211
11.5.4.2 Lower Self-controllable Voltage Level ........................ 211
11.5.4.3 Upper Self-controllable Voltage Level ......................... 211
11.6 Conclusion .................................................................................................. 213
References .............................................................................................................. 213
11.1 INTRODUCTION
In the current chip technology, the capability of silicon-on-chip (SOC) memory is
quickly developing to increment worldwide execution. As a more prominent reserve
memory is required, SRAM (static random access memory) assumes a continuously
critical job in current chip frameworks, compact gadgets, and cell phones [1]. To
achieve more noteworthy speed, SRAM-dependent reserve memories and SOC
memories are typically utilized [2]. The gadget scaling, SRAM cell configuration
has a few difficulties such as power utilization issues, steadiness, and area. Several
kinds of research have been carried out for low-power stable SRAM cell opera-
tion using many circuit design techniques 6T, 7T, 9T, and 11T bit cell using CMOS
203
204 Advanced VLSI Design and Testability Issues
WL
VDD
M4 M8
M3 M7
WL
BL
Q QBAR
M9 BLB
M10
M2 M6
M1 M5
precharging the RBL and activating RWL. If 1 is stored at node Q, then M6 turns
ON and makes a low resistance path for the flow of cell current through RBL to
ground, which can be sensed by the sense amplifier [21].
WL
VDD
S
M11
M4 M8
M3 M7
BL WL
Q QBAR
M9 BLB
M10
M2 M6
M1 M5
S`
M12
C. Hold:
In this state, the SRAM maintains the data until the power supply is ON.
Depending on the data stored, i.e., if data contained in the cell is “1,” then Q
will be at VDD and Qb will be at “0 V” and vice versa if the stored data is “0.”
A. Power:
The complete power consumed by SRAM cell is the sum of power
drawn from the source, sources accustomed to charge and discharge the
BLs, and sources used for WL and RL. Power consumed by explicit supply
is considered a product of the average source current and source voltage.
The total power mainly has three components such as (1) static or leakage
power, (2) dynamic or charging–discharging power, and (3) short circuit
power consumption during switching (Figure 11.7).
Power Consumption(nW)
40
35
30
25
20
15
10
5
0
Write 0 Write 1 Read 0 Read 1
7T CMOS SRAM cell 6T SRAM cell on 20nm FinFET 7T SRAM Cell 18nm FinFET
FIGURE 11.7 Power comparisons of CMOS SRAM and FinFET SRAM cells.
210 Advanced VLSI Design and Testability Issues
11.6 CONCLUSION
Different SRAM cell design topologies of 7T, 8T, 10T, and 12T performances have
been compared in terms of leakage power. The speed of SRAM cell is increased, as
the process technologies continue to advance, but devices will be more vulnerable
to gaps, which damage the static noise margin of SRAM cells. An 18-nm FinFET-
based SRAM cell circuit is implemented to reduce the leakage power. It is found
that by applying SVL technique, there was a reduction of leakage current in 18-nm
FinFET-based SRAM cell design.
REFERENCES
1. N. Verma, A. P. Chandrakasan 2008. A 256 kb 65 nm 8T subthreshold SRAM employ-
ing sense-amplifier redundancy. IEEE Journal of Solid-State Circuits 43(1): 141–149.
2. T.-H. Kim, J. Liu, J. Keane, C. H. Kim 2008. A 0.2 V, 480 kb subthreshold SRAM
with 1 k cells per bitline for ultra-low-voltage computing. IEEE Journal of Solid-State
Circuits 43: 518–529.
3. B. H. Calhoun, A. P. Chandrakasan 2007. A 256-kb 65-nm sub-threshold SRAM design
for ultra-low-voltage operation. IEEE Journal Solid-State Circuits, 42: 680–688.
4. M.-H. Tu, J.-Y. Lin, M.-C. Tsai, S.-J. Jou 2010. Single-ended sub-threshold SRAM with
asymmetrical write/read-assist. IEEE Transactions on Circuits and Systems, 57(12):
3039–3047.
5. M. Sharifkhani, M. Sachdev 2009. An energy efficient 40 Kb SRAM module with
extended read/write noise margin in 0.13 um CMOS. IEEE Journal Solid-State
Circuits, 44: 620–630.
6. S. K. Jain, P. Agarwal 2006. A low leakage SNM free SRAM cell design in deep sub
micron CMOS technology. 19th International Conference on VLSI Design (VLSID’06),
IEEE: 1063–9667.
7. I. J. Chang, J. J. Kim, S. P. Park, K. Roy 2008. A 32 kb 10 T subthreshold SRAM array
with bit-interleaving and differential read scheme in 90 nm CMOS. ISSCC Digest of
Technical Papers, 3–7: 388–622.
8. P. Athe, S. Dasgupta 2009. A comparative study of 6T, 8T and 9T Decanano SRAM
cell. IEEE Symposium on Industrial Electronics & Applications: 889–894.
9. A. A. Mazreah, M. R. Sahebi, M. T. Manzuri, S. J. Hosseini 2008. A novel zero-
aware four-transistor SRAM cell for high density and low power cache application.
International Conference on Advanced Computer Theory and Engineering, IEEE:
571–575.
10. S. Lin, Y.-B. Kim and F. Lombardi 2008. A 32 nm SRAM design for low power and
high stability. 51st Midwest Symposium on Circuits and Systems, IEEE: 422–425.
214 Advanced VLSI Design and Testability Issues
11. P. Upadhyay, R. Mehra, N. Thakur 2010. Low power design of an SRAM cell for por-
table devices. International Conference on Computer and Communication Technology
(ICCCT), IEEE: 255–259.
12. S. S. Rathod, A. K. Saxena, S. Dasgupta 2010. A proposed DG-FinFET based SRAM
cell design with RadHard capabilities. Microelectronics Reliability, 50(8): 1039–1190.
13. C. B. Kushwah, S.K. Vishvakarma, D. Dwivedi 2016. A 20 nm robust single-ended
boostless 7T FinFET sub-threshold SRAM cell under process–voltage–temperature
variation. Microelectronics Journal 51: 75–88.
14. Kumar T. S., Tripathi S. L. 2019. Implementation of CMOS SRAM cells in 7, 8, 10 and
12-transistor topologies and their performance comparison. International Journal of
Engineering and Advanced Technology, 8(2S2).
15. G. Chen, D. Sylvester, D. Blaauw, T. Mudge 2010. Yield-driven near-threshold SRAM
design. IEEE Transaction on VLSI System, 18: 1590–1598.
16. U. R. Karpuzcu, A. Sinkar, N.S. Kim, J. Torrellas 2013. Energy smart: Toward energy
efficient many cores for near-threshold computing, Proceedings of IEEE HPCA:
542–553.
17. R. G. Dreslinski, M. Wieckowski, D. Blaauw, D. Sylvester, T. Mudge 2010.
Nearthreshold computing: Reclaiming Moore’s law through energy efficient integrated
circuits. Proceedings of the IEEE, 98: 253–266.
18. B. H. Calhoun, D. Brooks 2010. Can subthreshold and near-threshold circuits go main-
stream?. IEEE Micro, 30: 80–85.
19. K. Takeda 2006. Low-Vdd static-noise-margin-free for read SRAM cell in high-speed
applications. Journal of Solid-State Circuits: 113–121.
20. L. Chang, D. M. Fried, J. Hergenrother, J. W. Sleight, R. H. Dennard, R. K. Montoye, L.
Sekaric, S. J. McNab, A. W. Topol, A. D. Adams, K. W. Guarini, and W. Haensch 2005.
Stable SRAM cell design for the 32 nm node and beyond. Proceedings of the IEEE on
Very Large Scale Integration (VLSI) Technology: 128–129.
21. C. Visweswariah 2003. Death, taxes and failing chips. Proceedings of the IEEE on
Design Automation Conference: 343–347.
22. Y. Taur, X. Liang, W. Wang, and H. Lu 2004. A continuous, analytic drain current model
for DG MOSFETs. IEEE Transactions on Electron Device Letters. 25(2): 107–109.
23. V. Sikarwar, S. Khandelwal and S. Akashe 2013. Analysis of leakage reduction tech-
niques in independent gate DG FinFET SRAM Cell. Chinese Journal of Engineering:
2013: 1–8. https://doi.org/10.1155/2013/738358
24. S. Birla, N. Kr. Shukla, R. K. Singh and M. Pattanaik 2010, June. Leakage current
reduction in 6T single cell SRAM at 90 nm technology. Proceedings of the IEEE
International Conference on Advances in Computer Engineering: 292–294.
25. C. Duari and S. Birla 2018, October. Leakage power improvement in SRAM Cell with
clamping diode using reverse body bias technique. Proceedings of the 2nd International
Conference on Data Engineering and Communication Technology, Advances in
Intelligent Systems and Computing: 828.
26. M. Manorama, S. Khandelwal, S. Akashe 2013, 12–14 April. Design of a FinFET based
inverter using MTCMOS and SVL leakage reduction technique, Students Conference
on Engineering and Systems (SCES).
27. S. Akashe, S. Sharma 2013. Leakage current reduction techniques for 7T SRAM Cell
in 45 nm technology. Wireless Personal Communication 71: 123–136.
28. K. Endo, S-I. O’inchi, Y. Ishikawa, E. Suzuki 2009. Independent-double-gate FinFET
SRAM for leakage current reduction. IEEE Electron Device letters 30(7): 757–759.
12 Performance Analysis
of AlGaN/GaN
Heterostructure Field-
Effect Transistor (HFET)
Yogesh Kumar Verma
Lovely Professional University
CONTENTS
12.1 Introduction ................................................................................................ 215
12.2 Model Description ...................................................................................... 216
12.3 Results and Discussions .............................................................................. 216
12.4 Conclusion .................................................................................................. 221
References .............................................................................................................. 221
12.1 INTRODUCTION
AlGaN and GaN materials are being exploited in the semiconductor electronics
market from the past decade. The materials are popularly used in heterojunc-
tion bipolar transistors and heterojunction field-effect transistors (HFETs) [1–9].
Because of the merits of FETs over bipolar transistors, the HFETs are still under
research. The difference between the electron affinities of both the materials
causes high inversion charge at the heterointerface. The AlGaAs/GaAs HFETs
were extensively a main area of research before the past decade. However, the
magnitude of 2-DEG density was limited only to 1016 m−2, which is 10 order of
magnitude lesser than the AlGaN/GaN HFET [2–19]. The formation of 2-DEG
density in AlGaAs/GaAs HFET depends on the doping concentration in the barrier
layer. There is AlN buffer layer to reduce the leakage current, thereby minimizing
the leakage issue in the device. The geometry of the device is also responsible for
enhancing the performance of the device. The spacing between source–gate and
gate–drain is deterministic to improve the performance of the device [20–34]. In
this chapter, the spacing between source–gate and gate–drain is varied, and its
effect on the current–voltage characteristics of the device is analyzed. This chapter
215
216 Advanced VLSI Design and Testability Issues
is divided into four sections: first section is introduction, second section is model
description, third section represents the results and discussions, and fourth section
represents the results.
FIGURE 12.2 Output characteristics: (a) VG = 0 V and (b) VG = −1 V with respect to different
source spacing.
218 Advanced VLSI Design and Testability Issues
FIGURE 12.3 Output characteristics of AlGaN/GaN HFET for 1-1, 1-2, 1-3, 1-4, 1-5, and
1-6 at gate voltage of (a) 0 V and (b) −1 V. HFET, heterostructure field-effect transistor.
Performance Analysis of AlGaN/GaN HFET 219
FIGURE 12.4 (a, b) Transfer characteristics of AlGaN/GaN HFET for 1-1, 2-1, 3-1, 4-1, 5-1,
and 6-1 at drain voltage of (a) 1 V and (b) 2 V. HFET, heterostructure field-effect transistor.
220 Advanced VLSI Design and Testability Issues
FIGURE 12.5 Transfer characteristics for 1-1, 1-2, 1-3, 1-4, 1-5, and 1-6 at (a) VD = 1 V and
(b) VD = 2 V.
Performance Analysis of AlGaN/GaN HFET 221
FIGURE 12.6 VTC curve for different values of load resistance. VTC, voltage transfer
characteristics.
12.4 CONCLUSION
In this chapter, the analysis of AlGaN/GaN HFET for different device geometries is
performed. The invertor circuit is also designed to utilize the present analysis. It has
been revealed that the drain current reduces significantly, when the spacing between
source and gate is reduced. However, the effects of reducing the spacing between
gate and drain are less significant on the current–voltage characteristics.
REFERENCES
1. S. Khandelwal, Y. S. Chauhan, and T. A. Fjeldly. Analytical modeling of surface-
potential and intrinsic charges in AlGaN/GaN HEMT devices. IEEE Transactions
on Electron Devices. 2012, vol. 59, No. 10, pp. 2856–2860. ISSN: 0018-9383. DOI:
10.1109/TED.2012.2209654.
2. Y. K. Verma, V. Mishra, P. K. Verma, and S. K. Gupta. Analytical modelling and
electrical characterisation of ZnO based HEMTs. International Journal of Electronics
(IJE, Taylor and Francis Group). 2019, November, vol. 106, No. 5, pp. 707–720. DOI:
10.1080/00207217.2018.1545931.
3. L. Wang, J. Liu, W. Zhou, Z. Xu, Y. Wu, and H. Tao. A novel method to dynamic
thermal impedance and channel temperature extraction of GaN HEMTs. International
Journal of Numerical Modeling: Electron Networks, Devices, and Fields. 2019, p. 2599.
ISSN: 1099-1204, DOI: 10.1002/jnm.2599.
4. N. Mendiratta, and S. L. Tripathi. A review on performance comparison of advanced
MOSFET structures below 45 nm technology node. Journal of Semiconductor, IOP
science. 2020, vol. 41, pp. 1–10.
5. Y. K. Verma, V. Mishra, and S. K. Gupta. Analog/RF and linearity distortion analysis
of MgZnO/CdZnO quadruple-gate field effect transistor (QG-FET). Silicon, Springer.
2020. DOI: 10.1007/s12633-020-00406-4.
6. S. L. Tripathi, R. Patel, and V. K. Agrawal. Low leakage pocket junction-less DGTFET
with bio sensing cavity region. Turkish Journal of Electrical Engineering and Computer
Sciences. 2019, vol. 27, No. 4, pp. 2466–2474.
222 Advanced VLSI Design and Testability Issues
Pramod K. Singh
Sharda University
CONTENTS
13.1 Introduction ................................................................................................ 225
13.2 Polymer-Based Composites ........................................................................ 226
13.3 Methods of Synthesis .................................................................................. 228
13.3.1 Solution Casting Method .............................................................. 228
13.3.2 Copolymerization ......................................................................... 228
13.3.3 Addition of Ceramic Fillers .......................................................... 228
13.3.4 Sol–Gel Process ............................................................................ 229
13.3.5 Plasticization ................................................................................. 229
13.3.6 Nanofillers..................................................................................... 230
13.4 Conclusion .................................................................................................. 230
References .............................................................................................................. 231
13.1 INTRODUCTION
From almost the past decade, polymer-based composites have gained a lot of attrac-
tion on account of its interesting properties for the scientific community. Variation in
properties may be in the form of enhanced electrical properties, mechanical proper-
ties, magnetic properties, or optical properties. On account of some excellent prop-
erties, these materials have been widely used in different applications in the field of
medical, food, or consumer goods, thus serving the society. In regard to a huge num-
ber of research articles, scientific reports, or journals, this chapter aims at narrating
various methods of synthesis of these polymer-based composites.
Composite materials are defined as materials that are made by combination of two
or more than two materials, leading to some enhanced properties. While synthesizing
225
226 Advanced VLSI Design and Testability Issues
a composite material, one must make sure that there is no chemical reaction among
components of a composite. All the components making a complete composite are
superficially placed without any chemical reaction between them. Nowadays, we have a
special class of composites known as nanocomposites. In nanocomposites, at least one
of the components of composite must have a nanoscale dimension [1,2]. Also for appli-
cation in FETs, we use a special class of composites known as polymer electrolyte [3].
a. Liquid electrolytes
b. Gel electrolytes
c. Solid electrolytes
Liquid electrolytes are ion-containing solutions that allow free movement of ions.
But they suffer major setback on account of their leakage problems. On the other
hand, gel electrolytes are in the form of a semisolid mass that show more stability and
equivalent conduction as in liquid electrolytes. In liquid and gel electrolytes, charge
carriers, i.e., ions conduct while moving in a polar solvent such as ethylene carbonate,
propylene carbonate, butyrolactone, and others or their mixtures [5–7]. Value of ionic
conductivity of these electrolytes is sufficient for their application in the commercial
lithium battery (for laptops or and cellular phones) working at ambient temperature.
Another drawback associated with these electrolytes is formation of an additional
insulating layer during their application in batteries. The decomposition products of
these electrolyte get deposited on the electrode, form insulating layer at electrode–
electrolyte interface, and hinder the movement of ions. This ultimately reduces the
life span of the product. Furthermore, it is very difficult to transport devices based on
liquid electrolyte due to its bulky nature. Usually, liquid electrolytes have low boiling
• Polymer electrolytes are very easy to process as compared with liquid elec-
trolytes. This ensures high flexibility in designing these polymers.
• They have adequate mechanical properties along with good thermal and
electrochemical stability. Also, they form acceptable electrode/electrolyte
interfacial contact.
• They are more safe as they lack the existence of any organic solvent that is
flammable.
For its use as a solid-state device for electrochemical applications, a polymer electro-
lyte composite must have the following properties:
a. Polymers that possess atoms or have special groups that are basically elec-
tron donors show a strong capability of making coordinate bonds with salts
(cations present in salt).
b. Host polymer must have low value of glass transition temperature.
c. Polymers that show flexibility and possess low values of cohesive energy
density have higher likelihood to interact with metal salts.
d. The metal salts that have low values of lattice energy have higher tendency
to form a polymer salt complex. Such salts generally have monovalent small
sized alkali ions and a large size anions (e.g., NH4I, NaI, KI, etc.).
228 Advanced VLSI Design and Testability Issues
13.3.2 COPOLYMERIZATION
In this method, one more polymer with a low value of Tg is added in addition to the
original host. This tends to produce a flexible polymer backbone and hence lowers
crystallinity and Tg of original host polymer and enhances conductivity (σ) (Druger
et al., 1971). PEO–PPO copolymer complex with alkali metal salts has also been
reported (Figure 13.3).
(PMMA) (West et al., 1985) or (ii) some inorganic/organic or ceramic fillers (Al2O3,
LiAlO2, SiO2, Li3N, Nascon, PMMA, etc.) during casting process are added to the
polymer solution to form composites [13–15], which have high conductivity also.
13.3.5 PLASTICIZATION
The addition of low-molecular-weight polar molecules called plasticizers improves
the segmental flexibility of original host polymer backbone chain and hence
13.3.6 NANOFILLERS
The existence of polymer nanocomposites can be dated a number of decades back.
The extensive research shows that polymer nanocomposites existing in nature have
nanofillers in the form of diatoms, black carbon, and silica, which had been intro-
duced as additives in polymer matrices. The augmented properties introduced by the
addition of nanofillers were not fully apprehended at that time due to lack of interest
and lack of available techniques. Extensive study on the outcome of incorporation of
nanofillers in polymer matrix started only after the first published research work on
polyamide doped with nanoclays by Park (2014) and Torsi et al. (2003) [17,18]. Both
the mentioned research publications used the term “hybrid” material instead of nano-
composite material. Swiftly, investigation in the field increased manifold and led to
the appearance of introductory use of the denomination “nanocomposites.” After the
aforementioned voyage, a large number of young researches initiated the work on var-
ious nanosized fillers. The emergence of this field was on account of continual devel-
opment in the practical development of thermoplastic as well as thermoset polymeric
materials, which has further led to development of new technologies. List of various
nanofillers being used has increased at an enormous rate from the past few years like
we started with nanoclays, nanooxides and are moving towards carbon nanotubes,
etc. With the increase in number of fillers, the matrix in which they are incorporated
has also increased manifold ranging from polymers and ceramics to glasses.
As per present scenario, the evolution of polymer nanocomposites is the most
active area and key aspect of progress of fields related to nanomaterial [19,20]. The
existing properties introduced by the incorporation of nanoparticles are large in
number and primarily focus on enhancing the electrical conduction and resist vari-
ous barrier properties to effects of temperature, presence of various gases and liquids
along with providing fire-retardant properties. Nanofillers significantly enhance or
adjust various properties of the base materials into which they are introduced, such
as optical property, electrical property, mechanical property, thermal property, or
fire-retardant properties. These properties of nanocomposite materials are signifi-
cantly affected by ratio of organic matrix to the nanofillers.
13.4 CONCLUSION
Organic FET is a three-terminal device consisting of drain, source, and gate.
Dielectric material of gate can be synthesized using a polymer-based composite that
assists in charge accumulation around gate on application of source voltage. These
polymer composites are synthesized by a number of methods such as wet chemical
method, ball milling, solution casting method, and so on. Out of all the methods,
solution casting method is widely known for synthesis of polymer-based composites
on account of ease of fabrication, low cost, and better output.
Synthesis of Polymer-Based Composites 231
REFERENCES
1. Cavallari, M.R.; Izquierdo, J.E.E.; Braga, G.S.; Dirani, E.A.T.; Pereira-da-Silva, M.A.;
Rodriguez, E.F.G.; Fonseca, F.J. Enhanced sensitivity of gas sensor based on poly(3-
hexylthiophene) thin-film transistors for disease diagnosis and environment monitor-
ing. Sensors 2015, 15, 9592–9609.
2. Klug, A.; Denk, M.; Bauer, T.; Sandholzer, M.; Scherf, U.; Slugovc, C.; List, E.J.W.
Organic field-effect transistor based sensors with sensitive gate dielectrics used for low-
concentration ammonia detection. Org. Electron. 2013, 14, 500–504.
3. Lu, C., Fu, Q.; Huang, S.; Liu, J. Polymer electrolyte-gated carbon nanotube field-effect
transistor. Nano Lett. 2004, 4, no. 4, 623–627.
4. Lienerth, P.; Fall, S.; Leveque, P.; Soysal, U.; Heiser, T. Improving the selectivity to
polar vapors of OFET-based sensors by using the transfer characteristics hysteresis
response. Sens. Actuator B Chem. 2016, 225, 90–95.
5. Sandberg, H.G.O.; Backlund, T.G.; Osterbacka, R.; Jussila, S.; Makela, T.; Stubb, H.
Applications of an all-polymer solution-processed high-performance, transistor. Synth.
Met. 2005, 155, 662–665.
6. Pacher, P.; Lex, A.; Proschek, V.; Etschmaier, H.; Tchernychova, E.; Sezen, M.; Scherf,
U.; Grogger, W.; Trimmel, G.; Slugovc, C.; et al. Chemical control of local doping in
organic thin-film transistors: From depletion to enhancement. Adv. Mater. 2008, 20,
3143–3148.
7. Ryu, G.S.; Park, K.H.; Park, W.T.; Kim, Y.H.; Noh, Y.Y. High-performance diketopyr-
rolopyrrole-based organic field-effect transistors for flexible gas sensors. Org. Electron.
2015, 23, 76–81.
8. Yang, Y.; Zhang, G.X.; Luo, H.W.; Yao, J.J.; Liu, Z.T.; Zhang, D.Q. Highly sensitive
thin-film field-effect transistor sensor for ammonia with the DPP-bithiophene conju-
gated polymer entailing thermally cleavable tert-butoxy groups in the side chains. ACS
Appl. Mater. Interfaces 2016, 8, 3635–3643.
9. Han, S.J.; Zhuang, X.M.; Shi, W.; Yang, X.; Li, L.; Yu, J.S. Poly(3-hexylthiophene)/
polystyrene (P3HT/PS) blends based organic field-effect transistor ammonia gas sen-
sor. Sens. Actuator B Chem. 2016, 225, 10–15.
10. Besar, K.; Yang, S.; Guo, X.G.; Huang, W.; Rule, A.M.; Breysse, P.N.; Kymissis, I.J.;
Katz, H.E. Printable ammonia sensor based on organic field effect transistor. Org.
Electron. 2014, 15, 3221–3230.
11. Chen, D.J.; Lei, S.; Chen, Y.Q. A single polyaniline nanofiber field effect transistor and
its gas sensing mechanisms. Sensors 2011, 11, 6509–6516.
12. Yu, S.H.; Cho, J.; Sim, K.M.; Ha, J.U.; Chung, D.S. Morphology-driven high-
performance polymer transistor-based ammonia gas sensor. ACS Appl. Mater. Interfaces
2016, 8, 6570–6576.
13. Wang, L.; Swensen, J.S. Dual-transduction-mode sensing approach for chemical detection.
Sens. Actuator B Chem. 2012, 174, 366–372.
14. Liao, F.; Yin, S.; Toney, M.F.; Subramanian, V. Physical discrimination of amine vapor
mixtures using polythiophene gas sensor arrays. Sens. Actuator B Chem. 2010, 150,
254–263.
15. Das, A.; Dost, R.; Richardson, T.; Grell, M.; Morrison, J.J.; Turner, M.L. A nitrogen
dioxide sensor based on an organic transistor constructed from amorphous semicon-
ducting polymers. Adv. Mater. 2007, 19, 4018–4023.
16. Cheon, K.H.; Cho, J.; Kim, Y.H.; Chung, D.S. Thin film transistor gas sensors incorpo-
rating high-mobility diketopyrrolopyrole-based polymeric semiconductor doped with
graphene oxide. ACS Appl. Mater. Interfaces 2015, 7, 14004–14010.
17. Park, C.E. A composite of a graphene oxide derivative as a novel sensing layer in an
organic field-effect transistor. J. Mater. Chem. C 2014, 2, 4539–4544.
232 Advanced VLSI Design and Testability Issues
18. Torsi, L.; Tanese, M.C.; Cioffi, N.; Gallazzi, M.C.; Sabbatini, L.; Zambonin, P.G.; Raos,
G.; Meille, S.V.; Giangregorio, M.M. Side-chain role in chemically sensing conducting
polymer field-effect transistors. J. Phys. Chem. B 2003, 107, 7589–7594.
19. Torsi, L.; Tafuri, A.; Cioffi, N.; Gallazzi, M.C.; Sassella, A.; Sabbatini, L.; Zambonin,
P.G. Regioregular polythiophene field-effect transistors employed as chemical sensors.
Sens. Actuator B Chem. 2003, 93, 257–262.
20. Lv, A.; Wang, M.; Wang, Y.; Bo, Z.; Chi, L. Investigation into the sensing process of
high-performance H2S sensors based on polymer transistors. Chem. Eur. J. 2016, 22,
3654–3659.
14 Power Efficiency
Analysis of Low-
Power Circuit Design
Techniques in 90-nm
CMOS Technology
Yelithoti Sravana Kumar, Tapaswini
Samant, and Swati Swayamsiddha
Kalinga Institute of Industrial Technology,
KIIT Deemed to be University
CONTENTS
14.1 Introduction ................................................................................................ 234
14.2 Existing Low-Power Techniques ................................................................ 235
14.2.1 Conventional Complementary Metal Oxide Semiconductor ........ 235
14.2.2 Pass-Transistor Logic Style ........................................................... 236
14.2.3 Differential Pass-Transistor Logic Style ....................................... 237
14.2.4 Transmission Gate Logic Style ..................................................... 237
14.2.5 Gate Diffusion Input Logic Style ................................................. 238
14.3 Proposed Low-Power Adiabatic Logic Techniques .................................... 239
14.3.1 Conventional Positive-Feedback Adiabatic Logic ........................ 240
14.3.2 Two-Phase Adiabatic Static Clocked Logic ................................. 241
14.4 Existing Design........................................................................................... 241
14.4.1 4×1 Multiplexer Using Conventional Complementary Metal
Oxide Semiconductor ................................................................... 241
14.4.2 4×1 Multiplexer Using Pass-Transistor Logic Style ..................... 241
14.4.3 4×1 Multiplexer Using Differential Pass-Transistor Logic Style .... 242
14.4.4 4×1 Multiplexer Using Transmission Gate Logic Style................ 242
14.4.5 4×1 Multiplexer Using Gate Diffusion Input Logic Style ............ 242
14.5 Proposed Design ......................................................................................... 243
14.5.1 4×1 Multiplexer Using Conventional Positive-Feedback
Adiabatic Logic ............................................................................ 243
14.5.2 4×1 Multiplexer Using Two-Phase Adiabatic Static
Clocked Logic ............................................................................... 245
233
234 Advanced VLSI Design and Testability Issues
14.1 INTRODUCTION
The growing demand for low-power very large-scale integration (VLSI) can be
addressed at various levels of design, including architecture, circuit, design, and pro-
cess technology levels. At the circuit design level, there is a significant energy sav-
ing potential by properly selecting the logical style to implement the combinational
circuit [1]. This is because all the important parameters that manage the switching
capacity, the transition activity, the power consumption, and the short-circuit current
are strongly influenced by the preferred logical style. Depending on the application,
different types of circuits are implemented, and various design methods are used. As
several aspects of performance are considered important, it is impossible to formu-
late universal optical rules of logical style.
This chapter analyzes 4-to-1 multiplexer using complementary metal oxide semi-
conductor (CMOS), transmission gate (TG), pass-transistor logic (PTL), dual pass-
transistor logic (DPTL) styles, and gate diffusion input [2]. These implementations
are compared based on the basis of transistor count, power dissipation, and delay [3].
A device that selects one of several analog or digital input signals and transfers
the selected input to a single line is called as multiplexer. A multiplexer of inputs has
n select lines, which are used to select which input line to send to the output; that
is why, it is also called a data selector. Multiplexer can also be used to implement
any combinational circuit [4]. So by simplifying the design of multiplexer, design of
many combinational circuits can be simplified. Figure 14.1 and Table 14.1 show the
block diagram and truth table for 4-to-1 multiplexer [5].
Power consumption is an important parameter in today’s VLSI technology. In the
adiabatic approach, the energy is stored in the charge capacitor instead of discharg-
ing to ground. The term “thermodynamics” means that there is no energy dissipation
[6]. Decades of CMOS technology has created low-power apparatus, and the power
consumption mainly occurs due to changing operation of the charge capacitor [7].
The two main types of dissipation are static and dynamic . Dynamic loss occurs due
to the operation of the charge capacitor. Internal leakage occurs in the device that
forms static electricity if circuit is not working, and dynamic power dissipation plays
TABLE 14.1
Truth Table of 4 × 1 Multiplexer
Selection Inputs Inputs Output
S1 S0 P Q R S Z
0 0 1 0 0 0 P
0 1 0 1 0 0 Q
1 0 0 0 1 0 R
1 1 0 0 0 1 S
a significant role in the circuit [8]. By decreasing the values of terminal capacitance
and reducing the voltage supply in CMOS logic circuits, the power dissipation can be
reduced [9]. Therefore, the effective model requires low power consumption. In this
way, the energy is recovered by sending it back to the power supply, and this reduces
overall power consumption. In this chapter, the performance of different adiabatic
styles is evaluated, and also the simulation results show that it is the best approach
in adiabatic logic [10].
FIGURE 14.2 CMOS logic schematic diagram. CMOS, complementary metal oxide
semiconductor.
236 Advanced VLSI Design and Testability Issues
CMOS circuits are constructed with the goal that each NMOS transistor requires
contribution from a voltage source or other PMOS transistors. Likewise, all NMOS
transistors require a contribution from the beginning additional NMOS transistor.
When a low gate voltage is applied to the arrangement of PMOS transistor, it creates
low resistance between source and drain, and when a high gate voltage is applied,
it creates high resistance between source and drain. Similarly, the NMOS transistor
arrangement makes high resistance between source and drain when a low gate volt-
age is applied, and when high gate voltage applied, it creates low resistance between
source and drain [11]. However, during the switching time, when gate voltage passes
from one state to another state, both MOSFETs (metal oxide semiconductor field-
effect transistors) work for a short time.
This induces a short peak in energy consumption, which is a serious problem at
high frequencies. CMOS complements all nMOSFETs with pMOSFET and con-
nects both gates and both drains to achieve current reduction [12]. A high gate volt-
age causes the nMOSFET to drive and the pMOSFET to become nonconductive,
but the opposite occurs when the gate voltage is low. This arrangement significantly
reduces energy consumption and heat generation [13].
When XOR is implemented using simple logic gate, it requires more transistors in
completely CMOS logic, but using PTL, it needs a less number of transistors.
A twin PMOS transistor branch has been added to the N-tree DPL to avoid the
problem of noise margin degradation in the CPL. This addition increases the input
capacity.
In any case, the symmetric operation procedure and dual transmission character-
istics compensate for the consequences of reduced speed and increased load. Full
swing operation limits the threshold voltage scale and improves circuit performance
at low supply voltages [16].
As a general rule, unlike conventional individual field effect transistors, TGs com-
posed of two FETS do not have the substrate terminal (mass) inside associated with
the source. In the TG logic, two transistors nMOSFET and pMOSFET are arranged
in parallel, but source and drain terminals of the two transistors are connected to one
another. Its gate terminals connected to one another through NOT gate (inverters) to
frame control terminals.
Unlike discrete FET transistors, to represent the TG, two varieties of symbols
are commonly used in this arrangement; the substrate terminal does not have the
connection with the source [18]. Since the terminal of the substrate is connected to
a specific supply voltage, the diode of the parasitic substrate (gate to the substrate)
always reverses the polarization and does not affect the signal flow. Subsequently,
the p-channel MOSFET substrate terminal is connected with the positive supply
voltage potential, and the n-channel MOSFET substrate terminal is connected with
the negative supply voltage potential.
1. The GDI cell contains three inputs and one output: G (common gate input
for NMOS and PMOS), P-type source (source input/PMOS drain), N-type
source (source input/NMOS drain), and D (common drain for NMOS and
PMOS).
2. Both NMOS and PMOS transistors are connected to N or P (respectively),
so they can be arbitrarily polarized instead of CMOS inverters.
Power Efficiency Analysis 239
Note that not all functions are probable with standard P-well CMOS forms; how-
ever, they can be effectively executed with two-well CMOS or silicon-on-insulator
(SOI) technology.
The different logical elements of the GDI cell for various input configurations
significant to output are conventional polar, and there is a short circuit between N
and P, resulting in static energy consumption, which is normal CMOS configuration.
There are inconveniences in implementation of OR, AND, and MUX. The impact
can be reduced by executing the design in floating majority SOI technology that a
full GDI library can implement.
The amount of power worthless throughout the discharge of the load capacitor is
given in Equation (14.1). CMOS assumes a significant job in decreasing control.
However, later, a dynamic approach was intended to weary the weakness of CMOS
logic. The adiabatic methodology used consumes less power and has low power
dissipation. By using this adiabatic technique, the wasted power within the CMOS
circuit is recovered [20]. There are two types of adiabatic logic square measure men-
tioned: partially adiabatic and fully adiabatic logic. The positive-feedback adiabatic
logic (PFAL) and two-phase adiabatic static clocked logic (2PASCL) come under the
partially adiabatic logic (Figure 14.6).
240 Advanced VLSI Design and Testability Issues
1. Evaluation stage
2. Hold stage
3. Recovery stage
4. Wait stage
Throughout the evaluation stage, the outputs are measured from a steady information
outstanding. Next during hold stage, the output is saved steady; in the recovery inter-
val, the energy is recovered, and finally in the wait stage, outputs are inserted for the
symmetry. PFAL is a dual-railing circuit that recognizes proportional commitments
about one another and gives enhanced outputs with divided essential recovery [21].
The general schematic of the PFAL structure is shown in Figure 14.7. It consists
of an adiabatic amplifier circuit and is created by the two PMOS and two NMOS
transistors; both output nodes are without any degradation in amplification level. The
pMOSFETs of the adiabatic electronic equipment and frame a TG.
FIGURE 14.9 Schematic design diagram for 4 × 1 multiplexer using conventional CMOS
logic. CMOS, complementary metal oxide semiconductor.
FIGURE 14.10 Schematic design diagram for 4 × 1 multiplexer using pass-transistor logic.
FIGURE 14.11 Schematic design diagram for 4 × 1 multiplexer using differential pass-tran-
sistor logic.
TABLE 14.2
Performance Comparison of Different Design Techniques with Respect to
the Number of Transistors, Power Analysis, and Speed
Power Analysis Speed
Number of
Name of the Logic Transistors 250 nm 90 nm 250 nm 90 nm
Conventional CMOS 26 10.511 × 10−8 8.56 × 10−8 21.892 ns 18.235 ns
Differential pass transistor 10 8.907 × 10−8 5.26 × 10−8 14.823 ns 10.549 ns
Pass transistor 6 8.657 × 10−8 5.02 × 10−8 16.548 ns 11.486 ns
Transmission gate 14 5.158 × 10−8 3.56 × 10−8 18.652 ns 13.586 ns
Gate diffusion input 6 4.163 × 10−8 3.12 × 10−8 19.548 ns 14.953 ns
Conventional PFAL 24 4.015 × 10−8 3.09 × 10−8 14.645 ns 10.561 ns
Proposed 2PASCL 26 3.374 × 10−8 2.26 × 10−8 12.562 ns 08.459 ns
CMOS, complementary metal oxide semiconductor; PFAL, positive-feedback adiabatic logic; 2PASCL,
two-phase adiabatic static clocked logic.
Power Efficiency Analysis 245
FIGURE 14.12 Schematic design diagram for 4 × 1 multiplexer using transmission gate
logic.
FIGURE 14.13 Schematic design diagram for 4 × 1 multiplexer using gate diffusion input
logic.
• The different multiplexer designs are enhanced on gate level through set-
ting an increasingly accurate aspect relation to accomplish a lower power
consumption.
• For the multiplexer, the power consumption is reduced by changing the sup-
ply voltages in the range of 0.1–5 V, so as to fit to the essential input dynamic
range. In this way, the power consumption improvement rate is 12%.
Power Efficiency Analysis 247
FIGURE 14.14 Schematic design diagram for 4 × 1 multiplexer using conventional positive-
feedback adiabatic logic.
14.7 CONCLUSION
In this chapter, low-power multiplexers are implemented using different logic designs.
From the simulation results, multiplexer design using 2PASCL gives low power con-
sumption compared with PFAL, and various low-power and high-speed techniques,
namely, TG logic, GDI logic, CMOS, DPTL, and PTL exist. The outcomes show that
when compared with other proposed structures, CMOS has more power consumption
248 Advanced VLSI Design and Testability Issues
FIGURE 14.15 Schematic design diagram for 4 × 1 multiplexer using two-phase adiabatic
static clocked logic.
and transistor count. The outcomes were simulated using Tanner EDA, and comparisons
are carried out with respect to various parameters such as power dissipation, delay, and
the total number of transistor. These points raise awareness about proposed techniques
over CMOS, which make them effective and suitable to be utilized in digital circuits.
REFERENCES
1. E. Chitra, N. Hemavathi, V. Ganesan, “Energy Efficient design of logic circuits using
adiabatic process”, International Journal of Engineering and Technology, Vol. 9, no. 6,
pp. 4504–4505, 2018.
Power Efficiency Analysis 249
20. S. Sayedsalehi, K. Navi, “A novel architecture for quantum-dot cellular automata mul-
tiplexer”, International Journal of Computer Science, Vol. 8, no. 6, pp. 55–60, 2011.
21. A. Agarwal, T. K. Gupta, A. K. Dadoria, “Ultra low power adiabatic logic using diode
connected DC biased PFAL logic”, Advances in Electrical and Electronic Engineering,
Vol. 55, no. 1, pp. 46–54, 2017.
15 Macromodeling
and Synthesis of
Analog Circuits
B. S. Patro and Sushanta Kumar Mandal
Kalinga Institute of Industrial Technology,
KIIT Deemed to be University
CONTENTS
15.1 Introduction ................................................................................................ 251
15.2 Parametric-Based Macromodeling ............................................................. 252
15.2.1 Symbolic Modeling ...................................................................... 253
15.2.2 Posynomial Templates/Geometric Programming ........................ 256
15.2.3 Model Order Reduction ................................................................ 256
15.3 Nonparametric Macromodeling ................................................................. 259
15.3.1 Artificial Neural Network ............................................................. 259
15.3.2 Support Vector Machine ............................................................... 261
15.3.3 Extreme Learning Machine .......................................................... 262
15.4 Conclusions ................................................................................................. 264
References .............................................................................................................. 264
15.1 INTRODUCTION
The analog, mixed-signal, radio frequency (RF), and digital circuits are the major
components of modern system-on-chips (SoCs). Demand for very high-end smart
devices such as smart phones with multitasking features has been increased to
improve the quality of life. Day by day the complexities of the SoCs are increas-
ing, as more functions are included and as there is an increase in the number of
transistors in a chip, which follows Moore’s law (Schaller 1997; Patro and Vandana
2016). Due to a rapid growth in SoC functions and competition for time-to-market
(TTM), understanding the features and designing an optimized analog circuit for a
specific task have become challenging jobs for the circuit designers. Existing design
and verification tools are not able to provide the feasible solution in desired limited
time (Chen 2009). One of such powerful tools is simulation program with integrated
circuit emphasis (SPICE)–based simulation modeling. It is a widely used simulation
tool, able to handle many types of analysis with very high accuracy (Najm 2010).
But this type of simulation method cannot handle the increasing complexities of the
circuits and consumes lot of time, which increases TTM.
251
252 Advanced VLSI Design and Testability Issues
the circuits and their mathematical analysis. So this modeling method is mostly pre-
ferred by the researchers who have sound notions about the circuits. In this technique,
human interference plays a greater role in formulating the models. Various techniques
are reported in literature. But in this section, three important methods are discussed
which are on the basis of their approach to formulate the model. They are (i) symbolic
modeling, (ii) posynomial templates, and (iii) model order reduction (MOR).
The end result is a set of algebraic equations for the performance parameters
which describe the behavior of the circuits. So this technique can be related to
numerical analysis.
Some good symbolic analyzers are reported in Kolka, Biolek, and Biolkova
(2008); McConaghy and Gielen (2009); Gielen and Sansen (2012); Rutenbar,
Gielen, and Roychowdhury (2007); and Shi and Tan (2000). In these analyzers,
the modified nodal analysis (MNA) method was principally used to formulate the
system of equations. In Vazzana, Grasso, and Pennisi (2017), the importance and
usefulness of MNA can be easily observed from its implemented tool using matrix
laboratory (MATLAB). Then, the recursive determinant expansion techniques
provide the performance of these generated symbolic models. Other methods
such as dead rows and V/I methods were also reported (Fakhfakh, Tlelo-Cuautle,
and Fernández 2012). These methods remove the redundancy and hence compli-
cation of the circuits to some extent. Some other methods such as nodal admit-
tance matrix (NAM) (Sánchez-López, Cante-Michcol, et al. 2013; Sánchez-López,
Ochoa-Montiel, et al. 2013) are also used for such type of reduction mechanism for
the circuits. Similarly, pathological equivalents have been used instead of active
devices for synthesis purposes (Tlelo-Cuautle, Sánchez-López, and Moro-Frías
2010; Saad and Soliman 2010). Binary decision diagram (Shi 2013; Zhang and
Shi 2011) was earlier used for logic synthesis and verification. But afterward, this
method is used more frequently for symbolic model generation and synthesis of
analog circuits. This method not only removes the data redundancy but also pro-
vides a mechanism for explicit enumeration, which decreases the complexity to a
certain extent.
There are also some works (Ferent and Doboli 2013; Shokouhifar and Jalali 2014,
2015; Shi, Hu, and Deng 2017; Vazzana, Grasso, and Pennisi 2017) that concentrate
mainly on automation of analog circuit synthesis. Ferent and Doboli (2013) provide
a method for comparison of analog circuits on the basis of symbolic performance
data generation. Shokouhifar and Jalali (2014, 2015) provide the automation meth-
ods for increasing the efficiency of symbolic models during the synthesis process by
the introduction of evolutionary algorithms. Shi, Hu, and Deng (2017) discussed the
advancement of automation for a symbolic macromodel generation for the ICs espe-
cially for the RLC type of elements. Similarly, a data structure–based determinant
decision diagram has shown some fruitful results in developing automatic symbolic
models [33].
Literature shows that this approach for generation of macromodels is most com-
plicated and time taking. Also, automated tools are limited to simple circuits. For
complex circuits, it uses various heuristic methods that somewhat simplifies, but
still the models generated are highly intricate. This led to the generation of com-
plex equations which further increases the memory storage and hence computing
time. Again, these models show unexpected behavior in several regions of the design
space. While reducing the redundancy using NAM, MNA, or other methods, some
of the key parameters will not be taken care of for the removed values during syn-
thesis time. These key features may be the vital parameters for the circuits during the
final fabrication time. But still, it is not able to satisfy the present growing demand
for fast design of circuits.
256 Advanced VLSI Design and Testability Issues
f ( x1 ,..., x n ) = ∑c x
k =1
α1 k
k 1 x 2α 2 k ... x nα nk , xi > 0 a (15.1)
number of small-size models. When these models are again grouped together, then
it will behave as similar to the original circuit. Like posynomial expressions, the
final reduced equations provide the models that are represented in the closed form of
mathematical equations. This technique has a very good advantage of modeling the
circuits whose behavior is linear in nature. It provides robust models that are helpful
for solving the mathematical equation in a faster manner. Pillage et al. (Shi, Hu, and
Deng 2017) used the explicit moments matching via Pade approximations (Antoulas
2005), and based on that, they provided the asymptotic waveform evaluation (AWE)
approach. But this approach becomes unstable and provides less accurate results
for transfer functions of higher order. Gallivan et al. (1994) provided the Pade via
Lanczos method that applies Krylov space projections (Freund 2000) for MOR in
circuit simulation. This helps to overcome the drawbacks of the AWE approach.
The Lanczos algorithm (Antoulas and Sorensen 2001) implicitly computes the first
moment or a transfer function. The passive reduced-order interconnect macromodel-
ing (PRIMA) (Odabasioglu, Celik, and Pileggi 1997) and the structure-preserving
reduced-order interconnect macromodeling (SPRIM) (Freund 2004) MOR meth-
ods help in reducing large-scale RLC networks having multiple terminals without
hampering the structure and passivity nature of the circuits. This is not observed
in Feldmann’s method (Feldmann 2004) who uses the SVDMOR technique. In this
method, the correlation between circuit responses at different terminals is used to
obtain the model, but the passiveness of the circuit is not preserved during the MOR
process. This limitation is encountered by Liu et al. (2006) and resolved with the
258 Advanced VLSI Design and Testability Issues
FIGURE 15.5 Generation of reduced-order equations using MOR technique. MOR, model
order reduction. ODE, ordinary differential equation; PDE, partial differential equation.
help of extended-SVDMOR method. It reduces the number of input and output ter-
minals separately. Figure 15.5 shows the procedure to generate reduced-order equa-
tions using the MOR technique.
MOR has the major limitation for nonlinear circuit. Chen (1999) proposed
a method for the reduction of the nonlinear circuit using quadratic Taylor series
expansion and Krylov space projections. The method is limited to weakly nonlin-
ear circuits and is inaccurate for strongly nonlinear circuits. Gu (2011) proposed
separate approaches for generating different reduced models for each circuit analy-
sis domain called the ManiMOR and the QLMOR. These methods are difficult to
use, scale poorly with the number of nonlinear terms, and still need some investi-
gations to be put in practice for complex nonlinear circuit models. Philips (2000)
presented a method for automatically extracting macromodels for weakly nonlinear
circuits based on the Volterra series and variational analysis theory (Rugh 1981).
Feng (2005) proposed a two-sided projection method to enhance this class of MOR
methods based on the variational analysis. De Jonghe and Gielen (2012) presented a
methodology to approximate nonlinear analog circuit models with a compact set of
analytical behavioral models. The trajectory piece-wise linear (TPWL) (Rewieński
and White 2006) method is a MOR method that consists of aggregating local linear
approximations around expansion points from state trajectories driven by training
inputs. An enhancement of the TPWL MOR method, which consists of an adaptive
Macromodeling and Synthesis of Analog Circuits 259
sampling of the linearization points across the model trajectory based on the error
between the nonlinear model and its linearized form, is proposed in Nahvi, Nabi,
and Janardhanan (2012).
Farooq et al. (2013) and Dong and Roychowdhury (2008) followed the main lines
of the TPWL MOR method but replaced the local linear models with piecewise poly-
nomial and Chebyshev interpolating polynomial models, respectively. This approach
improved the accuracy of the local reduced models but increased their on-the-fly
evaluation time. De Jonghe and Gielen (2012) presented a methodology to approxi-
mate nonlinear analog circuit models with a compact set of analytical behavioral
models. This method is based on transient simulations and curve-fitting techniques
and enhances the MOR method automation but does not overcome the input depen-
dency problem. Although some recent works (Aridhi, Zaki, and Tahar 2015, 2016)
have shown significant improvement in improving the modeling analysis of MOR
using some heuristic approaches, still it needs a lot of development for solving dif-
ferent types of circuits.
• Input nodes
• Hidden nodes
• Output nodes
The input layer or nodes are composed of input variables. The hidden layer or nodes
are composed of hidden neurons with calculated weights. They can be multilayers.
Output layer is composed of output parameters of the model. There can be one node
for a single output and many nodes for a multioutput model based on macromodeling
technique.
A basic ANN composed of three types of parameters:
The connection pattern is referred by the parameter “bias.” The weights of the con-
nections are updated during the training of the neural network, which is also called
as a learning process. The activation function converts the neuron’s weighted input
to its output activation.
Pandit, Mandal, and Patra 2009, 2010; Khandelwal, Garg, and Boolchandani 2015).
In these papers, robust and accurate macromodels have been generated, which is
observed from the optimized synthesis results (Boolchandani, Kumar, and Sahula
2009; Barros, Guilherme, and Horta 2007). Here, one vital thing is inferred that to
maintain the robustness of the model generated by LSSVM, kernel functions play a
major role.
Kernel functions are the type of functions used for pattern analysis and estima-
tion. They establish a relationship for different data sets and provide effective results
even if the data size and dimensions are very large. They provide a feature map
that is explicitly obtained from the data sets. They help in providing a very high-
quality performance along with SVM. They reduce the complex computation and
hence SVM implementation for generating the macromodels with a cheap computa-
tional cost (Chang and Lin 2011; Basak, Pal, and Patranabis 2007; Patro and Mandal
2017a,c; Patro, Panigrahi, and Mandal 2012).
The main aspect of this SVM to solve the regression type of problems is ε sensi-
tive loss. This means that the function estimated by the model must not deviate from
the actual targets by the factor of ε. In other words, the errors less than ε can be
ignored. The deviation of more than ε is the cost factor that needs to be optimized.
So this optimization problem is solved by its dual formulation and with the help of
Karush–Kuhn–Tucker condition. This is the key to the extension of this problem
even toward the nonlinear type of regression problems.
is a huge scope in the field of analog circuits macromodeling, and recently Patro and
Mandal (2016) focused on developing macromodel using this ML method for opera-
tional transconductance amplifier (OTA).
An overview of the structure of an ELM is shown in Figure 15.7. This is similar to
neural network, or in other words, it is the improvement of single-layer feedforward
network. But the main difference is that the hidden layer is not tuned.
Table 15.1 summarizes different modeling techniques with their applications and
complexities. ANN, SVM, and ELM have large area of applications. Due to low
complexity and less time required for model generation, these ML techniques are
used to generate macromodels for analog circuits.
TABLE 15.1
List of Macromodeling Techniques with Their Complexities
Sl. Macromodeling
No. Technique Applications Performance Analysis Complexity
1. Symbolic Different types of electrical and Less time for known Medium to
electronics circuits and networks, elements/circuits large
mathematical models, etc.
2. Posynomial Circuits, aerospace and structural Less time for linear Low to high
design, etc. behavioral systems
3. MOR Control system, MEMS, etc. Less time for small and Low to high
linear systems
4. ANN Almost all fields where mathematical Less time for small Low to high
or data analysis is required systems
5. SVM Almost all fields where mathematical Less time for small Low to
or data analysis is required systems medium
6. ELM Almost all fields where mathematical Less time for small Low
or data analysis is required systems
ANN, artificial neural network, ELM, extreme learning machine; MEMS, microelectromechanical system;
MOR, model order reduction; SVM, support vector machine.
264 Advanced VLSI Design and Testability Issues
15.4 CONCLUSIONS
This chapter provides a detailed analysis of several existing techniques for mac-
romodeling and generation of macromodels of analog circuits. These generated
macromodels can be used in the synthesis flow of analog circuits using various opti-
mization techniques such as genetic algorithm, particle swarm optimization, and so
on. At the end, comparison of different macromodeling techniques has been made
based on performance and complexity. It is found that ANN, SVM, and ELM show
better performance due to less complexity.
REFERENCES
Adhikari, Shyam Prasad, Kim Hyongsuk, Budhathoki Ram Kaji, Yang Changju, and O Chua
Leon. 2015. “A Circuit-Based Learning Architecture for Multilayer Neural Networks
with Memristor Bridge Synapses.” IEEE Transactions on Circuits and Systems I:
Regular Papers 62 (1). IEEE: 215–223.
Akhmadeev, Konstantin, Aya Houssein, Said Moussaoui, Einar A Høgestøl, Steffan D
Bos-Haugen, Hanne F Harbo, David Laplaud, Jennifer Graves, and Pierre-Antoine
Gourraud. 2018, July. “SVM-based Tool to Detect Patients with Multiple Sclerosis
Using a Commercial EMG Sensor.” In 2018 IEEE 10th Sensor Array and Multichannel
Signal Processing Workshop (SAM) (pp. 376–379). IEEE..
Antoulas, Athanasios C. 2005. Approximation of Large-Scale Dynamical Systems. SIAM.
Antoulas, Athanasios C, and Dan C Sorensen. 2001. “Approximation of Large-Scale
Dynamical Systems: An Overview.” International Journal of Applied Mathematics
and Computer Science 11: 1093–1121.
Aridhi, Henda, Mohamed H Zaki, and Sofiene Tahar. 2016. “Enhancing Model Order
Reduction for Nonlinear Analog Circuit Simulation.” IEEE Transactions on Very
Large Scale Integration (VLSI) Systems 24 (3). IEEE: 1036–1049.
Aridhi, Henda, Mohamed H Zaki, and Sofiène Tahar. 2015. “Fast Statistical Analysis of
Nonlinear Analog Circuits Using Model Order Reduction.” Analog Integrated Circuits
and Signal Processing 85 (3). Springer: 379–394.
Barros, Manuel, Guilherme Jorge, and Horta Nuno. 2007. GA-SVM feasibility model and
optimization kernel applied to analog IC Design automation. In Proceedings of the 17th
ACM Great Lakes Symposium on VLSI, 469–472.
Basak, Debasish, Pal Srimanta, and Patranabis Dipak Chandra. 2007. “Support Vector
Regression.” Neural Information Processing-Letters and Reviews 11(10): 203–224.
Bishop, Christopher M. 2006. Pattern Recognition and Machine Learning. Springer.
Boolchandani, D, Kumar Anupam, and Sahula Vineet. 2009. Multi-Objective Genetic
Approach for Analog Circuit Sizing Using SVM Macro-Model. In Proceedings of the
TENCON 2009-2009 IEEE Region 10 Conference, 1–6.
Borchani, Hanen, Varando Gherardo, Bielza Concha, and Larrañaga Pedro. 2015. “A Survey
on Multi-Output Regression.” Wiley Interdisciplinary Reviews: Data Mining and
Knowledge Discovery 5 (5). Wiley Online Library: 216–233.
Chang, Chih-Chung, and Lin Chih-Jen. 2011. “LIBSVM: A Library for Support Vector
Machines.” ACM Transactions on Intelligent Systems and Technology (TIST) 2 (3).
ACM: 27.
Chen, Wai-Kai. 2009. Analog and VLSI Circuits. CRC Press.
Chen, Yong. 1999. Model Order Reduction for Nonlinear Systems. Massachusetts Institute
of Technology.
Chen, Pai-Yu, Peng Xiaochen, and Yu Shimeng. 2018. “NeuroSim: A Circuit-Level Macro
Model for Benchmarking Neuro-Inspired Architectures in Online Learning.” IEEE
Macromodeling and Synthesis of Analog Circuits 265
Liu, Xueyi, Gao Chuanhou, and Li Ping. 2012. “A Comparative Analysis of Support Vector
Machines and Extreme Learning Machines.” Neural Networks 33. Elsevier: 58–66.
Liu, Pu, Tan Sheldon X-d, Yan Boyuan, and McGaughy Bruce. 2006. An extended SVD-
based terminal and model order reduction algorithm.” In Behavioral Modeling and
Simulation Workshop, Proceedings of the 2006 IEEE International, 44–49.
Manavalan, Balachandran, Shin Tae Hwan, and Lee Gwang. 2018. “DHSpred: Support-
Vector-Machine-Based Human DNase I Hypersensitive Sites Prediction Using the
Optimal Features Selected by Random Forest.” Oncotarget 9 (2). Impact Journals,
LLC: 1944.
Mandal, Sushanta K, Sural Shamik, and Patra Amit. 2008. “ANN-and PSO-Based Synthesis
of on-Chip Spiral Inductors for RF ICs.” IEEE Transactions on Computer-Aided
Design of Integrated Circuits and Systems 27 (1). IEEE: 188–192.
Mandal, Pradip, and Visvanathan V. 2001. “CMOS Op-Amp Sizing Using a Geometric
Programming Formulation.” IEEE Transactions on Computer-Aided Design of
Integrated Circuits and Systems 20 (1). IEEE: 22–38.
Martins, R, Lourenço N, Póvoa R, Canelas A, Horta N, Passos F, Castro-López R, Roca
E, and Fernández F. 2017. Layout-aware challenges and a solution for the automatic
synthesis of radio-frequency IC blocks. In 2017 14th International Conference on
Synthesis, Modeling, Analysis and Simulation Methods and Applications to Circuit
Design (SMACD), 1–4.
McConaghy, Trent, and Gielen Georges G E. 2009. “Template-Free Symbolic Performance
Modeling of Analog Circuits via Canonical-Form Functions and Genetic Programming.”
IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems 28
(8). IEEE: 1162–1175.
McCowan, Iain, Moore Darren, and Fry Mary-Jane. 2006. Classification of cancer stage
from free-text histology reports. In Engineering in Medicine and Biology Society,
2006. EMBS’06. 28th Annual International Conference of the IEEE, 5153–5156.
McCulloch, Warren S, and Pitts Walter. 1943. “A Logical Calculus of the Ideas Immanent in
Nervous Activity.” The Bulletin of Mathematical Biophysics 5 (4). Springer: 115–133.
Nahvi, S A, M Nabi, and Janardhanan S. 2012. Adaptive sampling of nonlinear system trajec-
tory for model order reduction. In 2012 Proceedings of International Conference on
Modelling, Identification & Control (ICMIC), 1249–1255.
Najm, Farid N. 2010. Circuit Simulation. John Wiley & Sons.
Odabasioglu, Altan, Celik Mustafa, and Pileggi Lawrence T. 1997. PRIMA: Passive reduced-
order interconnect macromodeling algorithm. In Proceedings of the 1997 IEEE/ACM
International Conference on Computer-Aided Design, 58–65.
Pandit, Soumya, Bhattacharya Sumit K, Mandal Chittaranjan, and Patra Amit. 2008. “A
Fast Exploration Procedure for Analog High-Level Specification Translation.” IEEE
Transactions on Computer-Aided Design of Integrated Circuits and Systems 27 (8).
IEEE: 1493–1497.
Pandit, Soumya, Mandal Chittaranjan, and Patra Amit. 2009. Systematic methodology for
high-level performance modeling of analog systems. In 2009 22nd International
Conference on VLSI Design, 361–366.
Pandit, Soumya, Mandal Chittaranjan, and Patra. 2010. “An Automated High-Level Topology
Generation Procedure for Continuous-Time ΣΔ Modulator.” Integration, the VLSI
Journal 43 (3). Elsevier: 289–304.
Patro, B Shivalal, and Mandal Sushanta K. 2017b. “Macro-Modeling of OTA Using ANN
for Fast Synthesis.” International Journal of Engineering Sciences and Management.
Patro, Shivalal, and Mandal Sushanta Kumar. 2017c. “A Multi Output Formulation for
Analog Circuits Using MOM-SVM.” Indonesian Journal of Electrical Engineering
and Computer Science 7(1): 90–96.
268 Advanced VLSI Design and Testability Issues
Patro, B Shivalal, and Mandal Sushanta K. 2016. “A Novel Modeling Technique for Operational
Amplifier Using RBF-ELM.” Journal of Engineering Science and Technology Review
9(4): 74–76.
Patro, B Shivalal, and Mandal Sushanta K. 2017a. “Support Vector Machine Based Macro-
Modeling of Voltage Controlled Oscillator for Fast Synthesis Purpose.” Journal of
Advanced Research in Dynamical and Control Systems, no. Special: 655–663.
Patro, B S, Panigrahi J K, and Mandal Sushanta K. 2012. A 6--17 GHz linear wide tuning
range and low power ring oscillator in 45nm CMOS process for electronic warfare.
In 2012 International Conference on Communication, Information & Computing
Technology (ICCICT), 1–4.
Patro, B Shivalal, and Vandana B. 2016. “Low Power Strategies for beyond Moore’s Law Era:
Low Power Device Technologies.” Design and Modeling of Low Power VLSI Systems.
IGI Global, 27.
Phillips, Joel R. 2000. Automated extraction of nonlinear circuit macromodels. In Custom
Integrated Circuits Conference, 2000. CICC. Proceedings of the IEEE 2000, 451–454.
Rewieński, Michał, and White Jacob. 2006. “Model Order Reduction for Nonlinear
Dynamical Systems Based on Trajectory Piecewise-Linear Approximations.” Linear
Algebra and Its Applications 415 (2–3). Elsevier: 426–454.
Rosenblatt, Frank. 1958. “The Perceptron: A Probabilistic Model for Information Storage
and Organization in the Brain.” Psychological Review 65 (6). American Psychological
Association: 386.
Rugh, Wilson John. 1981. Nonlinear System Theory. Johns Hopkins University Press.
Rutenbar, Rob A, Gielen Georges G E, and Roychowdhury Jaijeet. 2007. “Hierarchical
Modeling, Optimization, and Synthesis for System-Level Analog and RF Designs.”
Proceedings of the IEEE 95 (3). IEEE: 640–669.
Saad, Ramy A, and Soliman Ahmed M. 2010. “A New Approach for Using the Pathological
Mirror Elements in the Ideal Representation of Active Devices.” International Journal
of Circuit Theory and Applications 38 (2). Wiley Online Library: 148–178.
Sáenz Noval, Jorge Johanny, Roa Fuentes Elkim Felipe, Ayala Pabón Armando, and Van
Noije Wilhelmus. 2010. “A methodology to improve yield in analog circuits by using
geometric programming.” In Proceedings of the 23rd Symposium on Integrated
Circuits and System Design, 140–145.
Sánchez-López, C, Cante-Michcol B, Morales-López F E, and Carrasco-Aguilar M A. 2013.
“Pathological Equivalents of CMs and VMs with Multi-Outputs.” Analog Integrated
Circuits and Signal Processing 75 (1). Springer: 75–83.
Sánchez-López, C, Ochoa-Montiel R, Ruiz-Pastor A, and González-Contreras B M. 2013.
Symbolic nodal analysis of fully-differential analog circuits. In 2013 IEEE Fourth
Latin American Symposium on Circuits and Systems (LASCAS), 1–4.
Schaller, Robert R. 1997. “Moore’s Law: Past, Present and Future.” IEEE Spectrum 34 (6).
IEEE: 52–59.
Scholkopf, Bernhard, and Smola Alexander J. 2001. Learning with Kernels: Support Vector
Machines, Regularization, Optimization, and Beyond. MIT Press.
Shi, Guoyong. 2013. “A Survey on Binary Decision Diagram Approaches to Symbolic
Analysis of Analog Integrated Circuits.” Analog Integrated Circuits and Signal
Processing 74 (2). Springer: 331–343.
Shi, C-JR, and Tan Xiang-Dong. 2000. “Canonical Symbolic Analysis of Large Analog
Circuits with Determinant Decision Diagrams.” IEEE Transactions on Computer-
Aided Design of Integrated Circuits and Systems 19 (1). IEEE: 1–18.
Shi, Guoyong, Hu Hanbin, and Deng Shuwen. 2017. “Topological Approach to Automatic
Symbolic Macromodel Generation for Analog Integrated Circuits.” ACM Transactions
on Design Automation of Electronic Systems 22 (3). ACM: 1–25. doi:10.1145/3015782.
Macromodeling and Synthesis of Analog Circuits 269
Shokouhifar, Mohammad, and Jalali Ali. 2014. Automatic symbolic simplification of analog
circuits in MATLAB using ant colony optimization. In 2014 22nd Iranian Conference
on Electrical Engineering (ICEE), 407–412.
Shokouhifar, Mohammad, and Jalali Ali. 2015. “An Evolutionary-Based Methodology for
Symbolic Simplification of Analog Circuits Using Genetic Algorithm and Simulated
Annealing.” Expert Systems with Applications 42 (3). Elsevier: 1189–1201.
Singh, Vishal Kumar, and Bhardwaj Shweta. 2018. “Spam Mail Detection Using Classification
Techniques and Global Training Set.” Intelligent Computing and Information and
Communication, 623–632. Springer.
Smola, Alex J, and Bernhard Schölkopf. 2004. “A Tutorial on Support Vector Regression.”
Statistics and Computing 14 (3). Springer: 199–222.
Suykens, Johan A K, and Vandewalle Joos. 1999. “Least Squares Support Vector Machine
Classifiers.” Neural Processing Letters 9 (3). Springer: 293–300.
Suykens, Johan A K, Van Gestel Tony, and De Brabanter Jos. 2002. Least Squares Support
Vector Machines. World Scientific.
Tlelo-Cuautle, E, Sánchez-López C, and Moro-Frías D. 2010. “Symbolic Analysis of (MO)
(I) CCI (II)(III)-Based Analog Circuits.” International Journal of Circuit Theory and
Applications 38 (6). Wiley Online Library: 649–659.
Vapnik, Vladimir, Golowich Steven E, Smola Alex, and others. 1997. “Support Vector Method
for Function Approximation, Regression Estimation, and Signal Processing.” Advances
in Neural Information Processing Systems. Morgan Kaufmann Publishers, 281–287.
Vazzana, Giorgio Antonino, Grasso Alfio Dario, and Pennisi Salvatore. 2017. A tool-
box for the symbolic analysis and simulation of linear analog circuits. In 2017 14th
International Conference on Synthesis, Modeling, Analysis and Simulation Methods
and Applications to Circuit Design (SMACD), 1–4.
Wang, Shuheng, Li Guohao, and Bao Yifan. 2018. “A Novel Improved Fuzzy Support Vector
Machine Based Stock Price Trend Forecast Model.” ArXiv Preprint ArXiv: 1801.00681.
Werbos, Paul J. 1988. “Generalization of Backpropagation with Application to a Recurrent
Gas Market Model.” Neural Networks 1 (4). Elsevier: 339–356.
Winters-Hilt, Stephen, and Merat Sam. 2007. “SVM Clustering.” BMC Bioinformatics, 8: S18.
Wolfe, Glenn, and Vemuri Ranga. 2003. “Extraction and Use of Neural Network Models in
Automated Synthesis of Operational Amplifiers.” IEEE Transactions on Computer-
Aided Design of Integrated Circuits and Systems 22 (2). IEEE: 198–212.
Xi-Zhao, Wang, Qing-Yan Shao, Qing Miao, and Jun-Hai Zhai. 2013. “Architecture Selection
for Networks Trained with Extreme Learning Machine Using Localized Generalization
Error Model.” Neurocomputing 102. Elsevier: 3–9.
Xiong, Jian, Tian Shulin, and Yang Chenglin. 2016. “Fault Diagnosis for Analog Circuits
by Using EEMD, Relative Entropy, and ELM.” Computational Intelligence and
Neuroscience 2016. Hindawi.
Yu, Wen Xin, Sui Yongbo, and Wang Junnian. 2016. “The Faults Diagnostic Analysis for
Analog Circuit Based on FA-TM-ELM.” Journal of Electronic Testing 32 (4). Springer:
459–465.
Zhang, He, and Shi Guoyong. 2011. Symbolic behavioral modeling for slew and settling anal-
ysis of operational amplifiers. In 2011 IEEE 54th International Midwest Symposium on
Circuits and Systems (MWSCAS), 1–4.
Zhang, Ailin, and Shi Guoyong. 2018. “A Fast Symbolic SNR Computation Method and
Its Verilog-A Implementation for Sigma-Delta Modulator Design Optimization.”
Integration, the VLSI Journal 60. Elsevier: 190–203.
Zhang, Qi-Jun, Gupta Kuldip C, and Devabhaktuni Vijay K. 2003. “Artificial Neural
Networks for RF and Microwave Design-from Theory to Practice.” IEEE Transactions
on Microwave Theory and Techniques 51 (4). IEEE: 1339–1350.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
16 Performance-Linked
Phase-Locked Loop
Architectures: Recent
Developments
Umakanta Nanda
VIT-AP University
Biswajit Jena
Koneru Lakshmaiah Education Foundation
CONTENTS
16.1 Introduction ................................................................................................ 271
16.2 Performance-Linked Phase-Locked Loop Components ............................ 272
16.3 Recent Performance-Linked Architectures of Phase-Locked Loop
Blocks ......................................................................................................... 274
16.4 Design Challenges ...................................................................................... 284
16.5 Conclusion .................................................................................................. 285
References .............................................................................................................. 285
16.1 INTRODUCTION
Phase-locked loops (PLLs) [1–5] are used in almost all communication systems as
frequency synthesizers and clock generators. Sometimes the PLL itself is deliber-
ated as a unique system, and its design is one of the most challenging tasks, requir-
ing dedicated efforts. Design of high-performance PLLs contains combination of
several components of different features, covering the entire field of circuit designs
that are analog, digital, and mixed signal. To minimize the trade-off between power
consumption and phase noise [6,7] is the major challenge. The techniques and strate-
gies essential for implementing each building block of the PLL shown in Figure 16.1
271
272 Advanced VLSI Design and Testability Issues
CKRef Up
CPOut VCont VCOout
PFD Down CP LPF VCO
CKOut
FIGURE 16.1 Basic CP-PLL. CP-PLL, charge pump phase-locked loop; LPF, low-pass filter;
PFD, phase frequency detector; VCO, voltage-controlled oscillator.
differ. Each circuit needs a specific methodology, and it has to be driven by an appro-
priate design flow that certifies that the final design meets the traditional require-
ments. The charge pump PLL (CP-PLL) is the most promising one for realization in
CMOS technology.
A basic CP-PLL contains four major components: a phase frequency detector
(PFD), a CP, a loop filter, and a voltage-controlled oscillator (VCO). The PFD com-
pares the input reference signal (CK Ref) and the feedback signal (CKOut) with respect
to their phase and frequency and generates two signals, namely, Up and Down in
response [8]. The CP gets these two signals as its input and generates a constant
current into the loop filter (second order) having a capacitor CP and two resistors
R1 and R2 that filter the error voltage produced due to the phase difference between
CK Ref and CKOut. The filtered output voltage is termed as control voltage of the VCO
that changes the VCO frequency in the direction that reduces the phase difference
between the CK Ref and CKOut signals. When the control voltage matches the average
frequencies of the CK Ref and CKOut, the loop is said to be locked, and there exists a
cycle of CKOut for each cycle of CK Ref. However, they may not exactly match with
respect to their phase and have a constant or fluctuating phase difference; however,
excessive phase difference can cause loss of lock.
PLL Design
Techniques
Phase noise Spur Low noise & Wide lock Low noise
minimization Reduction Fast Locking range and power
FIGURE 16.2 Classification of PLL design techniques and their achievement. AFC, adap-
tive frequency calibration; PFD, phase frequency detector; PLL, phase-locked loop; VCO,
voltage-controlled oscillator.
element in the PFD, a dead zone–free PLL is reported in Refs. [13–19] with
low phase noise and fast locking capability.
For high-speed microprocessors and high-performance digital commu-
nication systems with clock generators, a fast locking PLL [20–22] is gen-
erally adopted. These kinds of application can sustain a certain amount of
phase noise but need to be fast. On the other hand, frequency synthesizers
for RF applications need to have ultralow phase noise. Hence, in modern
ICs, fast frequency acquisition, low phase noise, and wide lock range are
required in all most all PLL applications.
To reduce noise and jitter, designers focused on individual PLL compo-
nent’s optimization [9,23, and 24]. They also minimized the loop bandwidth
[25,26] to acquire improved noise performance, whereas bandwidth should
be enhanced to minimize the lock time. Hence, there is a tight trade-off
between the phase noise and frequency acquisition time. Hence, to achieve
simultaneously fast and low noise capability, dual/multi-PFD architectures
[29–31] are designed and reported in the past decade.
B. Charge Pump:
CP-PLLs are widely deployed in modern communication systems for
their larger gain, wide frequency acquisition range, and fast locking capa-
bility. However, nonideal effects such as current mismatch between the Up
and Down and increased glitches at the CP output motivated the designers
[32–38] to reduce current mismatch between Up and Down networks of the
CP and supply a signal with absolute no glitch in the control voltage to the
VCO. For this, they adopted only one current source that supplies both the
Up and Down currents. Transmission gates (TGs) were also used to dimin-
ish the nonideal effects such as clock feedthrough and charge injection to
overcome the glitches in VCO control voltage. Current mismatch can also
occur due to charge sharing [37–42] effect.
C. Voltage-Controlled Oscillator:
Due to the increasing demand for high-frequency multiband and mul-
tistandard transceivers in modern wireline, wireless communication, and
274 Advanced VLSI Design and Testability Issues
CKRef Up
D FF
Reset
D FF
CKOut Down
The transfer function of the above PFD is illustrated in Figure 16.4 that
shows the dead zone (φdz) [13] area in phase difference axis. It is the mini-
mum pulse width of the PFD output that is needed to turn on the CP com-
pletely. Here, the PFD is unable to detect the phase difference smaller than
dead zone. So the output of the PLL fluctuates within this range, triggering
jitter and phase noise further [63].
To prevent dead zone, a delay element block having delay of TD is usually
inserted in the reset path of the PFD. A new technique is deployed in Ref.
[13] to minimize the reference spur by maintaining the dead zone. To begin
with, they experimented with fixed delay elements. Then, controlling the
delay length by the feedback from the CP, a variable delay element is used.
The reference spurs stated here are reduced by around 20 and 24 dB at 50
and 100 MHz frequencies, respectively, (compare to Ref. [38]).
Dead zone, alternatively blind zone [14], appears when the input phase
difference goes near 2π. Here, the next rising edge of one of the inputs
following that of the other input at the time of reset cannot be detected by
the PFD. This topology uses only 16 transistors that eliminates blind zone
FIGURE 16.4 Dead zone (transfer function of PFD). PFD, phase frequency detector.
276 Advanced VLSI Design and Testability Issues
and achieves high acquisition speed. This topology avoids the reset process
when the phase difference ranges between π and 2π. In 0.5 μm CMOS pro-
cess, this PLL can operate at 800 MHz by using this PFD.
In Ref. [15], the precharging time for the internal parasitic capacitances is
responsible for the blind zone. Hence, using two extra transistors, they have
proposed a PFD that minimizes the precharging problem. A high-speed
PFD [12] with a delay cell and two additional transistors is designed here.
After simulation, the blind zone is compared and listed among the PFD of
this work [11,12]. In this work, the blind zone is stated to be 61 ps, whereas
in Refs. [11] and [12], it is reported as 156 and 221 ps, respectively. In Ref.
[16], the authors have adopted a modified dynamic logic style PFD for a
delay-locked loop (DLL) design. Here, they make one output of the PFD
high at a time even for a small phase difference of the inputs to the PFD.
Hence, the dead zone of the designed PLL at 800 and 84 MHz frequencies
is limited to 10.25 and 127.3 ps, respectively. As the jitter is directly trans-
lated from the dead zone, it is also limited in this design. In Ref. [17], the
two types of PFDs have been developed where the operation of the PFDs
does not depend on any reset process, thereby completely eliminating the
dead zone concept. The designers of Ref. [18] have considered double-gate
MOSFET (DG-MOSFET) for the NOR gate of the CP PFD, making it more
area efficient than its conventional counterpart. The reduced transistor count
also helps in lowering the parasitic capacitance that increases the speed of
the PFD. More importantly, for a phase difference of 60 and 80 ps, this PFD
generates the output to initiate the CP, which reaches a required threshold
level of logic high in 32- and 45-nm technologies, respectively. Hence, this
faster rise time at a very small phase error makes PFD to avoid dead zone in
a better manner than the PFD having conventional NOR gate. A new tech-
nique is adopted by deploying a modified Dickson CP design with charge
transfer switches (CTSs) in Ref. [19] to shrink the blind zone of a latch-
based PFD. Here the MOS transistor switches with proper on/off cycles are
employed as an alternative to the diodes to have the flow of charges in pump-
ing process, mentioned to as CTSs. The CTSs have been deployed to realize
the CPs and display better voltage pumping gain than the diodes.
The dead zone in Ref. [64] is reduced by making the pulse widths of Up
and Down half of their actual size. A predelay element is injected in Ref.
[65] for the reference clock and the local signals, respectively, before they
are alternatively provided to the reset inputs of the S–R latches. In this
work, the dead zone is removed and merged into the intervals of π or −π.
By inserting a delay of 1.2 ns, a reduced dead zone of 0.42 ns is achieved.
A new difference PFD [66] that can operate at high frequency uses an edge
detector circuit to avoid dead zone between two input signals. A comple-
mentary PFD is used in Ref. [67], where a phase detector and a frequency
detector operate in parallel. When the phase difference reaches below 180°,
the frequency detector is turned off. To make it dead zone free even at
4 GHz operating frequency, the AND gates between the inputs of the phase
detector are replaced by TGs. In Ref. [68], the reset part and the delay part
Performance-Linked PLL Architectures 277
are separated in the PFD block ensuring no erroneous condition at the ris-
ing edge of the input when the delay is active. In a very simple design, two
inverters acting as a buffer are used in Ref. [69] to get rid of the dead zone.
A double-edge-checking PFD is implemented in Ref. [70] that avoids Up
and Down signals to rise simultaneously reducing the dead zone. A delayed
version of the two inputs is generated in Ref. [73] where only the rising edge
of the inputs is delayed to minimize the blind zone of the PLL.
To get rid of the dead zone in Refs. [74,75] the PFD performs a fast reset
operation from modified tristate inverters acting as a D flip-flop where the
design can detect input phase differences as small as 750 fs at input operat-
ing frequencies of 38 kHz–2.5 GHz in 1.2 V and 90 nm CMOS technology.
Some recent PFD performance parameters are summarized in Table 16.1.
⎛ 2 ICP R ϕ K ⎞
⎜ 2π
e VCO
⎟ fCK Ref
Sr = 20 log ⎜ ⎟ − 20log ( dBc )
2 fCK Ref fPLPF (16.1)
⎜⎝ ⎟⎠
Ip
D
CKREF Up
Clk Q
S1
Reset VCont
CKOut Cp
Clk Q
Down
D
Ip
FIGURE 16.5 Conventional PFD and CP. CP, charge pump; PFD, phase frequency detector.
278
TABLE 16.1
Summary of the Literature Reporting Dead Zone Minimization in PFD
[14] [71] [15] [72] [66] [16] [17] [19] [74]
Performance Parameters 2007 2009 2010 2010 2010 2012 2013 2014 2015
Simulation type Post layout Schematic Experimental Schematic Schematic Schematic Post layout Experimental
Technology (nm) 500 180 130 130 130 180 180 180 90
VDD (V) 5 1.8 1.2 1.2 1.2 1.8 1.8 1.8 1.2
Lock-in time (ns) 150 – – 2800 90 265 4500 – –
Maximum frequency (GHz) 0.8 2.5 2.9 1.25 1.5 0.8 4 1 –
Dead zone/blind zone (ps) Nil Nil 52 Nil Nil 10.25 Nil Nil 0.7
Power consumption (mW) – 0.006 @ 50 MHz 0.496 @ 128 MHz 0.062 0.01 5 12.1 0.381 –
where R is the resistor value of the loop filter, ICP is the CP output current, φe
is the phase offset, fCK Ref is the frequency of the reference signal, and fPLPF is
the frequency of the pole of the loop filter.
Nonidealities (leakage current, current mismatch, and timing mismatch)
of the CP give rise to the glitches indicated in Figure 16.6. Even after 80 ns,
although the PLL is locked, the glitches are yet present, which can bring a
challenging amount of reference spur.
Architectures proposed in literature to improve the performance of CP
are broadly categorized into two types: “differential CP” and “single-ended
CP.” Differential CPs are more advantageous than single-ended [20]. For
example, the overall performance is not affected by the switch mismatch
between the NMOS and PMOS transistors. Moreover, the switches are
made up of only NMOS transistors, and due to its fully symmetric opera-
tion, the inverter delays between Up and Down signals do not generate any
offset. Compared with the single-ended, this CP doubles the range of output
compliance. Since the leakage current is a result of common mode off-
set, this configuration is not affected by it. Better immunity is provided by
two on-chip loop filters. However, differential CPs also suffer from some
critical drawbacks. To control the current mismatch, gain boosting circuits
[21,22] are used. When the Down is active, MN2 and MN3 work in concur-
rence with MN1 to deliver a gain boosting circuit. It rises the output resis-
tance of the CP and improves the current matching features; however, to
improve reference spur, the additional switching errors were not analyzed.
To minimize the reference spur level, the phase offset can be reduced, as
these are directly proportional to each other (Equation 16.1). To lessen the
phase offset, the turn-on time of the PFD can be reduced. Perfect current
matching characteristics are achieved in Ref. [24] by using an error ampli-
fier and reference current sources. To reduce the current mismatch, a second
compensation circuit is deployed in a CP, making it to have two push–pull
CPs and two replica-feedback biasing circuits working as compensators
[26]. In Ref. [27], the CP having a rail-to-rail error Op Amp (operational
amplifier) with reference circuit and cascade current mirror matches the
280 Advanced VLSI Design and Testability Issues
( ) (1 + λ ) ⋅ W (V )2 (1 + λNVDS )
2
I Down VGSP 3 − VTP P VDSP 3 N1 V
GS N 1 TN (16.2)
= N1
( ) (1 + λ ) W (V )2 (1 + λNVDS )
2
IUp VGSP1 − VTP P VDSP1 N3 V
GS N 3 TN N3
where IDown and IUp are the currents due to Down and Up networks, respec-
tively, VGSP3, VGSP1, VDSP3, and VDSP1 are the gate-to-source voltages and
drain-to-source voltage of PMOS 3 and 1, respectively, λP and λN are the
channel length modulation coefficients of PMOS and NMOS, respectively,
V TP and V TN are the threshold voltages of PMOS and NMOS, respectively,
and WN1 and WN3 are the widths of NMOS 1 and 3, respectively.
As the glitches in the CP output current are mostly responsible in the
reference spur, in Ref. [77], an additional buffer stage having a delay ele-
ment, a delay capacitor, and a MOSFET is introduced so that the glitches
will appear with some time delay. Moreover, the capacitance offered by
the delay capacitor and the transistor (parasitic) attenuates the magnitude
of the glitches. In Ref. [78], a differential CP design technique is adopted
to accomplish low charge sharing and charge injection. Incorporating a
sampled data common-mode feedback circuit, the output voltage range of
the CP is increased while lowering the mismatch between charging and
discharging current is realized. A dynamic current follow technology is
implemented in Ref. [79] to match the Up and Down currents, and two dif-
ferential pair inverters are incorporated to make the CP faster. The authors
in Ref. [80] claim that as the Up and Down currents are derived from a com-
mon reference current source, as long as the aspect ratios of the transistors
are same, there will not be any mismatch between Up and Down currents.
Introducing a new CP [81] to reduce charge sharing, current mismatch, and
charge injection problems, using differential current-steering switches with
DC reference voltage biasing at one side, feedthrough of the input pulses
can be eliminated. In Ref. [82], current mismatch compensator circuits are
introduced so that the CP will not be influenced by process variation and
power supply noise, which are also responsible for current mismatch and
Performance-Linked PLL Architectures 281
TABLE 16.2
Summary of the Literature Reporting Spur Reduction Using New Charge Pump Circuits
Performance [21] [22] [23] [24] [26] [27] [28] [80] [84] [87] [88]
Parameters 2006 2007 2008 2000 2009 2009 2010 2005 2007 2009 2011
Simulation type Schematic Post layout Post layout Post layout Schematic Post layout Experimental Schematic Schematic Schematic Experimental
Technology (nm) 350 90 180 250 130 180 130 180 180 180 180
VDD (V) 3.3 2.5 1.8 2.5 1.2 1.8 1.5 1 1.2 1.8 1.8
Reference spur – – – −75 – – – – – −55 –
(dBc/Hz)
Current mismatch 0.1 0.6 0.01 1 3.2 0.1 1 Nil 0.5 - 0.4
(%)
Area (mm2) – 0.289 15 40 – 0.036 3.96 – – – 0.036
Power – – – – 0.132 3 1 0.028 0.85 1 0.9
consumption
(mW)
Advanced VLSI Design and Testability Issues
Performance-Linked PLL Architectures 283
span of time. A PLL with phase error detector (PED) circuit is reported in
Ref. [42] to reduce both the power consumption and acquisition time. The
PED circuit delivers a dual-slope PFD and CP characteristic that effec-
tively shrinks the power consumption. A coarse-tuning current is acti-
vated to track the large phase difference for fast locking, and to complete
the fine adjustment near small phase difference, a fine-tuning current is
initiated.
Similarly, in another work [91], a dual-slope PFD and CP architecture
to accomplish fast locking of PLL is analyzed. The periodic ripples on
the control voltage of the VCO are randomized by a novel random clock
generator [40] where random selection of the PFDs is performed to reduce
the reference spur at the output of the PLL in locked state. In Ref. [92],
the PLL uses two PFDs (fast locking and low noise) for adaptively to cater
the need of locking performance and noise reduction purposes. The per-
formance parameters of the reported multi-PFD architectures are sum-
marized in Table 16.3.
D. Wide-Lock Range PLLs Using AFC Techniques:
Due to the increasing demand for high-frequency multiband and multi-
standard transceivers in modern communication systems, a wide-band fast
locking PLL achieving low phase noise is also highly essential. To improve
the locking process, lock range, and lock time and to reduce the total number
of comparisons, a code optimization along with a binary search algorithm
is used in Ref. [48]. The frequency range is found here to be more than
400 MHz, and the lock time is less than 65 μs when measured in 180-nm
technology. To enhance the frequency range, both the discrete and continu-
ous tuning mechanisms are adopted in Ref. [49]. An AFC technique is used
where an auxiliary digital loop to select a particular band of VCO is incor-
porated. This PLL was simulated and implemented on FPGA using Xilinx
TABLE 16.3
Summary of the Literature Reporting Performance Improvement Using
Multi-PFD Architectures
PLL with Multi PFD Architecture
Parameters [40] [41] [42] [91]
Simulation type Measured Measured Schematic Measured
Technology (nm) 180 1500 350 350
Frequency (GHz) 2.5 0.16 2.4 0.8
Lock range (GHz) 2.5–2.7 0.12–0.25 1.8–2.5 0.358–1.44
Lock-in time (ns) N/A 4500 150 3000
Phase noise (dBc/Hz at 1 MHz offset) −105 N/A N/A N/A
Power consumption (mw) 20 18.68 18.5 23.1
Layout area (mm2) 1.56 845 N/A 87.1
system generator. The lock-in time achieved here is 1.7 μs. In Ref. [50], the
VCO incorporates a 5-bit differential switched capacitor array to build a tun-
ing range from 2 to 3.2 GHz. The calibration time achieved here is less than
6 μs. Another AFC technique implemented in Ref. [51] works in two differ-
ent modes, namely, frequency calibration mode and store/load mode. In the
first mode, a new frequency detector is employed to reduce the lock time
to 16 μs, whereas in second mode of operation by loading the calibration
results stored after frequency calibration, the AFC makes the VCO come
back to the calibrated frequency in about 1 μs. In a 900-MHz PLL [52], an
automatic switched-capacitor discrete-tuning loop is deployed to have 20%
more tuning range than the conventional one with a calibration time of 2 ms
when measured in 0.6-μm technology. However, this method lacks in the
calibration time. A better solution is presented in Ref. [53] where both the
switched-capacitor bank LC VCO and the AFC technique are used to get a
tuning range of 600 MHz, which is as wide as 40% of the highest frequency.
However, this technique also suffers from high calibration time of the order
of tens of μs.
offset will always remain between the reference and feedback signals,
which increases the jitter.
D. Delay in Feedback Loop:
In mixed-signal PLL circuits, the delay introduced due to the digital
blocks in the feedback loop such as buffer, divider, and PFD produces a lin-
ear fluctuating phase shift as a function of frequency where the phase shift
is directly proportional to the operating frequency. Due to this, the phase
margin is also degraded that puts a question mark on the stability of the PLL.
Further for large-bandwidth PLL, the feedback loop is always troublesome.
E. PLL Sampling Effect:
PFD does not compare the phases of the reference and feedback signal
continuously. Practically, it is done only based on their edge position which
makes the PFD a sampled data system. A hold process is thought of during
the inactive time of the PFD when there is no signal to compare. The pro-
cess of phase sample and hold in a PLL leads a phase lag in feedback loop.
This degrades the PLL phase margin and again puts the doubt on stability.
16.5 CONCLUSION
PLL is a very widely used circuit technique in almost all modern electronic systems.
In this chapter, recent PLL architectures are discussed in relation with their perfor-
mance. The performance parameters are compared for each category of designing.
This chapter presents a comprehensive view of the PLL architectures.
REFERENCES
1. J. P. Frazier, J. Page, “Phase-lock loop frequency acquisition study,” IRE Transactions
on Space Electronics and Telemetry, vol. 8, no. 3, pp. 210–227, Sept. 1962.
2. F. M. Gardner, “Charge-pump phase-lock loops,” IEEE Transactions on Communications,
vol. 28, no. 11, pp. 1849–1858, Nov. 1980.
3. R. E. Best, Phase Locked Loops Design, Simulation and Applications, 5th ed. New York:
McGraw-Hill Publication, 2003.
4. U. Nanda, D. P. Acharya, S. K. Patra, Design of a low noise PLL for GSM applica-
tion. International Conference on Circuits, Controls and Communications (CCUBE),
Bengaluru, pp. 1–4, 2013.
5. F. M. Gardner, Phaselock Techniques, 3rd ed. New York: Wiley-Interscience, 2005.
6. C. Toumazou, G. S. Moschytz, B. Gilbert, Trade-Offs in Analog Circuit Design: The
Designer’s Companion. New York: Springer-Verlag, 2004.
7. P. K. Rout, D. P. Acharya, U. Nanda, Advances in analog integrated circuit optimi-
zation: a survey, Applied optimization methodologies in manufacturing systems. IGI
Global, USA, pp. 309–333, 2018.
8. I. Thompson, P. V. Brennan, “Phase noise contribution of the phase/frequency detec-
tor in a digital PLL frequency synthesiser,” IEE Proceedings of Circuits, Devices and
Systems, vol. 150, no. 1, pp. 1–5, Feb. 2003.
9. H. Johansson, “A simple precharged CMOS phase frequency detector,” IEEE Journal
of Solid-State Circuits, vol. 33, no. 2, pp. 295–299, Feb. 1998.
286 Advanced VLSI Design and Testability Issues
27. J. G. Maneatis, “Low-jitter process-independent DLL and PLL based on self-biased tech-
niques,” IEEE Journal of Solid State Circuits, vol. 31, no. 11, pp. 1723–1732, Nov. 1996.
28. C. S. Vaucher, “An adaptive PLL Tunning System architecture Combining High
Spectral purity and Fast settling time,” IEEE Journal of Solid State Circuits, vol. 35,
no. 4, pp. 490–502, Apr. 2000.
29. T.-W. Liao, J.-R. Su, C.-C. Hung, “Spur-reduction frequency synthesizer exploiting
randomly selected PFD,” IEEE Transaction on Very Large Scale Integration (VLSI)
Systems, vol. 21, no. 3, pp. 589–592, Mar. 2013.
30. Y. Woo, Y. M. Jang, M. Y. Sung, “Phase-locked loop with dual phase frequency detec-
tors for high-frequency operation and fast acquisition,” Microelectronics Journal,
Elsevier, vol. 33, pp. 245–252, 2002.
31. Y.-F. Kuo, R.-M. Weng, C.-Y. Liu, A fast locking PLL with phase error detector. IEEE
Conference on Electron Devices and Solid-State Circuits, 2005, pp. 423–426, 19–21
Dec. 2005.
32. W. Rhee, Design of high-performance CMOS charge pumps in phase-locked loops
Proceedings of the IEEE International Symposium on Circuits and Systems, ISCAS’99.
1999, vol. 2, pp. 545–548, Jul. 1999.
33. Y.-S. Choi, D.-H. Han, “Gain-boosting charge pump for current matching in phase-
locked loop,” IEEE Transaction on Circuits and Systems II: Express Briefs, vol. 53,
no. 10, pp. 1022–1025, Oct. 2006.
34. R. H. Mekky, M. Dessouky, Design of a low-mismatch gain-boosting charge pump
for phase-locked loops, International Conference on Microelectronics, ICM 2007.
pp. 321–324, 29–31 Dec. 2007.
35. J. Zhou, Z. Wang, A high-performance CMOS charge-pump for phase-locked loops
International Conference on Microwave and Millimeter Wave Technology, 2008, vol. 2,
pp. 839–842, 21–24 Apr. 2008.
36. J.-S. Lee, M.-S. Keel, S.-Il Lim, S. Kim, “Charge pump with perfect current matching
characteristics in phase-locked loops,” Electronics Letters, vol. 36, no. 23, pp. 1907–
1908, 9 Nov. 2000.
37. M. El-Hage, Y. Fei, Architectures and design considerations of CMOS charge pumps for
phase-locked loops. Canadian Conference on Electrical and Computer Engineering,
2003.IEEE CCECE 2003. vol. 1, pp. 223–226, 4–7 May 2003.
38. M.-S. Hwang, J. Kim, D.-K. Jeong, “Reduction of pump current mismatch in charge-
pump PLL,” Electronics Letters, vol. 45, no. 3, pp. 135–136, 29 Jan. 2009.
39. N. Hou, Z. Li, Design of high performance CMOS charge pump for phase-locked loops
synthesizer. 15th Asia-Pacific Conference on Communications, pp. 209–212, 8–10 Oct.
2009.
40. M. Jung, A. Ferizi, R. Weigel, A charge pump with enhanced current matching
and reduced clock-feedthrough in wireless sensor nodes Asia-Pacific Microwave
Conference Proceedings (APMC), pp. 2291–2294, 7–10 Dec. 2010.
41. C. Zhang, T. Au, M. Syrzycki, A high performance NMOS-switch high swing cascode
charge pump for phase-locked loops. IEEE 55th International Midwest Symposium on
Circuits and Systems (MWSCAS), pp. 554–557, 5–8 Aug. 2012.
42. N. Kamal, S. F. Al-Sarawi, D. Abbott, “Reference spur suppression technique using ratioed
current charge pump,” Electronics Letters, vol. 49, no. 12, pp. 746–747, 6 Jun. 2013.
43. M. Brandolini, P. Rossi, D. Manstretta, F. Svelto, “Toward multistandard mobile termi-
nals -fully integrated receivers requirements and architectures,” IEEE Transaction on
Microwave Theory and Techniques, vol. 53, no. 3, pp. 1026–1038, 2005.
44. U. Nanda, D. P. Acharya, S. K. Patra, “Low noise and fast locking phase locked loop
using a variable delay element in the phase frequency detector,” Journal of Low Power
Electronics, American Scientific Publishers, vol. 10, no. 1, pp. 53–57, 2014.
288 Advanced VLSI Design and Testability Issues
61. B. Han, J. Wu, C. Hu, A 2.5 GHz low phase noise LC VCO in 0.35 um SiGeBiCMOS
technology. 7th International Conference on ASICON’07 by IEEE, 22–25 Oct. 2007.
62. R. J. Baker, H. W. Li, D. E. Boyce, CMOS circuit design, layout, and simulation. IEEE
Press Series on Microelectronic Systems, 2002.
63. G. S. Singh, D. Singh, S. Moorthi, Low power low jitter phase locked loop for high
speed clock generation. Asia Pacific Conference on Postgraduate Research in
Microelectronics and Electronics (PrimeAsia), 2012, pp. 192, 196, 5–7 Dec. 2012.
64. G. B. Lee, P. K. Chan, L. Siek, “A CMOS phase frequency detector for charge pump
phase-locked loop,” 42nd Midwest Symposium on Circuits and Systems, vol. 2,
pp. 601–604, 1999.
65. D.-C. Juang, D.-S. Chen, J.-M. Shyu, C.-Y. Wu, A low-power 1.2 GHz 0.35 μm CMOS
PLL. Proceedings of the Second IEEE Asia Pacific Conference on ASICs, pp. 99–102,
2000.
66. K.-H. Cheng, T.-H. Yao, S.-Y. Jiang, W.-B. Yang, A difference detector PFD for low jit-
ter PLL. The 8th IEEE International Conference on Electronics, Circuits and Systems,
pp. 43–46, 2001.
67. M. Renaud, Y. Savaria, A CMOS three-state frequency detector complementary to an
enhanced linear phase detector for PLL, DLL or high frequency clock skew measure-
ment. Proceedings of the 2003 International Symposium on Circuits and Systems,
ISCAS '03., vol. 3, pp. III-148–III-151, 25–28 May 2003.
68. K.-S. Lee, B.-H. Park, H.-I. Lee, M. J. Yoh, Phase frequency detectors for fast fre-
quency acquisition in zero-dead-zone CPPLLs for mobile communication systems.
Proceedings of the 29th European Solid-State Circuits Conference, ESSCIRC '03.,
pp. 525–528, 16–18 Sept. 2003.
69. M. M. Kamal, E. W. El-Shewekh, M. H. El-Saba, Design and implementation of a
low-phase-noise integrated CMOS frequency synthesizer for high-sensitivity nar-
row-band FM transceivers. Proceedings of the 15th International Conference on
Microelectronics, 2003, pp. 167–175, 9–11 Dec. 2003.
70. C.-P. Chou, Z.-M. Lin, J.-D. Chen, A 3-ps dead-zone double-edge-checking phase-
frequency-detector with 4.78 GHz operating frequencies. Proceedings of the IEEE
Asia-Pacific Conference on Circuits and Systems, vol. 2, pp. 937–940, 6–9 Dec. 2004.
71. N. M. H. Ismail, M. Othman, CMOS phase frequency detector for high speed applica-
tions, International Conference on Microelectronics, pp. 201–204, 19–22, Dec. 2009.
72. J. Lan, Y. Wang, L. Liu, R. Li, A nonlinear phase frequency detector with zero blind zone
for fast-locking phase-locked loops. International Conference on Anti-Counterfeiting
Security and Identification in Communication, pp. 41–44, 18–20 Jul. 2010.
73. C. Zhang, M. Syrzycki, Modifications of a dynamic-logic phase frequency detector for
extended detection range. 53rd IEEE International Midwest Symposium on Circuits
and Systems (MWSCAS), pp. 105–108, 1–4 Aug. 2010.
74. J. Strzelecki, S. Ren, “Near-zero dead zone phase frequency detector with wide input
frequency difference,” Electronics Letters, vol. 51, no. 14, pp. 1059–1061, 2015.
75. M. K. Hati, T. K. Bhattacharyya, A PFD and charge pump switching circuit to optimize
the output phase noise of the PLL in 0.13-μm CMOS. International Conference on
VLSI Systems, Architecture, Technology and Applications, pp. 1–6, 8–10 Jan. 2015.
76. B. Razavi, Design of Analog CMOS Integrated Circuits. New York: McGraw-Hill, 2000.
77. R. A. Baki, M. N. El-Gamal, A new CMOS charge pump for low-voltage (1V) high-
speed PLL applications. Proceedings of the International Symposium on Circuits and
Systems, ISCAS '03., vol. 1, pp. I-657–I-660, 25–28 May 2003.
78. B. Terlemez, J. Uyemura, The design of a differential CMOS charge pump for high
performance phase-locked loops. Proceedings of the 2004 International Symposium
on Circuits and Systems, ISCAS '04., vol. 4, pp. IV, 56 1–4, 23–26 May 2004.
290 Advanced VLSI Design and Testability Issues
79. Z. Tao, Z. Xuecheng, S. Xubang, A CMOS charge pump with a novel structure in PLL.
Proceedings of 7th International Conference on Solid-State and Integrated Circuits
Technology, vol. 2, pp. 1555–1558, 18–21 Oct. 2004.
80. H. Yu, Y. Inoue, Y. Han, A new high-speed low-voltage charge pump for PLL applica-
tions. 6th International Conference on ASIC, vol. 1, pp. 387–390, Oct. 2005.
81. S. J. Byun, B. Kim, C.-H. Park, Charge pump circuit for a PLL. US Patent 6,952,126
B2, 4 Oct. 2005.
82. K. S. Ha, L. S. Kim, Charge pump reducing current mismatch in DLLs and PLLs.
Proceedings of ISCAS, Kos, Greece, pp. 2221–2224, May 2006.
83. N. T. Hieu, T. W. Lee, H. H. Part, A perfectly current matched charge pump of CP-PLL
for chip-to-chip optical link. Pacific Rim Conference on Lasers and Electro-Optics,
Seoul, Korea, pp. 1–2, Aug 2007.
84. Y. Sun, L. Siek, P. Song, Design of a high performance charge pump circuit for low
voltage phase-locked loops. International Symposium on Integrated Circuits, ISIC '07,
pp. 271–274, 26–28 Sept. 2007.
85. C.-F. Liang, S.-H. Chen, S.-I. Liu, “A digital calibration technique for charge pumps in
phase-locked systems,” IEEE Journal of Solid-State Circuits, vol. 43, no. 2, pp. 390–
398, Feb. 2008.
86. W.-M. Lin, S.-I. Liu, C.-H. Kuo, C.-H. Li, Y.-J. Hsieh, C.-T. Liu, “A phase-locked loop
with self-calibrated charge pumps in 3 um LTPS-TFT technology,” IEEE Transactions
on Circuits and Systems II: Express Briefs, vol. 56, no. 2, pp. 142–146, Feb. 2009.
87. W.-H. Chiu, T.-S. Chang, T.-H. Lin, A charge pump current miss-match calibration
technique for ΔΣ fractional-N PLLs in 0.18-μm CMOS. IEEE Asian Solid-State
Circuits Conference, A-SSCC 2009, pp. 73–76, 16–18 Nov. 2009.
88. L. Zhiqun, Z. Shuangshuang, H. Ningbing, “Design of a high performance CMOS
charge pump for phase-locked loop synthesizers,” Journal of Semiconductors, vol. 32,
no. 7, p. 075007, Jul. 2011.
89. Y.-W. Chen, Y.-H. Yu, Y.-J. E. Chen, “A 0.18 μm CMOS dual-band frequency synthe-
sizer with spur reduction calibration,” IEEE Microwave and Wireless Components
Letters, vol. 23, no. 10, pp. 551–553, Oct. 2013.
90. K. M. Ware, H.-S. Lee, C.G. Sodini, “A 200-MHz CMOS phase-locked loop with dual
phase detectors,” IEEE Journal of Solid-State Circuits, vol. 24, no. 6, pp. 1560–1568,
Dec. 1989.
91. K.-H. Cheng, W.-B. Yang, “A dual-slope phase frequency detector and charge pump
architecture to achieve fast locking of phase-locked loop,” IEEE Transactions on Circuits
and Systems II: Analog and Digital Signal Processing, vol. 50, no. 11, pp. 892–896, Nov.
2003.
92. U Nanda, D P Acharya, Adaptive PFD “Selection Technique for Low Noise and Fast
PLL in Multi- standard Radios,” Microelectronics Journal, Elsevier, vol. 64, pp. 92–98,
2017.
17 Review of Analog-to-
Digital and Digital-to-
Analog Converters
for A Smart Antenna
Application
B. S. Patro, A. Senapati, and T. Pradhan
Kalinga Institute of Industrial Technology,
KIIT Deemed to be University
CONTENTS
17.1 Introduction ................................................................................................ 291
17.1.1 Butler Matrix ................................................................................ 294
17.1.2 Architecture of Smart Antenna System........................................ 294
17.1.2.1 Receiver........................................................................ 294
17.1.2.2 Transmitter ................................................................... 296
17.2 Analog-to-Digital Converters and Digital-to-Analog Converters .............. 297
17.2.1 Basics of Analog-to-Digital Conversion ....................................... 297
17.2.2 Sampling ....................................................................................... 298
17.2.3 Quantization ................................................................................. 299
17.2.4 Successive Approximation Register Analog-to-Digital
Converter ...................................................................................... 300
17.2.4.1 Analog-to-Digital Converter ........................................ 301
17.2.5 Flash Analog-to-Digital Converter ............................................... 303
17.2.6 Pipelined Analog-to-Digital Converter ........................................ 304
17.2.7 Delta–Sigma Analog-to-Digital Converter ................................... 305
17.3 Digital-to-Analog Converter....................................................................... 305
17.3.1 Feedback Digital-to-Analog Converter ........................................ 305
17.3.2 Decimator ..................................................................................... 305
17.4 Conclusion .................................................................................................. 306
References .............................................................................................................. 307
17.1 INTRODUCTION
In recent years, there is an enormous rise in traffic for mobile and personal com-
munication systems. The rise in traffic put a demand on both manufacturers and
291
292 Advanced VLSI Design and Testability Issues
operators to increase capacity (Murch and Letaief 2002; Fletcher and Darwood
1998; Tsai and Woerner 2001; Poormohammad and Farzaneh 2018). Capacity
enhancement is a major challenge for service providers as the radiation environment
is affected by negative factors to limit the capacity (Consortium and others 2005;
Shivapanchakshari and Aravinda 2019).
The spectral efficiency of the networks is maximized by the wireless carriers
by exploring novel ways in order to improve their return on investment (Godara
2004). Lots of efforts have been put by the researchers to improve the wireless sys-
tems performance. Smart antennas (SAs) have shown improved quality and cover-
age efficiently in the wireless network systems (Chryssomallis 2000). SA systems
have received much attention in the past few years (Chryssomallis 2000; Andersson
et al. 1991; Tsoulos et al. 1998; Kohno 1998; Rappaport 1998; Tsoulos 2000) because
they can increase system capacity (very important in urban and densely populated
areas) by dynamically tuning out interference while focusing on the intended user
(Boukalov and Haggman 2000; Liberti and Rappaport 1999) along with impressive
advances in the field of digital signal processing.
Base station antennas in a cellular system are either omnidirectional or sector-
ized. There is a waste of resources because the majority of transmitted signal power
radiates in directions other than desired user directions. Also, the signal power radi-
ated over the coverage area of the network will experience interference by other
undesired users. An SA system solves this problem by transmitting power toward
the desired directions only. Adaptive beamforming capability in SA systems enables
flexible synthesis and steering of antenna beams for optimized signal-to-noise and
signal-to-interference ratio (SIR) performances.
SAs, also known as adaptive antennas and digital antenna arrays, are antenna
arrays with smart signal-processing algorithms. Reconfigurable antennas have a
similar capability as SAs, but they are single-antenna elements and not antenna
arrays. SA techniques are used in acoustic signal processing, track and scan radar,
radio astronomy, and radio telescope and cellular systems such as W-CDMA,
UMTS, and LTE.
The deployment of SAs for wireless communications has emerged as one of the
leading technologies for achieving high-efficiency networks that maximize capacity
and improve quality and coverage. SA systems have received much attention in the
past few years because they can increase system capacity (very important in urban
and densely populated areas) by dynamically tuning out interference while focusing
on the intended user along with impressive advances in the field of digital signal pro-
cessing. The International Telecommunication Union (ITU) suggested an SA system
as one of the key technologies for the fourth generation (4G) and beyond to improve
wireless network capacity and mobility. Selected control algorithms, with predefined
criteria, provide adaptive arrays (AAs) the unique ability to alter the radiation pat-
tern characteristics (nulls, side-lobe level, main beam direction, and beamwidth).
These control algorithms originate from several disciplines and target-specific appli-
cations (e.g., in the field of seismic, underwater, aerospace, and more recently cel-
lular communications). The commercial introduction of SAs is a great promise for a
big increase in system performance in terms of capacity, coverage, and signal qual-
ity, all of which will ultimately lead to increased spectral efficiency.
A Review of ADC and DAC for Smart Antenna 293
The main functions of SAs are direction-of-arrival (DOA) estimation and beam-
forming. DOA estimation: In order to maximize the performance of the adaptive
SA, the accurate estimation of the DOA of all signals transmitted to the AA antenna
is required. The DOA estimation methods is categorized as a nonsubspace method
and the subspace method. Nonsubspace methods are simpler, but performance is
not good in terms of resolution. It depends on the spatial spectrum, and DOAs are
obtained as locations of peaks in the spectrum. The main advantage of these tech-
niques is that it can be used in situations where there is a lack of information about
the properties of the signal. Various DOA estimation algorithms available in the
literature are MUSIC, Improved MUSIC, and ESPIRIT (Gross 2005; Balanis and
Ioannides 2007; Basha, Sridevi, and Prasad 2013).
Beamforming: After the DOA estimation, the function of the beamforming block
is to generate the main beam toward the user and nulls toward interferers. Various
beamforming algorithms available in literature are least mean square (LMS), sample
matrix inversion (SMI), recursive least square (RLS), constant modulus algorithm
(CMA), and conjugate gradient method (CGM) (Gross 2005; Patel, Makwana, and
Parmar 2016; Awan et al. 2017; Lakshmi, Sivvam, and Rajyalakshmi 2018). A func-
tional block diagram of the digital signal–processing part of an AA antenna sys-
tem is shown in Figure 17.1. The SA system downconverts the received signals to
baseband and digitizes them. Then using DOA algorithm, the signal-of-interest and
signal-not-of-interest is located by dynamically changing the complex weights [w0,
w1, …, wM-1]. Then the adaptive algorithm such as LMS, RLS, or SMI computes the
appropriate weights to produce an optimum radiation pattern, i.e., main beam toward
user and null toward interferer.
Smart antenna configurations:
FIGURE 17.1 Smart antenna block diagram. DOA, direction of arrival; DSP, digital signal
processor.
294 Advanced VLSI Design and Testability Issues
1. Switched Beam Antenna: In some SAs, the beam patterns are fixed. This
kind of SAs are known as switched beam (SB) antennas. Depending upon
the requirements of the system, which beam has to be selected has to be
decided accordingly. The benefits of SB antennas are that these can be eas-
ily fed into the present cell structures and are of low cost.
2. Adaptive Antenna Array: Adaptive SAs are the ones that let the beam
drive itself in any direction and at the same time eliminate the interfering
signals. The beam direction can be evaluated by the DOA estimation meth-
ods. It is a highly rated SA and consists of a complex transceiver.
FIGURE 17.2 Smart antennas: (a) switched beam antenna and (b) adaptive antenna array.
A Review of ADC and DAC for Smart Antenna 295
and multiplied with complex weights [w1, w2, w3 …, wM], which are calculated by
the signal-processing unit. The weight updating is done using various adaptive
signal processing algorithms such as LMS and its variants, SMI, RLS, CMA, and
CGM. Linearly constrained minimum variance (LCMV) is used for multiple-input
multiple-output (MIMO) antenna system (Wang, Chen, and Jiang 2018). SA with
electrically steerable parasitic array radiator (ESPAR) is implemented using an adap-
tive beamforming algorithm based on simultaneous perturbation stochastic approxi-
mation (Ganguly, Ghosh, and Kumar 2019). Particle swarm optimization is used for
power and phase of excitation delivered to array elements using field-programmable
gate array (FPGA) (Greda et al. 2019). Weights are estimated to maximize the power
of the received signal from the desired user (SB or phased array [PA]) and to max-
imize the SIR by suppressing the received signal from interfering sources (AA).
FPGA is used for real-time implementation of SA arrays with ADC and DAC.
The method for calculating the weights differs depending on the type of optimiza-
tion criterion. When the SB is used, the receiver will test all the predefined weight
vectors (corresponding to the beam set) and choose the best one giving the strongest
received signal level. If the PA approach is used, which consists of directing a maxi-
mum gain beam toward the strongest signal component, the weights are calculated after
the DOA is first estimated. A number of well-documented methods exist for estimating
the DOA. In the AA approach, where maximization of SIR is needed, the optimum
weight vector (of dimension M) wopt can be computed using a number of algorithms
such as optimum combining and others that will follow (Balanis and Ioannides 2007).
When the beamforming is done digitally (after ADC), the beamforming and
signal-processing units can normally be integrated with the same unit (digital signal
processor). The separation in Figure 17.3 is done to clarify the functionality. The
beamforming can be performed at either RF or intermediate frequency.
17.1.2.2 Transmitter
The transmission part of an SA system is shown in Figure 17.4. It is schematically
similar to its reception part. The signal is split into N branches, which are weighted
by the complex weights [w1, w2 , …, wN] in the beamforming unit. The weights are
calculated by the signal-processing unit. The radio unit consists of D/A converters
and the upconverter chains. The principal difference between uplink and downlink
is that since there are no SAs applied to the user terminals (mobile stations), there
is only limited knowledge of the channel state information available. Therefore, the
optimum beamforming in downlink is difficult, and the same performance as the
uplink cannot be achieved (Balanis and Ioannides 2007).
The main blocks for any SA systems are ADC and DAC. Without this improved
and faster antenna system, to provide and process the signal is a tedious task. The
components that are essential for the receiver part as well as transmitter part are an
ADC and a DAC. ADC is used at the receiver side and DAC is used at the transmitter
side of the SA array. The receiver will receive the RF signal, but it is needed to be
translated to an accurate digital signal at each antenna element. The job of translat-
ing an analog RF signal to an accurate digital signal is carried out by an appropriate
ADC. Similar is the case with transmitter of the antenna where the reverse of ADC
happens. At the transmitter end, the input signal is mostly digital, whereas the output
is of real-time analog signal, and this conversion of digital to analog is done by a
DAC. The next section will cover about this essential part of an antenna with various
types of ADCs and DACs.
FIGURE 17.5 General block diagram of an ADC. ADC, analog-to-digital converter; S/H,
sample and hold.
If we consider any basic ADC, there are three basic blocks as sampler, quantizer,
and coder. The sampler determines the output data rate and is responsible for issues
such as antialiasing in Nyquist ADC. Quantizer takes the analog-sampled signal as
input and maps it to the discrete quantized value. Quantization noise is added to the
ADC at the quantizer block not at the sampler. Coder is basically a digital filter or
digital coder that collects the data from the quantizer. So the most important compo-
nents of an ADC are sampler and quantizer.
17.2.2 SAMPLING
It is a process that converts the continuous input to a discrete signal in time and imposes
a limit on the bandwidth of the analog input signal. According to the Nyquist theorem,
to prevent information loss, x(t) must be sampled at a minimum rate of f N = 2BW, often
referred to as the Nyquist frequency, and thus the ADCs in which analog input signal
is sampled at the minimum rate (fs = f N) are called Nyquist ADCs. Conversely, ADCs
in which fs > f N are called oversampling ADCs. The speed prior to the Nyquist rate for
an ADC is expressed in terms of the oversampling ratio, defined as
fs
OSR =
2 Bw
FIGURE 17.7 Antialiasing filter for (a) Nyquist-rate ADC and (b) oversampling ADC.
ADC, analog-to-digital converter.
Bw aligns with fs/2, then aliasing will occur if xa(t) contains frequency components
above fs/2. High-order analog AAFs are thus required to implement sharp transition
bands capable of removing out-of-band components with no significant attenuation
of the signal band, as illustrated in Figure 17.7a for the low-pass case. Conversely,
as fs/2 > Bw in oversampling ADCs, the replicas of the input signal spectrum that
are created by the sampling process are farther apart than in Nyquist ADCs. As
illustrated in Figure 17.7b, frequency components of the input signal in the range
[Bw, fs − Bw] do not alias within the signal band, so that the filter transition band can
be smoother, which greatly reduces the order required for the AAF and simplifies
its design.
17.2.3 QUANTIZATION
The quantization process also introduces a limitation on the performance of an ideal
ADC, because an error is generated while performing the continuous-to-discrete
transformation of the input signal in amplitude, commonly referred to as quanti-
zation error. The operation of quantizers is illustrated in Figure 17.8. As a matter
of example, Figure 17.8c depicts the I/O characteristic of a quantizer with N = 2,
although results apply to a generic N-bit quantizer. Input amplitudes within the full-
scale input range [−XFS /2, + XFS /2] are rounded to 1 out of the 2N different output
levels, which are usually encoded into a binary digital representation. If these levels
are equally spaced, the quantizer is said to be uniform, and the separation between
adjacent output levels is defined as the quantization step:
YFS
Δ=
2N − 1
where YFS stands for the full-scale output range. As XFS and YFS are not necessar-
ily equal, the quantizer may exhibit a gain different from unity, as indicated in
300 Advanced VLSI Design and Testability Issues
FIGURE 17.8 Illustration of the quantization process: (a) multibit quantizer block, (b) single-
bit quantizer block, (c) I/O characteristic of a multibit quantizer, (d) I/O characteristic of a
single-bit quantizer, (e) multibit quantization error, and (f) single-bit quantization error.
Figure 17.8c by the slope kq. As shown in Figure 17.8e, the quantizer operation thus
inherently generates a rounding error that is a nonlinear function of the input. Note
that, if q(n) is kept within the range [−XFS /2, +XFS /2], the quantization error e(n) is
bounded within [−Δ/2, + Δ/2]. The former input range is known as the nonoverload
region of the quantizer, as opposed to ranges with |q(n)|> Δ /2, for which the magni-
tude of e(n) grows monotonously.
Figure 17.8 also shows the operation of a single-bit quantizer (N = 1). From Figure
17.8d, it can be noted that, compared with the multibit case, the output of a single-bit
quantizer is determined by the input sign only, regardless of its magnitude. Therefore,
the gain kq is undefined and can be arbitrarily chosen.
The following section will discuss about various ADCs and DACs with their ori-
gins and various characteristics. Figure 17.9 shows the classification of various types
of ADCs.
is again converted to respective analog value and compared with the real-time input
value. The compared value is again fed to SAR logic where it again approximated
the appropriate digital value, and this process is repeated until a minimum accuracy
is reached. The main disadvantage of this ADC is low speed; it is slower than other
ADCs due to the approximation procedure.
The research on SAR-ADC became stagnant with the invention of the pipelined
ADC. The first pipelined ADC was invented in the late 1980s (Lewis and Gray 1987;
Lewis et al. 1992). The later attracted more researchers due to its speed and resolu-
tion bit production at that time. Thus, SAR-ADC was out of the research scope for
almost more than a decade.
In the year 2004, Dieter Draxelmayr presented a paper on time-interleaved SAR-
ADC, which was a low-power high-speed SAR-ADC at that time (Draxelmayr
2003). Dieter Draxelmayr presented the paper in ISSCC (International Solid-State
Circuits Conference). This was the time when flash ADCs were among the fastest
ADCs but with a disadvantage of high-power consumption. Flash ADCs use a 2N
number of comparators to compare VIN with their reference voltages simultane-
ously, thus making itself a power hunger design. SAR-ADC on the other hands uses
only a single comparator with some switches, which made SAR-ADC compatible
for CMOS process miniaturization and future proof. With the presentation of the
paper by Dieter Draxelma at ISSCC, researchers were again inclined toward the
SAR-ADC as the concern about the speed had been eliminated. This has been again
proved from the increased number of papers published on SAR-ADCs after 2004
(Matsuura 2016).
Craninckx and Van der Plas in 2007 at IEEE ISSCC presented a paper on a
dynamic low-power SAR-ADC with a passive charge-sharing technique and asyn-
chronous controller. The prototype was designed without using any active circuits,
which are in general used for increasing the speed. Thus, they cut off any chances of
static power loss in their design. Although the concerns remained with the resolution
that was just below 10 ENOB and speed of the design. Their prototype achieved a
power consumption of 290 mu W in 90-nm digital CMOS technology at that time
(Craninckx and Van Der Plas 2007).
Pieter Harpe, Eugenio Cantatore, and Arthur van Roermund published a journal
with a theory of the Data-Driven Noise-Reduction Method (DDNR) at IEEE Journals
of Solid-State Circuits in 2013 (Harpe, Cantatore, and Van Roermund 2013). In this
paper, they introduced DDNR and also claimed that it has enhanced the noise perfor-
mance of the comparator, selectively. They have sacrificed the power consumption
factor to achieve a higher resolution in their design. With a self-oscillating compara-
tor, they had managed to generate an internal clock for bit cycling, which limited the
use of an external clock to generate only sample rate frequency. The design used a
segmented capacitive DAC to reduce differential nonlinearity (DNL) error as well
as to save power. The prototype gained a resolution up to 12 bit and also reduced the
leakage power to ensure a maintained efficiency even at low sample rates.
In 2012, Akira Shikata, Ryota Sekimoto, Tadahiro Kuroda, and Hiroki Ishikuro
proposed the use of a trilevel comparator for SAR-ADC with their published journal
at IEEE Journal of Solid-State Circuits, where they used the trilevel comparator for
reducing the load of the comparator with reference to the speed requirement. With
that, they reduced the resolution of internal DAC by 1 bit. The paper implies that
the authors had managed to operate their proposed design at a single low voltage
of 0.5 V, and therefore, their design achieved a low power of 1.2 μW with 1.1 MHz
sampling frequency. In this paper, the capacitor mismatch problem at the level of
DAC was addressed by the capacitor array with the ability to be reconfigured and
A Review of ADC and DAC for Smart Antenna 303
open-loop gain. The second stage of this prototype used SAR-ADC to reduce power
and increase resolution, which eliminated the use of more pipelines.
In 2019, Kyoung-Jun Moon, Dong-Shin Jo, Wan Kim, Michael Choi, Hyung-Jong
Ko, and Seung-Tak Ryu at the IEEE Journal of Solid-State Circuits published their
work regarding a pipeline SAR-ADC where their prototype has used current-mode
residue processing (Moon et al. 2019). This proposed state of the art used differ-
ent building blocks such as degenerated gm-cell as an open-loop residual amplifier;
the S/H circuit has been replaced by a switched-current mirror and use of a split-
current DAC for current-domain SAR conversion. There fabricated 28-nm proto-
type achieved a signal-to-noise dynamic range ratio (SNDR) of 56.6 dB with a figure
of merit of a 21.7-fJ/conversion step. The residue processing used in this prototype
enhanced the operation speed.
17.3.2 DECIMATOR
A decimator is a digital filter with more complexity which is responsible for downs-
ampling the oversampled output of the CTΔ∑M to the Nyquist rate. The purposes of
decimator are as follows:
Thus, increasing the resolution of complete CTΔ∑ ADC, Amrith Sukumaran and
Shanthi Pavan in their published journal at IEEE Journal of Solid-State Circuits in
2014 studied the use of finite impulse response (FIR) feedback DAC in single-bit
continuous-time delta–sigma ADCs. They proposed a method to stabilize a CTDSM
using an FIR feedback DAC (Sukumaran and Pavan 2014).
In 2010, in IEEE Journal of Solid-State Circuits, Gerry Taylor, Member, and Ian
Galton published a paper where they have proposed a CTDSM with mostly having
digital circuitry. The design was based on a voltage-controlled ring oscillator. It was
claimed to be free from an analog integrator, feedback DACs, comparators, refer-
ence voltages, and requirement of a low-jitter clock. Because most of the parts are
digital in this proposed prototype, the area of the design was claimed to be less than
that of a conventional CTDSM (Taylor and Galton 2010).
In October 2019, Antonios Nikas, Sreenivas Jambunathan, Leonhard Klein,
Matthias Voelker, and Maurits Ortmanns published a paper in IEEE Journal of
Solid-State Circuits where they proposed a CTDSM using a current-reuse DAC and
modified instrumentation amplifier (IA). The prototype used the modified IA as the
integrator for the delta–sigma modulator, whereas the current-reuse DAC was used
as the feedback circuit to that integrator. The authors claimed that the design con-
sumes 22% less power in 180-nm CMOS technology with a dynamic range of 90 dB
(Nikas et al. 2019).
In September 2019, a letter was published in IEEE Solid-State Circuits Letters by
John Bell and Michael P. Flynn proposing a multiband continuous-time delta–sigma
ADC. In the proposed prototype, a new class of continuous-time delta–sigma ADC
demonstrated two simultaneous bands such as baseband and bandpass. The total
added bandwidth of the ADC was 90 MHz while the bands were shown to be sepa-
rated by 500 MHz. The authors claimed that this prototype was first to be presented
by them (Bell, Member, and Flynn 2019).
In 2014, a paper by Astria Nur Irfansyah, Long Pham, Andrew Nicholson,
Torsten Lehmann, Julian Jenkins, and Tara Julia Hamilton entitled “Nauta OTA
in a Second-Order Continuous-Time Delta–Sigma Modulator” proposed the use of
Nauta OTA as an integrator in a second-order CTDSM. They studied the structure
for high-bandwidth operation and simple inverter-based structure (Irfansyah et al.
2014).
In 1992, a journal published in IEEE Journal of Solid-State Circuits by Bram
Nauta entitled “A CMOS Transconductance-C Filter Technique for Very High
Frequencies” first proposed the use of inverter-based tranconductors. He first studied
the work of differential amplifier–based integrator and then proposed how can an
improved result be achieved by inverter-based operational transconductance ampli-
fier (Nauta 1992).
17.4 CONCLUSION
ADC and DAC are part of SA transmitter and receiver. This chapter elaborates dif-
ferent types of ADCs and DACs, which can be used for SA applications. SA is very
much required to meet the growing demand of the users for faster communication
and networking. So choosing the appropriate ADC and DAC is also an important
A Review of ADC and DAC for Smart Antenna 307
task for the researchers considering and fulfilling all the design requirements of the
users and service providers. The future scope includes performance comparison;
further analysis can be done with different types of antenna arrays such as linear,
planar, circular, and so on and their performance comparison.
REFERENCES
Andersson, Sören, Mille Millnert, Mats Viberg, and Bo Wahlberg. 1991. A study of adap-
tive arrays for mobile communication systems. In Proceedings ICASSP 91: 1991
International Conference on Acoustics, Speech, and Signal Processing, 3289–92.
Awan, Adnan Anwar, Shahid Khattak, Aqdas Naveed Malik, and others. 2017. Performance
comparisons of fixed and adaptive beamforming techniques for 4G smart antennas. In
2017 International Conference on Communication, Computing and Digital Systems
(C-CODE), 17–20.
Balanis, Constantine A, and Panayiotis I Ioannides. 2007. “Introduction to Smart Antennas.”
Synthesis Lectures on Antennas 2 (1). Morgan & Claypool Publishers: 1–175.
Basha, T S Ghouse, P V Sridevi, and M N Giri Prasad. 2013. “Beam Forming in Smart
Antenna with Precise Direction of Arrival Estimation Using Improved MUSIC.”
Wireless Personal Communications 71 (2). Springer: 1353–64.
Bell, John, Student Member, and Michael P Flynn. 2019. “A Simultaneous Multiband
Continuous-Time ADC With 90-MHz Aggregate Bandwidth in 40-Nm CMOS” 2 (9):
91–94. doi:10.1109/LSSC.2019.2933159.
Boukalov, Adrian O, and S-G Haggman. 2000. “System Aspects of Smart-Antenna
Technology in Cellular Wireless Communications-an Overview.” IEEE Transactions
on Microwave Theory and Techniques 48 (6). IEEE: 919–29.
Butler, Jesse. 1961. “Beam-Forming Matrix Simplifies Design of Electronically Scanned
Antenna.” Electron. Design 9: 170–73.
Chryssomallis, Michael. 2000. “Smart Antennas.” IEEE Antennas and Propagation
Magazine 42 (3). IEEE: 129–36.
Consortium, International Engineering, and others. 2005. “Smart Antenna Systems.” http://
read.pudn.com/downloads157/doc/comm/699605/smart_antenna%20systems.pdf.
Craninckx, Jan, and Geert Van Der Plas. 2007. A 65fJ/conversion-step 0-to-50MS/s
0-to-0.7mW 9b charge-sharing SAR ADC in 90 nm digital CMOS. In Digest of
Technical Papers - IEEE International Solid-State Circuits Conference. doi:10.1109/
ISSCC.2007.373386.
Dingwall, Andrew G.F. 1979. “Monolithic Expandable 6 Bit 20 MHz CMOS/SOS A/D
Converter.” IEEE Journal of Solid-State Circuits 14 (6): 926–32. doi:10.1109/
JSSC.1979.1051299.
Draxelmayr, Dieter. 2003. “A 6b 600MHz 10mW ADC Array in Digital 90nm CMOS.” In
Digest of Technical Papers - IEEE International Solid-State Circuits Conference,
47:212–213+536. doi:10.1109/isscc.2004.1332695.
Fletcher, P N, and P Darwood. 1998. “Beamforming for Circular and Semicircular
Array Antennas for Low-Cost Wireless Lan Data Communications Systems.” IEE
Proceedings-Microwaves, Antennas and Propagation 145 (2). IET: 153–157.
Ganguli
Ganguly, Saurav, Jayanta Ghosh, and Puli Kishore Kumar. 2019. “Performance Analysis of
Array Signal Processing Algorithms for Adaptive Beamforming.” Simulation 2: 15.
Godara, Lal Chand. 2004. Smart Antennas. Boca Raton, FL: CRC press.
Greda, Lukasz A, Andreas Winterstein, Daniel L Lemes, and Marcos V T Heckler. 2019.
“Beamsteering and Beamshaping Using a Linear Antenna Array Based on Particle
Swarm Optimization.” IEEE Access, 7: 141562–73.
308 Advanced VLSI Design and Testability Issues
Gross, F B. 2005. Smart Antenna for Wireless Communication. New York: Graw-Hill. Inc.
Harpe, Pieter, Eugenio Cantatore, and Arthur Van Roermund. 2013. “A 10b/12b 40 KS/s
SAR ADC with Data-Driven Noise Reduction Achieving up to 10.1b ENOB at 2.2
FJ/Conversion-Step.” IEEE Journal of Solid-State Circuits 48 (12). IEEE: 3011–18.
doi:10.1109/JSSC.2013.2278471.
Hieu, Bui Van, Seunghwan Choi, Jongkug Seon, Youngcheol Oh, Chongdae Park, Jaehyoun
Park, Hyunwook Kim, and Taikyeong Jeong. 2011. A new approach to thermometer-
to-binary encoder of flash ADCs- bubble error detection circuit. In 2011 IEEE 54th
International Midwest Symposium on Circuits and Systems (MWSCAS), 1–4. IEEE.
doi:10.1109/MWSCAS.2011.6026403.
Irfansyah, Astria Nur, Long Pham, Andrew Nicholson, Torsten Lehmann, Julian Jenkins,
and Tara Julia Hamilton. 2014. Nauta OTA in a second-order continuous-time delta-
sigma modulator. In Midwest Symposium on Circuits and Systems, 849–52. Institute of
Electrical and Electronics Engineers Inc. doi:10.1109/MWSCAS.2014.6908548.
Kohno, Ryuji. 1998. “Spatial and Temporal Communication Theory Using Adaptive Antenna
Array.” IEEE Personal Communications 5 (1). IEEE: 28–35.
Kumar, Pradeep, and Amit Kolhe. 2011. “Design & Implementation of Low Power 3-Bit Flash
ADC in 0. 18μm CMOS.” International Journal of Soft Computing and Engineering
1 (5): 71–74.
Lakshmi, T S Jyothi, Sandeep Sivvam, and V Rajyalakshmi. 2018. Performance evaluation
of smart antennas using non blind adaptive algorithms. In 2018 Conference on Signal
Processing and Communication Engineering Systems (SPACES), 66–71.
Lee, Chun C., and Michael P. Flynn. 2010. “A 12b 50MS/s 3.5mW SAR Assisted 2-Stage
Pipeline ADC.” IEEE Symposium on VLSI Circuits, Digest of Technical Papers. IEEE,
239–40. doi:10.1109/VLSIC.2010.5560243.
Lehne, Per H. 1999. “An Overview of Smart Antenna Technology for Mobile Communications
Systems.” IEEE Communications Surveys, Fourth Quarter 2 (4).
Lewis, Stephen H., and Paul R. Gray. 1987. “A Pipelined 5-Msample/s 9-Bit Analog-to-
Digital Converter.” IEEE Journal of Solid-State Circuits 22 (6): 954–61. doi:10.1109/
JSSC.1987.1052843.
Lewis, Stephen H, H Scott Fetterman, George F Gross, R Ramachandran, T R Viswanathan,
and Senior Member. 1992. “A 10-b 20-Msample / s Analog-to-Digital Converter.”
IEEE Journal of Solid-State Circuits 27 (3): 351–58.
Liberti, Joseph C, and Theodore S Rappaport. 1999. Smart Antennas for Wireless
Communications: IS-95 and Third Generation CDMA Applications. Upper Saddle
River, NJ: Prentice Hall PTR.
Matsuura, Tatsuji. 2016. “Recent Progress on CMOS Successive Approximation ADCs.” IEEJ
Transactions on Electrical and Electronic Engineering 11 (5): 535–48. doi:10.1002/
tee.22290.
Mccreary, James L, and P R Gray. n.d. [1975] “All-MOS Charge Redistribution Analog-to-
Digital Conversion Techniques.” IEEE Journal of Solid-State Circuits, SC-10 (12): 371–79.
Moon, Kyoung-Jun, Dong-Shin Jo, Wan Kim, Michael Choi, Hyung-Jong Ko, and Seung-Tak
Ryu. 2019. “A 9.1-ENOB 6-MW 10-Bit 500-MS/s Pipelined-SAR ADC with Current-
Mode Residue Processing in 28-Nm CMOS.” IEEE Journal of Solid-State Circuits 54
(9). IEEE: 2532–42. doi:10.1109/jssc.2019.2926648.
Murch, Ross D, and K Ben Letaief. 2002. “Antenna Systems for Broadband Wireless Access.”
IEEE Communications Magazine 40 (4). IEEE: 76–83.
Nauta Bram. 1992. “A CMOS Transconductance-C Filter Technique for Very High Frequencies.”
IEEE Journal of Solid-State Circuits 27 (2): 142–53.
A Review of ADC and DAC for Smart Antenna 309
Nikas, Antonios, Sreenivas Jambunathan, Leonhard Klein, Matthias Voelker, and Maurits
Ortmanns. 2019. “A Continuous-Time Delta-Sigma Modulator Using a Modified
Instrumentation Amplifier and Current Reuse DAC for Neural Recording.” IEEE
Journal of Solid-State Circuits 54 (10). IEEE: 2879–91. doi:10.1109/jssc.2019.2931811.
Patel, Dhaval N, B J Makwana, and P B Parmar. 2016. Comparative analysis of adaptive beam-
forming algorithm LMS, SMI and RLS for ULA smart antenna. In 2016 International
Conference on Communication and Signal Processing (ICCSP), 1029–33.
Poormohammad, Sarah, and Forouhar Farzaneh. 2018. “Proposed 2D and 3D Geometries
Intended for Smart Antenna Applications, Including Direction Finding and
Beamforming Implementation.” IET Radar, Sonar & Navigation 13 (5). IET: 673–81.
Pradhan, Tuhinansu. 2017. “Design of a low-voltage low power dynamic latch comparator for
a 1.2-V 0.4-mW CT delta sigma modulator with 41-dBm SNDR”. In 2017 International
Conference on Trends in Electronics and Informatics (ICEI) (pp. 815–20). IEEE.
Pradhan, T. 2017. Implementation of a low power second order continuous time delta sigma
modulator in 0.09μm technology using Cadence. M. Tech Thesis, KIIT Deemed to be
University.
Rappaport, Theodore S. 1998. Smart Antennas: Adaptive Arrays, Algorithms, & Wireless
Position Location. Piscataway, NJ: Institute of Electrical and Electronics Engineers Inc.
Shelton, J, and K Kelleher. 1961. “Multiple Beams from Linear Arrays.” IRE Transactions on
Antennas and Propagation 9 (2). IEEE: 154–61.
Shikata, Akira, Ryota Sekimoto, Tadahiro Kuroda, and Hiroki Ishikuro. 2012. “A 0.5 v
1.1 MS/Sec 6.3 FJ/Conversion-Step SAR-ADC with Tri-Level Comparator in 40 Nm
CMOS.” IEEE Journal of Solid-State Circuits 47 (4). IEEE: 1022–30. doi:10.1109/
JSSC.2012.2185352.
Shivapanchakshari, T G, and H S Aravinda. 2019. An efficient mechanism to improve the
complexity and system performance in OFDM using switched beam smart antenna
(SSA). Computer Science On-Line Conference, 60–66.
Sukumaran, Amrith, and Shanthi Pavan. 2014. “Low Power Design Techniques for Single-Bit
Audio Continuous-Time Delta Sigma ADCs Using FIR Feedback.” IEEE Journal of
Solid-State Circuits 49 (11). IEEE: 2515–25. doi:10.1109/JSSC.2014.2332885.
Taylor, Gerry, and Ian Galton. 2010. “A Mostly-Digital Variable-Rate Continuous-Time
Delta-Sigma Modulator ADC.” IEEE Journal of Solid-State Circuits 45 (12). IEEE:
2634–46. doi:10.1109/JSSC.2010.2073193.
Tsai, Jiann-An, and Brian D Woerner. 2001. Adaptive beamforming of uniform circular arrays
(UCA) for wireless CDMA system. In Conference Record of Thirty-Fifth Asilomar
Conference on Signals, Systems and Computers (Cat. No. 01CH37256), 1: 399–403.
Tsoulos, George V. 2000. Adaptive Antennas for Wireless Communications. Wiley-IEEE Press.
Tsoulos, George V, Georgia E Athanasiadou, Mark A Beach, and Simon C Swales. 1998.
“Adaptive Antennas for Microcellular and Mixed Cell Environments with DS-CDMA.”
Wireless Personal Communications 7 (2–3). Springer: 147–69.
Varghese, George Tom, and K. K. Mahapatra. 2012. A high speed low power encoder for a 5
bit flash ADC.” In 2012 International Conference on Green Technologies, ICGT 2012,
41–5. doi:10.1109/ICGT.2012.6477945.
Wang, Xiaodong, Changlin Chen, and Weidong Jiang. 2018. Implementation of real-time
LCMV adaptive digital beamforming technology. In 2018 International Conference
on Electronics Technology (ICET), 134–7.
Yip, Marcus, and Anantha P. Chandrakasan. 2011. A resolution-reconfigurable 5-to-10b
0.4-to-1V power scalable SAR ADC. Digest of Technical Papers - IEEE International
Solid-State Circuits Conference. IEEE, 190–1. doi:10.1109/ISSCC.2011.5746277.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
18 Active Inductor–Based
VCO for Wireless
Communication
Aditya Kumar Hota
Veer Surendra Sai University of Technology
Kabiraj Sethi
Veer Surendra Sai University of Technology
CONTENTS
18.1 Introduction ................................................................................................ 311
18.1.1 Inductor–Capacitor Voltage-Controlled Oscillator ...................... 312
18.1.1.1 Linear Feedback Approach .......................................... 312
18.1.1.2 Cross-Coupled Approach ............................................. 313
18.2 Ring Voltage-Controlled Oscillator............................................................ 314
18.3 Active Inductor ........................................................................................... 314
18.4 Voltage-Controlled Oscillator Using Active Inductor ................................ 315
18.5 Discrete Fourier Transform for Voltage-Controlled Oscillator .................. 319
18.6 Summary .................................................................................................... 320
References .............................................................................................................. 322
18.1 INTRODUCTION
The recent wireless communication systems require very wide-band voltage-
controlled oscillators (VCOs) to be integrated with other multimode, multiband sys-
tems, which is very difficult to achieve with a passive, untuned inductor-based VCO.
The gigabit-per-second (Gbps) serial links demand small-width clock pulses, which
are fulfilled by the ring VCOs. In fiber optic communication, maintaining the phase
noise of the oscillator is a very challenging task. So the requirement is a very-high-
Q-value VCO to keep the phase noise low. This can be done by designing a high-Q
tank circuit having high-Q inductor or capacitor. In millimeter wave application, it
is not possible to generate a fundamental frequency; rather the higher-order harmon-
ics are required for that purpose. So heterojunction bipolar transistor (HBT)–based
311
312 Advanced VLSI Design and Testability Issues
A(Vosc , ω)
Vo = V (18.1)
1 − A(Vosc , ω)β(Vosc , ω) in
Vin Vout
A
The close loop gain should be infinite to get an output signal out of nothing. The
broadband noise from the active and passive components of the circuit upon biasing
causes the oscillation to start. Initially, Vin = 0, and the gain is independent of the
voltage, but gradually Vin increases to Vosc where the gain, as well as a function of
input voltage, becomes linear. The conditions for sustained oscillation are as follows:
the loop gain phase shift should be 180o, and magnitude should be greater than 1.
Hartley oscillator, Armstrong oscillator [1], Colpitts oscillator [2], and Clapp oscil-
lator [3] fall under this category.
Figure 18.2 shows a low-power Colpitts VCO [2]. It employs the current-reused
series–shunt feedback topology, which takes care of the power requirement of the
VCO. The Q-factor of the circuit is improved by the inversion metal oxide semicon-
ductor (MOS) varactor (enhanced varactor) along with the metal–insulator–metal
(MIM) capacitor and so the improvement of noise performance is there.
Vg
Rg VDD
VDD
IDD
VDD
L1
P S S P
L1 VDD CENTRE-TAP
TRANSFORMER OUTP
OUTM- TRANSFORMER
Rg2
M7 M8
M6
C3
C3
M5
Enhanced
Varactor
Vc
C1 C2
M1 M2
A B H
1 2 N
Gm1 Gm1
V2 V2
Vin 2 Vin 2
V1 V1
1
Y 1
C1 2 Y
1
Gm2 Gm2
(a) (b)
FIGURE 18.4 (a) Lossless AI and (b) lossy AI. AI, active inductor.
floating AI, none of the terminals are connected to either power supply or ground.
Looking into port 2, the admittance is
I in 1
Y= = (18.2)
Vin ⎛ C1 ⎞
S⎜
⎝ Gm1Gm 2 ⎟⎠
⎛ C1 ⎞
which behaves as an inductor having inductance value L = ⎜⎜ ⎟ , i.e., the
⎟
⎝ Gm 1Gm 2 ⎠
inductance is directly proportional to the load capacitance and inversely propor-
tional to the product of transconductances. This gyrator-C AI is inductive over all
the frequency spectra. The equivalent circuit of the lossy single-ended gyrator-C AI
is shown in Figure 18.5. The AI has many advantages over the passive counterpart,
such as tunability, high quality factor, and small in size. But it has also the linearity
issue along with noise-producing factor, which makes the designer to design care-
fully to overcome these issues.
Vin
Iin
LS
GP CP
GS
Vdd
Mp I1
n bits Control
Inverter
Module
M2 M1
Vbias
IN
Vctrl
Monitor Vcntrl
Interface I2
Cvar
FIGURE 18.6 Reconfigurable VCO with AI. AI, active inductor; VCO, voltage-controlled
oscillator.
Vdd
Vdd
InV1
n bits
Vctrl M4
I1 Mp InV2
InV3
M3
out
in
InVn M2
M1 M2
Vbias IN
Vctrl Vctrl M1
Vcntrl Monitoring
I2 Interface
Cvar
(a) (b)
FIGURE 18.7 (a) Active inductor structure and (b) inverter structure.
Active Inductor–Based VCO 317
⎛ f ⎞
FOM = L ( Δf ) + 10 log ( Pdis ( mW )) − 20 log ⎜ osc ⎟ (18.3)
⎝ foffset ⎠
The inverter is designed by domino logic as per Figure 18.7b. The tuning range of
this VCO is from 1.22 to 2.6 GHz with a very low power consumption of 4 mW and
die area of .0031 μm2. The VCO shown in Figure 18.8 is based on a PMOS cross-
coupled pair of transistor core, connected to the supply through a current source to
avoid the supply noise, minimizing the phase noise [6]. The two inductors L1 and
L2 are realized by the AI. The tail current sources for the AI are implemented by
two transistors. The AI has a resistive feedback between the two transconductors
to increase the quality factor of the AI, thus reducing the phase noise of the VCO.
Here with this design, there is a 94% of tuning range from 120 MHz to 2 GHz with
a phase noise variation of −80 to −90dbc Hz at a frequency offset of 1 MHz. The
power consumption of the VCO is 7 mW with a 2.2 mW for the AI from a supply of
1.8 V. The VCO with same concept is adopted in Ref. [7]. The only difference is that
here the AI uses an extra additive capacitor Ca at node 1 and the transistors operate
in subthreshold region instead of saturation, thus reducing the power consumption.
The figure of merit proposed in this work is
⎛ f F ⎞
FOM T = L ( Δf ) − 20 log ⎜ 0 T ⎟ + 10 (18.4)
⎝ Δf 10 ⎠
where L (Δf) is the phase noise, FT is the frequency tuning range, and Pdis is the
power dissipation. This VCO can be used in low-power, multiband, and wide-tuning
applications.
An ultrawideband CMOS (complementary metal oxide semiconductor) Hartley
VCO is designed as per Figure 18.9a, consisting both common source stages as
negative transconductor and pi-feedback through two parallel AIs in series with a
capacitor [8]. The AI in Figure 18.9b uses the active resistor in the feedback path.
The tuning range is 3.8–7.4 GHz with a high-quality factor of 90 and a little bit high
Vdd
I
Vdd Vdd
I1 M1 I1
M2
Rf 1 1 Rf
Mp1 V1 Mp1
V2
Mn2 Mn2
Va Va
C
Mn1 Mn1
Vb Vb
I2 I2
Active
Inductor
VDD
VDD
R1 L R2
C2 C1 I1
Rf
C3 MN5
MN4 C5
Vs MN11
Vc MN6
Vb Vtune
GND MN13
Vo2 C4 Vo1 GND
MN12 Vin
L1 L2 I2
GND
(a) (b)
FIGURE 18.9 (a) Hartley VCO and (b) active inductor. VCO, voltage-controlled oscillator.
MP2
MN4
MP9
Vin-
VDD
C1
V4
V2
MN9
Vb2
MP8
MN2
MN4
MN11
MP7
MP4
VDD
C3
MP2
V6
MP4
Vtune2
R2
I2
MN7
Vr2
MN8
I3
Mr1 Mr2
VDD
Vout-
Vout+
MN7
Vr1
MN6
R1
I1
MP3
Vtune1
MP1
V5
C4
VDD
MP3
MN1
MN9
MN6
MN1
Vb1
MP5
MP1
MN3
VDD
V1
V3
C2
MN3
Vin+
MP4
FIGURE 18.10 VCO with differential AI. AI, active inductor; VCO, voltage-controlled
oscillator.
power dissipation of 29.1 mW. The phase noise offered by this VCO is −92.05 dBc/
Hz. Better phase noise and VCO gain can be obtained by varying the biasing condi-
tion and by the use of active load. A LC VCO with differential AI [9] with a balance
between the phase noise and the power dissipation is designed in TSMC 180-nm
process as in Figure 18.10. The differential AI is designed with a current-mirror
feedback networks and a cascode structure as shown in the dotted box. The VCO
hunts a 13.6 mW power for a supply of 1.8 V. The frequency tuning ranges from 1.126
to 2.713 GHz, with a phase noise of −117.2 dBc/Hz at 1-MHz offset frequency. So
far, we have discussed the AI-based LC VCO that includes the LC and cross-coupled
Active Inductor–Based VCO 319
VDD
VDD
MP3 MP4
AI AI
MP5
OUT - OUT+
MN1 MN2
in+ In-- Mi
I2
(a) (b)
FIGURE 18.11 (a) Cross-coupled delay cell and (b) active Inductor.
VCOs. Now the circuit appeared in Figure 18.11a is a cross-coupled delay cell with
AI, as shown in Figure 18.11b, as load which is used to design a fully differential
VCO [10]. The VCO has eight stages and tested with resistive and current source load
and found smaller delay as compared with its passive counterpart. The AI is a self-
biased one, and the VCO provides full rail-to-rail voltage swing without any loss.
But there is a reduction in frequency tuning range of the VCOs to 0–1.1 GHz from
1.5 to 2.2 GHz with AI as load instead of PMOS transistor load.
TEST
SIGNATURE TRANSMITTER
GENERATOR
LOOP
BACK
CIRCUIT
SIGNATURE
RESPONSE RECEIVER
EVALUATOR
FIGURE 18.12 Loopback technique for testing RFIC. RFIC, radio frequency integrated
circuit.
and the baseband analog-to-digital converter. The test pattern is generated by a DSP,
and Gaussian minimum shift keying is used for the testing of GSM transceiver. A
BIST PLL technique is presented for the testing of VCO in Ref. [17] where no addi-
tional circuitry is required for the test pattern generation, but the evaluation circuitry
consists of the XOR networks, DRAM, and OR gate which monitor the unintended
frequency of oscillation due to internal faults rather the voltage of VCO. Ten num-
bers of catastrophic fault models for the VCO were generated and tested as
• gate-to-source short
• gate-to-drain short
• drain-to-source short
• gate open
• drain open
• source open
• resistor open
• resistor short
• capacitance short
• capacitance open
A discrete Fourier transform for LC VCO based on a single test signal is presented in
Ref. [18]. The fault model may be hard, i.e., due to shorts and opens in the circuit, or
soft, i.e., due to parametric fault and corner states. Here, sense transistors are intro-
duced along with the VCO to sense the fault. Under fault-free case, the current value
of the sense transistor was measured, and in the presence of the fault, it is observed
that there is a drop in the current value in the sensing transistor. This fault is reflected
as a compression in the VCO output waveform as well as a single-bit digital output
signal.
18.6 SUMMARY
This chapter emphasis on the different types of VOCs and their use in various fields.
The ring VCOs and digitally controlled oscillator are good for the optical fiber com-
munication and for the high-speed data rate applications, where there is a very high
TABLE 18.1
Comparison of AI-Based VCOs
[22]
Unit [19] [20] [21] With AC Without AC [8] [5] [6]
Active Inductor–Based VCO
Tech. μm 0.18 CMOS 0.18 SiGe 0.18 SiGe 0.18 CMOS 0.18 CMOS 0.18 CMOS 0.13 CMOS 0.18 CMOS
BiCMOS BiCMOS
VDD V 1.8 2.5 3.3 1.8 1.8 1.8 1.1 1.8
Tuning range GHz 0.5–3.0 2–15 8–17 1.13–2.67 1.3–3.24 3.80–7.40 1.22–2.6 0.1–2
% 143 – 72 – – – 72 94
Kvco MHz V 2500 – – – – 2400 – 2350
DC power mW 6–2.8 88–120 75–115 2.2–13 2.2–13 29.1 3.6–4.3 7
P0 dBm −14 to −22 −8 to −6 −15 to 12 – – – – −4.7 to 1.5
L (Δf) dBc/Hz −101 to 118 −112 – −82.8 to −92.2 −74.2 to −88.6 −75.42 to −82 to −87 −80 to −90
(10 MHz) −92.05
Active area mm2 0.15 × 0.3 0.67 × 0.25 0.65 0.171 × 0.171 0.171 × 0.171 – 0.0031 –
FOMT dBc Hz – – – −168 −166.3 – −151 –
AC, additive capacitor; AI, active inductor; FOM, figure of merit; VCO, voltage-controlled oscillator; CMOS, complementary metal oxide semiconductor; BiCMOS,
bipolar CMOS.
321
322 Advanced VLSI Design and Testability Issues
tuning range requirement and the phase noise can be tolerated to some extent. On
the other hand, the LC VCOs and the cross-coupled VCOs are on high demand for
the wireless communication systems such as Bluetooth, ZigBee, Wi-Fi, and so on.
As we discussed that the LC VCOs' tuning range can be enhanced by the implica-
tion of the AI with proper design, maintaining the phase noise least effected, such
designs are now adopted in industries with technology compatibilities. A comparison
of performance parameters of the various AI-based VCOs is presented in Table 18.1,
and the results are up to the industry standard. Also, a clear picture of the testing of
the wireless transceiver ICs is presented here, the DUT being VCO and PLL, as they
are the key modules for the signal generation in any RFIC, mostly using the loopback
techniques.
REFERENCES
1. C. C. Hsiao, C. W. Kuo, C. C. Ho, and Y. J. Chan, “Improved quality-factor of
0.18-μm CMOS active inductor by a feedback resistance design,” IEEE Microw. Wirel.
Components Lett., vol. 12, no. 12, pp. 467–469, 2002.
2. T. P. Wang, “A K-band low-power colpitts VCO with voltage-to-current positive-feed-
back network in 0.18 μm CMOS,” IEEE Microw. Wirel. Components Lett., vol. 21, no.
4, pp. 218–220, 2011.
3. N. Pohl, H. M. Rein, T. Musch, K. Aufinger, and J. Hausner, An 80 GHz SiGe bipo-
lar VCO with wide tuning range using two simultaneously tuned varactor pairs. 2008
IEEE CSIC Symp. GaAs ICs Celebr. 30 Years Monterey, Tech. Dig. 2008, pp. 1–4,
2008.
4. F. Yuan, CMOS active inductors and transformers, Switzerland. 2008.
5. F. Haddad, I. Ghorbel, and W. Rahajandraibe, “Design of reconfigurable inductorless
RF VCO in 130 nm CMOS,” Bionanoscience, vol. 9, no. 2, pp. 285–295, 2019.
6. H. B. Kia and A. K. A’ain, “A wide tuning range voltage controlled oscillator with a
high tunable active inductor,” Wirel. Pers. Commun., vol. 79, no. 1, pp. 31–41, 2014.
7. Y.-J. Jeong, Y.-M. Kim, H.-J. Chang, and T.-Y. Yun, “Low-power CMOS VCO with a
low-current, high-Q active inductor,” IET Microwaves, Antennas Propag., vol. 6, no. 7,
p. 788, 2012.
8. M. Mehrabian, A. Nabavi, and N. Rashidi, A 4~7GHz ultra wideband VCO with tun-
able active inductor. In Proceedings of The 2008 IEEE International Conference on
Ultra-Wideband, ICUWB 2008, vol. 2, pp. 21–24, 2008.
9. Y. Zhang et al., A Novel LC VCO with high output power and low phase noise using
differential active inductor. In 2018 IEEE 3rd International Conference on Integrated
Circuits and Microsystems, ICICM 2018, pp. 90–93, 2018.
10. F. Yuan, A fully differential VCO cell with active inductor load for GBPS serial links.
3rd Int. IEEE Northeast Work. Circuits Syst. Conf. NEWCAS 2005, vol. 2005, pp.
183–186, 2005.
11. D. Lupea, U. Pursche, and H. J. Jentschel, RF-BIST: Loopback spectral signature anal-
ysis. in Proceedings -Design, Automation and Test in Europe, pp. 478–483, 2003.
12. M. Jarwala, D. Le, and M. S. Heutmaker, End-to-end test strategy for wireless systems.
IEEE Int. Test Conf., pp. 940–946, 1995.
13. A. Halder, S. Bhattacharya, G. Srinivasan, and A. Chatterjee, A system-level alternate
test approach for specification test of RF transceivers in loopback mode. In Proceedings
of the IEEE International Conference on VLSI Design, no. 2003, pp. 289–294, 2005.
Active Inductor–Based VCO 323
14. G. Srinivasan, F. Taenzler, and A. Chatterjee, “Loopback DFT for low-cost test of
single-VCO-based wireless transceivers,” IEEE Des. Test Comput., vol. 25, no. 2,
pp. 150–159, 2008.
15. S. Ozev and C. Olgaard, Wafer-level RF test and DfT for VCO modulating transceiver
architecures. Proc. IEEE VLSI Test Symp., no. Vts, pp. 217–222, 2004.
16. G. Srinivasan, A. Chatterjee, and F. Taenzler, Alternate loop-back diagnostic tests for
wafer-level diagnosis of modern wireless transceivers using spectral signatures. In
Proceedings of the IEEE VLSI Test Symposium, vol. 2006, pp. 222–227, 2006.
17. I. Toihria, R. Ayadi, and M. Masmoudi, “High performance BIST PLL approach for
VCO testing,” 2014, International Conference on Advanced Technologies for Signal
and Image Processing (ATSIP), pp. 517–522.
18. L. Dermentzoglou, Y. Tsiatouhas, and A. Arapoyanni, “A design for testability scheme
for CMOS LC-tank voltage controlled oscillators,” J. Electron. Test. Theory Appl.,
vol. 20, no. 2, pp. 133–142, 2004.
19. L. H. Lu, H. H. Hsieh, and Y. Te Liao, “A wide tuning-range CMOS VCO with a dif-
ferential tunable active inductor,” IEEE Trans. Microw. Theory Tech., vol. 54, no. 9,
pp. 3462–3468, 2006.
20. T. Kanar and G. M. Rebeiz, “A 2–15 GHz VCO with harmonic cancellation for wide-
band systems,” IEEE Microw. Wirel. Components Lett., vol. 26, no. 11, pp. 933–935,
2016.
21. S. Shankar, S. Horst, and J. D. Cressler, Frequency- and amplitude-tunable X-to-Ku
band SiGe ring oscillators for multiband BIST applications. In Proceedings of the IEEE
Bipolar/BiCMOS Circuits and Technology Meeting, pp. 9–12, 2010.
22. J. Shim and J. Jeong, “A band-selective low-noise amplifier using an improved tunable
active inductor for 3–5 GHz UWB receivers,” Microelectron. J., vol. 65, no. January,
pp. 78–83, 2017.
Taylor & Francis
Taylor & Francis Group
http://taylorandfrancis.com
19 Fault Simulation
Algorithms: Verilog
Implementation
Sobhit Saxena
Lovely Professional University
CONTENTS
19.1 Introduction ................................................................................................ 325
19.2 Logic Simulation ........................................................................................ 325
19.3 Fault Simulation .......................................................................................... 327
19.4 Verilog Coding for Simulation ................................................................... 330
References .............................................................................................................. 338
19.1 INTRODUCTION
Simulation is a technique normally used for functionality verification during circuit
designing before fabrication. Fault is the effect of any defect created in the hardware
of the circuit during fabrication, which can cause the circuit malfunction.
Every hardware unit of the circuit needs to be tested in the production house,
and segregation between faulty and fault-free units is done before it is out for cus-
tomer use. Testing of hardware units can be done by applying all possible inputs, and
responses are recorded and matched with the expected results. This process requires
a lot of time. In order to reduce the time and also to identify the type of fault with the
location of the fault in the circuit, instead of all possible inputs, test vectors/patterns
need to be applied for testing. To identify these test vectors/patterns, simulation exer-
cises are required for every possible input to check their effectiveness in testing and
identification of the possible faults in the circuit. This type of simulation is known
as fault simulation.
Therefore, the simulation is classified into two broad categories: logic simulation
and fault simulation as depicted from the block diagram shown in Figure 19.1.
325
326 Advanced VLSI Design and Testability Issues
Among these, deductive fault simulation technique can be faster than the parallel
fault simulation but is slower than the concurrent fault simulation because concurrent
fault simulation simulates only the fault-affected parts of the circuits. However, dif-
ferential fault simulation is expected to be 12 times faster than the concurrent fault
simulation. The memory requirement of the concurrent fault simulation is quite high
than the deductive fault simulation because of the I/O values of every bad gates avail-
able in the target circuit. The advantage of concurrent fault simulation is that it can be
used for functional models and it can perform the sequential fault simulation with ease.
reduce the fault simulation time. To realize this bitwise parallelism, there are
two ways: one is parallel fault simulation and another one is parallel pattern
simulation. Assuming that binary logic is used, one bit is sufficient to store the
logic value of the signal. It means that the computer is using n-bit wide data
words; each signal is connected with a data word of which n − 1 bits are issued
for n − 1 faulty circuits and the remaining bit is reserved for fault-free circuit.
According to this, n − 1 faulty circuits and one fault-free circuit are processed
in parallel using bitwise logic operation that corresponds to a speedup factor of
approximately n − 1 compared with serial fault simulation. A fault is detected
if the bit value is different from the fault-free circuit at any way of the outputs.
c. Deductive Fault Simulation:
Deductive fault simulation is very different type of simulation in which
no fault injection is required. In this simulation, for a particular input pat-
tern, the possible excited faults at primary inputs are identified. The propa-
gation of these faults depends upon the passing gate. Faults that can be
excited at the output of that gate will be added to the list. The overall faults
that are propagated at the output can be detected by that input pattern
and become the test vector for the propagated faults. The same process is
repeated for different input patterns till all possible faults got propagated
at the output of the circuit. This deductive fault simulation can be very fast
because only fault-free simulations have to be performed [3].
d. Concurrent Fault Simulation:
Out of the available fault simulation techniques, the concurrent fault
simulation technique is best suited for the combinational and the sequential
circuits. Concurrent fault simulation is an event-driven simulation in which
both the fault-free and the faulty circuits are simulated together. Concurrent
fault simulation works by simulating the differential parts of the circuit that
are affected by the fault (Figure 19.3).
In concurrent fault simulation, every gate has a set of bad gates, and
these bad gates are basically the gates with I/O values in the presence of the
respective faults. In concurrent fault simulation, the bad gate and the good
gate are simulated concurrently, and then it is checked whether the fault is
propagated to the output or not. If the fault is visible at the output, then it
can be observed that the fault is propagated to the output for the given test
vector, which is called a good event. For some test vectors, the fault is not
visible at the output, which is called a bad event [4].
Advantages
1. Concurrent fault simulation is faster because it simulates the differen-
tial parts of the circuit.
2. Concurrent fault simulation can deal with multivalued simulations.
3. It can perform sequential simulation with ease.
4. It is suited for both the combinational and the sequential circuits.
The flow of the concurrent fault simulation is given in flowchart in which
there is a collapsed fault list, which consists of list of the faults and the test
Fault Simulation Algorithms 329
vector set that is given to simulate the fault-free and the faulty circuits.
These fault-free and the faulty gates are evaluated, and then the output
response of the circuit is compared with the fault-free circuit. If any fault
is detected, then that is deleted from the collapsed fault list. This process
will be continued until all the faults are covered in the collapsed fault list.
e. Differential Fault Simulation:
Differential fault simulation is the combination of concurrent fault simu-
lation and the single fault simulation technique for sequential fault simula-
tion [6]. Initially, for first input pattern, fault-free circuit simulation is done,
and the output is stored; then, first fault is injected in the circuit, and the
simulation is occurred. The difference between the states (of memory ele-
ments, e.g., flip-flops) of fault-free and faulty circuit is stored. Then, first
fault is removed and second fault is inserted. and the difference between
the states of faulty circuit with first fault and faulty circuit with second fault
is stored. This process is repeated till for all faults get simulated for first
input pattern. For the next input pattern, the state of the fault-free circuit
is first restored, and then pattern is applied. After simulation, the output is
stored. The state of faulty circuit with first fault is restored by inserting the
difference between the states of fault-free and faulty circuit with first input
pattern. After that, first fault is injected as an event, and the process contin-
ues as done for first input pattern. The state of every circuit is restored from
the last simulation. Those faults that are detected are dropped from the list.
TABLE 19.1
Truth Table of 2 × 1 Multiplexer
a b c y
0 0 0 0
0 0 1 0
0 1 0 0
0 1 1 1
1 0 0 1
1 0 1 1
1 1 0 0
1 1 1 1
the value of x-faulty and y-faulty circuits’ output responses (Figure 19.5).
x and y are two faults, one is “a” stuck at “1” and another one is “g” stuck
at “0.” To make a stuck at “1,” insert extra OR gate with one signal input
“1.” Similarly, for stuck at “0,” insert extra AND gate with one signal input
“0” (Figure 19.6).
While simulation, one fault is considered for one time. It means while
propagating the x-fault, y-fault is not considered. Similarly, according to
that, while propagating y-fault, x-fault is not considered. While propagat-
ing the x-faulty (Fx) circuit, output response is different from the fault-free
(FF) circuit’s output response. It means that x-fault is detected and “x” has
become high at output response. Similarly, while propagating the y-faulty
(Fy) circuit, output response is different from the fault-free circuit’s output
332 Advanced VLSI Design and Testability Issues
TABLE 19.2
Truth Table of Fault-Free and Faulty Circuits
Inputs Outputs
A B C FF Fx Fy x y
0 0 0 0 0 0 0 0
0 0 1 1 1 0 0 1
0 1 0 0 1 0 1 0
0 1 1 0 1 0 1 0
1 0 0 0 0 0 0 0
1 0 1 1 1 0 0 1
1 1 0 1 1 1 0 0
1 1 1 1 1 1 0 0
response. It means that the y-fault is detected and “y” has become high at
output response, which has been shown in Table 19.2.
In Table 19.2, a, b, and c are inputs to the circuit. FF, Fx, and Fy are
the fault-free, x-fault, and y-fault output responses. If Fx and FF output
responses are different, then “x” is high at output response. Similarly, if Fy
and FF output responses are different, then “y” is high at output response.
Verilog Code
module prall(a,b,c,f,x,y,fx,fy);
input a,b,c;
output f,x,y,fx,fy;
wire [8:0]w;
not (w[0],b);
and af1(w[1],a,b);
and af2(w[2],c,w[0]);
or af3(w[3],w[1],w[2]);
// fault-free circuit Programming //
or ax0(w[4],1,a);
and ax1(w[5],b,w[4]);
or ox2(w[6],w[2],w[5]);
// x-faulty circuit Programming //
Fault Simulation Algorithms 333
and oy0(w[7],0,w[2]);
or oy1(w[8],w[7],w[1]);
// y-faulty circuit programming //
assign fx=w[6];
assign fy=w[8];
assign f=w[3];
xor (x,w[6],w[3]);
xor (y,w[8],w[3]);
endmodule
Test Bench
module prall_tb();
reg a,b,c;
wire f,x,y;
wire fx,fy;
prall u(a,b,c,f,x,y,fx,fy);
initial
begin
a=0;b=0;c=0;
#5 a=0;b=0;c=1;
#5 a=0;b=1;c=0;
#5 a=0;b=1;c=1;
#5 a=1;b=0;c=0;
#5 a=1;b=0;c=1;
#5 a=1;b=1;c=0;
#5 a=1;b=1;c=1;
#10 $stop;
end
endmodule
Simulation Results:
Figure 19.7 shows that y-fault is detected. When [a, b, c] are equal to
[0, 0, 1],“FF” and “Fy” output responses are different. So that the y output
response is high (Figures 19.7 and 19.8).
Figure 19.8 shows that x-fault is detected. When [a, b, c] are equal to
[0, 1, 0], “FF” and “Fx” output responses are different. So that the x output
response is high.
Concurrent Fault Simulation:
To analyze the concurrent fault simulation, we have considered a 2×1
multiplexer as we have discussed earlier. This has inputs a, b, and c and the
output y. The d wire is output of the not gate whose input is b. f is the output
of and gate whose inputs are a and d. g is the output of the and gate whose
inputs are b and c. To understand the concurrent fault simulation simpler,
we have considered two single stuck at faults. One is the a stuck at 0 and
other is the g stuck at 1.
Let us first consider the case of a stuck at 0. At this point, the and gate with inputs
a and d will have the fault. The fault-free and gate and the faulty and gate are simu-
lated together, and this fault is further propagated to the output of the circuit, and
it is verified with the expected response. If the fault is detectable, then that event is
called a good event or else the event is called bad event. The 2×1 multiplexer with the
a stuck at 0 is given in Figure 19.9 where ff is the output of the faulty gate output. If
the a stuck at 0, then output of the and gate is ff = 0. The output yfa (faulty circuit out-
put with the fault a stuck at 0) now depends only on the g which is equal to b and c.
In the concurrent fault simulation, the fault list has single fault (a stuck at 0); then
the test vectors are applied until the fault is detected and propagated to the output.
The detailed propagation of values of the faulty circuit and fault-free circuit is given
in Table 19.3.
In Table 19.3, a, b, and c are the inputs, f is the fault-free output of the and gate,
y is the output of the fault-free circuit, ff is the faulty and gate output, and yfa is
the faulty circuit output. The fdet signal is used for detecting the fault of that gate.
The ydet signal is used for checking whether the fault is propagated to the output or
not. We can observe that the fault (a stuck at 0) is observed at only the test vectors
(abc = 100,101) at the and gate output (ff), so the ff = 1 at those test vectors, which are
good events. Now, we check if the fault due to the a stuck at 0 is propagated to the
output. We observe the ydeta = 1 only at the test vectors (abc = 100,101). The simu-
lated waveform for the 2×1 multiplexer with a stuck at fault is given.
Verilog Code
module ckt(a,b,c,k,y,yfa,yfg,fdet,ydeta,ydetg);
input a,b,c,k;
output y,yfa,yfg;
output reg fdet,ydeta,ydetg;
wire d,f,g;
wire ff; // faulty and gate
supply1 vdd;
supply0 gnd;
not n1(d,b);
and a1(f,a,d);// fault free circuit
and a2(g,b,c);
or o1(y,f,g);
and a3(ff,gnd,d); // faulty and gate (a|0)
or o2(yfa,ff,g);
or o3(yfg,f,vdd); // faulty and gate
always@(f,ff,y,yfa,yfg,fdet,ydeta,ydetg)
begin
case(k)
TABLE 19.3
Faulty Circuit Truth Table for a Stuck at 0
a b c f Y Ff yfa fdet ydeta
0 0 0 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0
0 1 0 0 0 0 0 0 0
0 1 1 0 1 0 1 0 0
0 0 0 1 1 0 0 1 1
0 0 1 1 1 0 0 1 1
0 1 0 0 0 0 0 0 0
0 1 1 0 1 0 1 0 0
336 Advanced VLSI Design and Testability Issues
1'b0 : begin
if(f!=ff)
begin
assign fdet= vdd;
if(y= =yfa)
assign ydeta = gnd;
else
assign ydeta = vdd;
end
else
begin
assign fdet= gnd;
assign ydeta= gnd;
end
end
1'b1 : begin
if(y= =yfg)
assign ydetg = gnd;
else
assign ydetg = vdd;
end
endcase
end
endmodule
Test Bench
module ckt_tb();
reg a,b,c,k;
wire y,yfa,yfg,fdet,ydeta,ydetg;
ckt u(a,b,c,k,y,yfa,yfg,fdet,ydeta,ydetg);
initial
begin
k=1;a=0; b=0; c=0;
#5 k=1; a=0; b=0; c=1;
#5 k=1; a=0; b=1; c=0;
#5 k=1; a=0; b=1; c=1;
#5 k=1; a=1; b=0; c=0;
#5 k=1; a=1; b=0; c=1;
#5 k=1; a=1; b=1; c=0;
#5 k=1; a=1; b=1; c=1;
#5 $stop;
end
endmodule
Simulation Results:
In the above simulation results, the signal k is given 0 because we have
used case statement in the Verilog program (Figure 19.10). So, if we give
k = 0, then the a stuck at 0 fault simulation is carried out, and if we give k = 1,
then g stuck at 1 is carried out.
Fault Simulation Algorithms 337
Now, we consider the g stuck at 1 fault; then the collapsed fault list now contains
the g stuck at 1. Since the output of the multiplexer is yfg = f + g, the output will be
yfg = 1 (Figure 19.11).
Since g stuck at 1, there are no gates to be evaluated, and only the output gate
has to be evaluated. The faulty output is yfg that is compared with output faulty
free circuit, which is y. The circuit is simulated with test vectors set and is best
understood by Table 19.4. The gf signal is the output of the and gate which is
stuck at 1, and yfg is the faulty output signal which is also stuck at 1 due to g
stuck at 1.
The fault of g stuck at 1 can be observed at the test vectors (abc = 000,001,010,110),
and these are called good events. The simulated results for the g stuck at 1 are given
(Figure 19.12).
Here to simulate the g stuck at 1, the k is given, so the signals fdet and ydeta are
in don’t care state.
338 Advanced VLSI Design and Testability Issues
TABLE 19.4
Faulty Circuit Truth Table for g Stuck 1
a b c G gf y yfg
0 0 0 0 1 0 1
0 0 1 0 1 0 1
0 1 0 0 1 0 1
0 1 1 1 1 1 1
1 0 0 0 1 1 1
1 0 1 0 1 1 1
1 1 0 0 1 0 1
1 1 1 1 1 1 1
REFERENCES
1. L.-T. Wang, N. E. Hoover, E. H. Porter, and J. J. Zasio, SSIM: A software levelized
compiled-code simulator, in Proc. Des. Automat. Conf., June 1987, pp. 2–8.
2. E. G. Ulrich, Exclusive simulation of activity in digital networks, Commun. ACM,
12(2), 102–110, 1969.
3. D. B. Armstrong, A deductive method for simulating faults in logic circuits, IEEE
Trans. Comput., C-21(5), 464–471, 1972.
4. E. G. Ulrich and T. Baker, Concurrent simulation of nearly identical digital networks,
IEEE Trans. Comput., 7(4), 39–44, 1974.
5. L.-T. Wang, C.-W. Wu and X. Wen, VLSI Test Principles and Architectures, Morgan
Kaufmann Publishers, Elsevier, San Francisco, 2006.
6. W. T. Cheng and M. L. Yu, Differential fault simulation: A fast method using minimal
memory, in Proc. Des. Automat. Conf., June 1989, pp. 424–428.
20 Hardware Protection
through Logic
Obfuscation
Jyotirmoy Pathak and Suman Lata Tripathi
Lovely Professional University
CONTENTS
20.1 Introduction ................................................................................................ 339
20.2 Logic Working ............................................................................................ 341
20.3 Comparisons ............................................................................................... 345
20.4 Future Works .............................................................................................. 346
20.5 Conclusions ................................................................................................. 347
References .............................................................................................................. 347
20.1 INTRODUCTION
With the recent burst in technology, the use of integrated circuits (ICs) in various
fields has seen a tremendous increase. Due to this high demand, certain steps in
the process of fabrication are done outside of the foundry, which makes it prone
to several attacks. There are several processes employed for securing ICs against
malicious attacks. One such method is known as hardware obfuscation. Hardware
obfuscation basically means hiding the IC’s structure and function, which makes it
much more difficult to reverse-engineer by the adversaries. The adversaries gener-
ally make use of reverse engineering to decipher the IC. Reverse engineering basi-
cally stands for the technology that is used to describe be the structure, functionality,
and design of an IC. Once there is adequate knowledge of the inner structure of the
circuit, it is quite easy to tamper it or use it illegally. Thereby, obfuscation provides
a means of making the circuit structurally and functionally difficult to comprehend,
which increases the cost and time required to reverse-engineer it, providing security.
Hardware obfuscation provides a security not only against reverse engineering but
also against IC piracy, overbuilding of chips, duplication, etc. According to Moore’s
law, the number of transistors in a circuit is increasing twice every 18 months. This
is made possible because the size of the transistor is reducing to a larger extent; cur-
rently, work is being done in 7-nm technology. With the shrinkage in the size of the
transistors, it was believed that reverse engineering of the design will get difficult,
339
340 Advanced VLSI Design and Testability Issues
but it is not so. Apple’s iPhone 6 and iPhone 6s were reverse-engineered successfully,
and the processing units present inside it were identified.
Reverse engineering has made IC theft and espionage a major threat to IC indus-
try. There has been a tremendous increase in cases involving hardware theft [1].
This makes it even more urgent to concentrate on the hardware security perspective.
Research has been going on obfuscation at many levels. The basic idea of obfuscation
is premanufacture safety, which includes prevention from the addition of hardware
Trojans while designing a circuit. This Trojan can easily leak the confidential infor-
mation of the IC or can make the IC work in an unfamiliar manner. There have been
many researches that have been going on to detect the presence of Trojan and also
prevent the addition of Trojan. Another area of concern is postmanufacture in which
an attacker can use the manufacturer IC in ways that incur a loss to the customer or
manufacturer. There have been researches on gate-level obfuscation [2] in which
researchers make use of AOI gates instead of simple AND–OR logic gates and, after
that, reverse the obtained output in order for obtaining a correct value. At layout-
level obfuscation, they make use of special camouflaged gates instead of simpler
library gates, which prevent the hacker to identify its correct function even through
an SEM [3]. They also add dummy contacts to the layout, making it much more dif-
ficult to see the connectivity of the circuit, thereby making it a time-consuming task
to reverse-engineer the design [4].
HDL (hardware description language)-level obfuscation includes obfuscation of
the design by modification of HDL codes. Here, the codes that are written in Verilog
or VHDL are made much more difficult to understand. This is achieved by remov-
ing comment lines in coding [5] or by using an encryption technique using the AES
algorithm. Later, this encrypted code can be decrypted at the user end with the help
of a decryption key [6].
Illegal and unauthorized use of IC can be prevented by various other techniques
such as watermarking [7, 8] and cryptography [9, 10].
Obfuscation can be done using two basic ways:
• Structurally
• Functionally
In the case of structural obfuscation, we complicate the structure of the given design,
making it a tedious task to find its functionality [11]. In Ref. [12], we have seen
that by making use of high-level transformation (HLT) techniques, circuit struc-
tures were modified to look alike but perform a different functionality. So the circuit
obtained looks structurally the same but are functionally different.
In functional obfuscation, we make use of a key. Until and unless the entered key
is correct, the design does not give the desired output. This key can be stored in the
form of fuses in the circuit [13]. However, saving the key in the nonvolatile memory
of the circuits in the form of fuses makes it insecure, as it can be obtained by the
adversaries. So there was a development of a technique called physical unclonable
function (PUF) that provides with the key to the circuitry [14]. The key can be pro-
vided either using an FSM or a PUF. On entering the correct key, the circuit oper-
ated in a normal mode, and on entering a wrong key, the circuit operated in an
Hardware Protection through Logic Obfuscation 341
Logic Obfuscation
Structural Obfuscation
- Use of HLT Functional Obfuscation
- Use of Loop Unrolling
Sequential Combinational
Obfuscation Obfuscation
- Use of FSM - Modification of
Structures Obfuscation
- Use of PUFs Cells.
- Logic Locking
FIGURE 20.1 Taxonomy for logic obfuscation. FSM, finite-state machine; HLT, high-level
transformation; PUF, physical unclonable function.
342 Advanced VLSI Design and Testability Issues
at various random site locations. If the provided key is incorrect, the output obtained
is fixed and incorrect. The key is stored in a tamper-free memory inside the chip.
The greater the key size, the longer the time required for the attacker to decipher the
key. The area overhead, in this case, is quite high even when only 5% of gates are
inserted as key gates compared with the total gates in the design. There has been
a constant development in the area of hardware security recently. Numerous tech-
niques and area are researched under this. Classification on the hardware security
on the different area has been described in Ref. [21] for a better understanding of
the people researching in this area. In Ref. [22], they developed a novel technique of
obfuscating a digital signal–processing circuit by use of HLT. HLT was mainly used
for area, delay, and power trade-offs, but it was observed that, apart from this, HLT
also provides structural and functional obfuscation. They also introduced multiple
modes that are meaningful but functionally incorrect, thereby providing a higher
level of security. When the incorrect key is applied from the FSM, the reconfigura-
tion can activate either a meaningful or nonmeaningful mode, which is functionally
incorrect. This technique provided an advantage of lower area overhead by provid-
ing a higher level of security. As we can observe from the papers mentioned earlier
the area overhead for these techniques is a concern, thereby in Ref. [23], there is a
mechanism for providing security to the circuit without affecting the area overhead
of the design. In this paper, the original net list of the IC is modified to obtain obfus-
cated net list. This obfuscated cell (OC) results in a correct output if the entered key
is correct. The advantage of OC is that the adversary cannot identify it on reverse
engineering by image processing techniques. This makes only the authorized chips
to work properly. The previous techniques of using XOR/XNOR gates resulted in a
high area overhead, which is not in this case. It resulted in an area overhead of 0.63%
and a power overhead of 2.6% in the circuit.
In Ref. [24], they have classified hardware obfuscation at different levels. They
have researched on various techniques employed to reverse-engineer a design,
checked the obfuscation techniques at various level, and then focused on the obfus-
cation at device level by manipulating the interconnects and the transistors in the
circuit. They have varied the doping concentration and introduce dummy contacts
for increasing the effects of reverse-engineering a design. They also showed that
changing the logic family from CMOS (complementary metal oxide semiconductor)
to pass transistor logic or transmission gates also increases the efforts of the adver-
sary to reverse-engineer a design. For a more detailed understanding of the layout-
and net list–level obfuscation, it must be noted that the survey paper in Ref. [25] has
very precisely classified all the possible techniques for various layout-level obfusca-
tion along with their demerits as well as the net list–level obfuscation techniques. In
Ref. [26], there is a slight modification of the technique that was presented in Ref.
[22]. This paper provides a structural obfuscation by using five basic techniques such
as loop unrolling, loop invariant code motion, tree height reduction, logic transfor-
mation, and removal of redundant logics. After application of these modifies HLT
techniques, particle swarm optimization (PSO) is performed for the optimization of
the design. This results in a low cost of obfuscation. Along with structural obfus-
cation using HLT, there have been researches going on various logic locking and
344 Advanced VLSI Design and Testability Issues
technique is not very secure as it is very easy to know that logic is locked to 1 if any
one of the inputs of the OR gate is 1. So to make it much more difficult to understand,
a modification is done in which the use of XNOR/NAND/NOR gate is done for
securing the logical functioning, but this raises another issue of a large area overhead
in the circuit. In Ref. [34], they have used the circuitry already present inside the IC
for logic-locking procedure. Nowadays, testing of an IC is equally important, and to
do a thorough test, extra circuitry is included in the design known as test points. This
paper makes use of those test points as the logic-locking gates. It introduces the PUF
and scrambler in the circuit. PUF is used to generate unclonable keys, and scrambler
makes the circuit operated in either test mode/functional mode or obfuscated mode.
20.3 COMPARISONS
In Ref. [35], protection is provided to IP cores by embedding a digital signature in
the IP. This consumes a very less area overhead but provides security against fake
ownership and does not affect the design from overbuilding and reverse-engineering
attacks. To provide protection against IP piracy, Rajendran (2012) [36] explored the
concept of EPIC in which they locked the design using combinational locking mech-
anism using a key. This rendered the design useless if the entered key is wrong. This
includes the area overheads of the wires and gates for additional locking, TRNG,
and so on but provides better security compared with digital signatures. Based on the
length of the key entered, the security becomes better. For a 64-bit key, it requires
a year to break it as there are more than 2^20 combinations available. The protec-
tion mainly focuses on IP vendors and the end users to modify it. In Ref. [16], they
described a means of providing security at manufacturers, IP vendors, chip design-
ers, and end users by obfuscation of net list. This is done by using XOR and OR
gates and randomly placing internal wires at a different location. This is performed
at different area constraints, and it is observed that the larger the area constraint, the
larger the failing vectors. The timing constraints and the power overhead were under
the average of 22%, which is well within the manageable limit. This was modified
in Ref. [17] to provide protection in RTL by using key, and it was observed that the
number of failing vectors, in this case, is very high and the area and power overhead
is under 10%. These circuits fail to provide protection against Trojan attacks, which
triggers at a rare event, so in Ref. [19], they devised an approach to make the present
Trojans benign and also prevent the rare triggering condition to occur. This technique
provides a high Trojan coverage to provide protection at an overall area and power
overhead of 11% on an average. Ref. [36] was modified in Ref. [37] by inserting the
extra gates at a random location, weighted and unweighted locations. It results in a
high area overhead and a power–delay product of 25% for random placement and
21% on unweighted placement. In Ref. [38], they have devised another clever method
of obfuscation by using dynamic traversal of paths by locking the mode using code
lock during state transition time with a nominal area overhead of 7.8%. Regardless
of the above safety measures, if an attacker inserts the hardware Trojan in the chip
with low controllability, it becomes very difficult to know the affected area, so in
Ref. [39], they developed a technique for encryption of the circuit that prevents the
introduction of the Trojans in the circuit areas with low controllability. In Ref. [22],
346 Advanced VLSI Design and Testability Issues
the one-level protection is modified to two-level protection, i.e., both structural and
functional protection. One of the major benefits of this process is the meaningful
output nodes that are functionally correct but still the output is wrong, which makes
it hard for the adversary to know the proper working of the design. In Ref. [40],
they have utilized the layout-level obfuscation by generating special gates called as
obfuscates, which are stored in the obfuscated library and used, which required the
adversary to provide extra effort in reverse engineering it. It gives an area overhead
of 5.67× on average, but it can be further reduced by using a less number of inputs
in obfuscates. The area and power overhead along with proper security is becoming
an issue when we talk about the obfuscation, so in Ref. [23], a scheme is developed
making use of inverter and mux to form an OC, which helps in providing a better
security even if the gate-level net list is obtained by hiding the functionality with an
average area overhead of 0.63% and a power overhead of 2.6%. To further modify
the structural obfuscation presented in Ref. [22], Yasin (2017) [26] developed a few
higher transformation techniques as well as a low cost–driven design by using PSO.
It provides an obfuscation of 22% within reduction in the cost of design by 55%. In
the aforementioned technique, we use the same OCs for obfuscation of the design;
it is easier for the adversary to reverse-engineer all of these once the adversary
reverse-engineers one, so Yasin et al. [32] developed a technique of using differ-
ent OCs in design, making it difficult to reverse-engineer. The area–power–delay
constraint in this technique is negligible. For further increasing the security of the
circuit, Chakraborty et al. [15] made use of dynamic obfuscation technique in which
when the entered key is wrong, the output of the design is inconsistent. The inconsis-
tency can be further randomized, and this prevents the circuits for SAT attacks. The
area and power overhead, in this case, is 1%, and security is increased effectively.
Another technique developed to make the circuit resilient to SAT-based attack is
discussed in Ref. [30]. They have made use of cyclic obfuscation by creating dummy
paths in which way this produces an area overhead of 40% less than the use of XOR
gate in camouflaging and a delay of 20% less than the delay of use of XOR gate in
camouflaging. The technique mentioned before obfuscated the design by adding an
additional gate in case of logic-locking technique, but in Ref. [34], they made use
of test points to function as the logic-locking architecture to function in testing or
normal operation as well as to lock the design in case of wrongly entered key. This
increases the design for testability along with providing necessary protection to the
circuit.
20.5 CONCLUSIONS
With the increase in the use of IC chips in numerous areas, the reverse engineering
of these chips has seen a tremendous increase in the past decade. This technique
can lead to IP piracy, overbuilding, leakage of secure information, Trojan introduc-
tion, malfunctioning of IC, and so on. Securing the IC has become more important
than ever. To ensure the safety of the device, many techniques have been researched
upon, among which logic obfuscation is most widely used. This provides security
to design both structurally and functionally. It makes two functionally different
designs look structurally the same, thereby making it difficult for the adversary
to reverse-engineer. It also includes the use of camouflaging, dummy gates, and
the introduction of different logic blocks. It makes use of a key to validate a cir-
cuit; in case of entering a wrong key, the circuit does not operate as intended. To
provide more security, the key length was focused upon, which came with a cost
of area overhead. This leads to the research in various other alternatives to make
the design much more secure and increases the complexity of the circuit keeping
the area, power, and delay overhead constraints within permissible limits. These
include making use of PUF, hybrid PUF, dynamic obfuscation, different OCs, and
so on. As per the need of each design with a specific area and delay constraint, these
techniques are utilized. Future works for security on different technologies such
as 3D ICs can be a major turn in providing a much higher level of security against
reverse engineering.
REFERENCES
1. Foreign infringement of intellectual property rights implications on selected U.S.
industries. http://www.usitc.gov/publications/332/working_papers/id_14_100505.pdf.
2. Becker, Georg T., Francesco Regazzoni, Christof Paar, and Wayne P., Burleson.
“Stealthy dopant-level hardware trojans: Extended version.” Journal of Cryptographic
Engineering 4, no. 1 (2014): 19–31.
3. Rajendran, Jeyavijayan, Michael Sam, Ozgur Sinanoglu, and Ramesh Karri. Security
analysis of integrated circuit camouflaging. In Proceedings of the 2013 ACM SIGSAC
Conference on Computer & Communications Security, pp. 709–720 (2013).
348 Advanced VLSI Design and Testability Issues
21. Lao, Yingjie, and Keshab K. Parhi. “Obfuscating DSP circuits via high-level transfor-
mations.” IEEE Transactions on Very Large Scale Integration (VLSI) Systems 23, no. 5
(2015): 819–830.
22. Zhang, Jiliang. “A practical logic obfuscation technique for hardware security.” IEEE
Transactions on VLSI Systems 24, no. 3 (2016): 1193–1197.
23. Vijayakumar, Arunkumar, Vinay C. Patil, Daniel E. Holcomb, Christof Paar, and
Sandip Kundu. “Physical design obfuscation of hardware: A comprehensive investi-
gation of the device and logic-level techniques.” IEEE Transactions on Information
Forensics and Security 12, no. 1 (2017): 64–77.
24. Becker, Georg T., Marc Fyrbiak, and Christian Kison. Hardware obfuscation:
Techniques and open challenges. In Foundations of Hardware IP Protection, pp.
105–123. Springer, 2017.
25. Sengupta, Anirban, Dipanjan Roy, Saraju P. Mohanty, and Peter Corcoran. “DSP
design protection in CE through algorithmic transformation based structural obfusca-
tion.” IEEE Transactions on Consumer Electronics 63, no. 4 (2017): 467–476.
26. Yasin, Muhammad, Bodhisatwa Mazumdar, Ozgur Sinanoglu, and Jeyavijayan
Rajendran. “Removal attacks on logic locking and camouflaging techniques.” IEEE
Transactions on Emerging Topics in Computing 1 (2017): 1–1.
27. Yasin, Muhammad, Bodhisatwa Mazumdar, Ozgur Sinanoglu, and Jeyavijayan
Rajendran. Security analysis of anti-sat. In Design Automation Conference (ASP-
DAC), 2017 22nd Asia and South Pacific, pp. 342–347 (2017).
28. Yasin, Muhammad, Bodhisatwa Mazumdar, Ozgur Sinanoglu, and Jeyavijayan
Rajendran. “Removal attacks on logic locking and camouflaging techniques.” IEEE
Transactions on Emerging Topics in Computing 1 (2017): 1.
29. Shamsi, Kaveh, Meng Li, Travis Meade, Zheng Zhao, David Z. Pan, and Yier Jin.
Cyclic obfuscation for creating sat-unresolvable circuits. In Proceedings of the on
Great Lakes Symposium on VLSI 2017, pp. 173–178 (2017).
30. Zhang, Grace Li, Bing Li, Bei Yu, David Z. Pan, and Ulf Schlichtmann.
TimingCamouflage: Improving circuit security against counterfeiting by unconven-
tional timing. In 2018 Design, Automation & Test in Europe Conference & Exhibition
(DATE), pp. 91–96 (2018).
31. Sumathi, G., L. Srivani, D. Thirugnana Murthy, Anish Kumar, and K. Madhusoodanan.
Hardware obfuscation using different obfuscation cell structures for PLDs. In SG-CRC,
pp. 143–157 (2017).
32. Yasin, Muhammad, Jeyavijayan, JV Rajendran, Ozgur Sinanoglu, and Ramesh Karri.
“On improving the security of logic locking.” IEEE Transactions on Computer-Aided
Design of Integrated Circuits and Systems 35, no. 9 (2016): 1411–1424.
33. Chen, Michael, Elham Moghaddam, Nilanjan Mukherjee, Janusz Rajski, Jerzy Tyszer,
and Justyna Zawada. “Hardware protection via logic locking test points.” IEEE
Transactions on Computer-Aided Design of Integrated Circuits and Systems, 37, no. 12
(2018): 3020–3030.
34. Castillo, Encarnacin, Uwe Meyer-Baese, Antonio García, Luis Parrilla, and Antonio
Lloris. “IPP@ HDL: Efficient intellectual property protection scheme for IP cores.”
IEEE Transactions on Very Large Scale Integration (VLSI) Systems 15, no. 5 (2007):
578–591.
35. Roy, Jarrod A., Farinaz Koushanfar, and Igor L. Markov. EPIC: Ending piracy of inte-
grated circuits. In Proceedings of the Conference on Design, Automation and Test in
Europe, pp. 1069–1074 (2008).
36. Rajendran, Jeyavijayan, Youngok Pino, Ozgur Sinanoglu, and Ramesh Karri. Security
analysis of logic obfuscation. In Proceedings of the 49th Annual Design Automation
Conference, pp. 83–89 (2012).
350 Advanced VLSI Design and Testability Issues
37. Desai, Avinash R., Michael S. Hsiao, Chao Wang, Leyla Nazhandali, and Simin Hall.
Interlocking obfuscation for anti-tamper hardware. In Proceedings of the Eighth
Annual Cyber Security and Information Intelligence Research Workshop, p. 8 (2013).
38. Dupuis, Sophie, Papa-Sidi Ba, Giorgio Di Natale, Marie-Lise Flottes, and Bruno
Rouzeyre. A novel hardware logic encryption technique for thwarting illegal overpro-
duction and hardware trojans. In On-Line Testing Symposium (IOLTS), 2014 IEEE 20th
International, pp. 49–54 (2014).
39. Malik, Shweta, Georg T. Becker, Christof Paar, and Wayne P. Burleson. Development
of a layout-level hardware obfuscation tool. In 2015 IEEE Computer Society Annual
Symposium on VLSI (ISVLSI), pp. 204–209 (2015).
40. Imeson, Frank, Ariq Emtenan, Siddharth Garg, and Mahesh V. Tripunitara. Securing
computer hardware using 3D integrated circuit (IC) technology and split manufactur-
ing for obfuscation. In USENIX Security Symposium, pp. 495–510 (2013).
41. Xie, Yang, Chongxi Bao, Yuntao Liu, and Ankur Srivastava. 2.5 D/3D integra-
tion technologies for circuit obfuscation. In 2016 17th International Workshop on
Microprocessor and SOC Test and Verification (MTV), pp. 39–44 (2016).
42. Yang, Jianlei, Xueyan Wang, Qiang Zhou, Zhaohao Wang, Hai Li, Yiran Chen, and
Weisheng Zhao. “Exploiting spin-orbit torque devices as reconfigurable logic for circuit
obfuscation.” IEEE Transactions on Computer-Aided Design of Integrated Circuits
and Systems (2018).
43. Mathew, Jimson, Rajat Subhra Chakraborty, Durga Prasad Sahoo, Yuanfan Yang,
and Dhiraj K. Pradhan. “A novel memristor-based hardware security primitive.” ACM
Transactions on Embedded Computing Systems (TECS) 14, no. 3 (2015): 60.
Index
A B
Active inductor (AI), 314–315 Beetle/chip FET, 182, 184
equivalent circuit of lossy, 316 Behavioral modeling, 54–55
gyrator-C, 314–315 Verilog, 107–108
reconfigurable VCO with, 316 VHDL, 73, 101
VCO with differential, 318 Bell, John, 306
voltage-controlled oscillator using, 315–319 Binary number systems, 19
Active inductor–based VCO Biochemical-based biosensor, 170
active inductor, 314–315 Bio-FETs
discrete Fourier transform for VCO, applications, 170, 178
319–320 cell-based, 180–183
overview, 311–313 classification and advances in, 176–179
ring voltage-controlled oscillator, 314 different improvement techniques, 177–178
voltage-controlled oscillator using active DNA-modified FET or GenFET, 178–179
inductor, 315–319 dual-gate ISFET (DG-ISFET), 172–173
for wireless communication, 311–322 enzyme FET (EnFET), 176
Adaptive antenna array, 294 FET, a Bio-FET’s basic structure, 170–171
Adder and subtractor circuit, 25–27 ISFET, a pH Sensor, 171–172
Addition of ceramic fillers, 228–229 ISFET bioapplications, 173
AHPL (A Hardware Programming major benchmarks in Bio-FET device, 177
Language), 95 types of, 184
AlGaN/GaN heterostructure field-effect working, 176–177
transistor (HFET) Bioreceptors, 148
model description, 216 Biosensors
overview, 215–216 advantages, 148–149
results and discussions, 216–221 biochemical-based, 170
Algorithmic modeling, 54–55 components, 148
Analog design defined, 169–170
described, 18 FET-based, 150–165
vs. digital design, 18 introduction, 147–148
Analog-to-digital conversion, 297–298 low-power FET-based (see Bio-FETs)
Analog-to-digital converters, 297–305 penicillin-sensitive biosensor, 176
basics of analog-to-digital conversion, silicon nanowire, 173–174
297–298 types of, 149–150
delta-sigma, 305 Blocking and nonblocking statements, 55–56
flash, 303–304 Boolean equation, 22–24
pipelined, 304–305 Bulk MOSFET, 115–116
quantization, 299–300 Butler matrices, 294
sampling, 298–299 Byte, defined, 2
Analytes, 148
APL (artificial language), 94–95 C
Architecture, 72
behavioral modeling, 73 Calorimetric biosensor, 150
dataflow modeling, 72 Cantatore, Eugenio, 302
structural modeling, 73–74 Carbon nanotube biosensor, 153–154
Armchairedge GNR (AGNR), 138–139 Carbon nanotube field-effect transistors
Artificial neural network (ANN), 259–261, 263 (CNTFETs), 137
Asynchronous counters, 33–34 Cell-based Bio-FETs, 180–183
Atmel, 9 application, 182–183
@ (sensitivity_list), 49 beetle/chip FET, 182
working, 181–182
351
352 Index
L M
Latches, 30–31 Macromodeling
Lattice Semiconductor, 8, 9, 10 nonparametric, 259–263
Leakage reduction techniques, 211–213 overview, 251–252
lower self-controllable voltage level, 211 parametric-based, 252–259
self-controllable voltage level technique, 211 Magnetic storage, 36
upper self-controllable voltage level, 211–213 Mahapatra, K.K., 304
Lehmann, Torsten, 306 McCulloch, Warren, 259
Lewis, Stephen H., 304 Memory
Lexical tokens, 74–75 described, 34
Libraries, VHDL, 98–99 digital circuits, 34–36
Liquid electrolytes, 226, 229 flash, 36
Logic blocks, 9 magnetic storage, 36
configurable, 10–12 optical storage, 36
Logic families RAM, 36
complementary metal oxide semiconductor, ROM, 35
22, 23 Memristor, 122–123
digital integrated circuit, 19–20 Metal gate electrodes, 119
diode–transistor logic, 20, 21 Metal oxide semiconductor field-effect transistor
emitter-coupled logic, 20 (MOSFET), 112
resistor–transistor logic, 20, 21 bulk, 115–116
transistor–transistor logic, 22 double-gate silicon-on-insulator, 117
Logic gate delays, 69 junctionless, 119–120
Logic simulation, 325–326 PDSOI, 116
compiled code simulation, 326 scaling down, 129–130
event-driven simulation, 326 SOI, 116
Logic working, 341–345 surrounding-gate silicon-on-insulator, 118
Loop filter or low-pass filter, 274 topologies, 119–124
Lower self-controllable voltage level, 211 triple-gate silicon-on-insulator, 118
Low-noise fast PLLs using multi-PFD Michael Choi, 305
architectures, 281–283 Microsemi (Actel), 10
Low-power adiabatic logic techniques Mixed model, 59
conventional positive-feedback adiabatic Model order reduction, 256–259, 263
logic, 240–241 Monitor, System Verilog, 81
proposed, 239–241 Moore, Gordon, 113
two-phase adiabatic static clocked Moore’s law, 113, 129, 131–132
logic, 241 More than Moore (MtM), 113
Low-power FET-based biosensors Multibridge channel MOSFET (MBCFET), 118
cell-based Bio-FETs, 180–183 Multiplexer, 27–29, 330–338
classification and advances in, 176–179
immunofet, 179–180
organic field-effect transistor, 174–176
N
overview, 169–170 Nanobeam stacked channels, 118
principle of operation, 170–173 Nanoelectronic devices, see Nanotechnology
silicon nanowire biosensor, 173–174 Nanofillers, 230
Low-power techniques Nanotechnology
comparative analysis, 245–247 applications, 143
conventional complementary metal oxide beyond CMOS, 136–142
semiconductor, 235–236 emerging improvements, 115–124
differential pass-transistor logic style, 237 limitations, 143
existing, 235–239 material technology, 118–119
existing design, 241–243 overview, 112–113
gate diffusion input logic style, 238–239 technical challenges and solutions, 142–144
pass-transistor logic style, 236–237 Nanowire (NW) FETs, 136, 139
proposed design, 243–245 Nanowire field-effect transistor biosensor, 153
transmission gate logic style, 237–238 Nanowire (NW) transistors, 136–137
356 Index
generate statement, 50 W
identifiers, 42, 74
introduction, 96–104 Wan Kim, 305
keywords of, 41, 75 Whitespace, 42, 74
lexical tokens, 74–75 Wide-lock range PLLs using AFC techniques,
libraries, 98–99 283–284
miniproject, 85–91 Wireless communication
modeling types, 54–59 active inductor, 314–315
number or constant value, 45 active inductor–based VCO for, 311–322
operator types, 52–54, 75 deployment of SAs for, 292–293
packages, 98–99 discrete Fourier transform, 319–320
parameters, 45 inductor–capacitor voltage-controlled
procedural assignment, 76–77 oscillator, 312–313
sequential statements, 45–49 ring voltage-controlled oscillator, 314
simulations and synthesis, 103–104 voltage-controlled oscillator using active
switch-level primitives, 51 inductor, 315–319
test bench, 60–62, 77–78
timescale, 75 X
time value (#), 49
tristate gates, 50–51 Xilinx, 9, 10
user-defined primitives, 59–60 Xueyi Liu, 257
variables, 42–44, 98
vector data, 44–45 Y
whitespace, 42, 74
Verilog IEEE standard, 41 Yip, Marcus, 303
VHSIC (very highspeed integrated circuits), 40, 95 Yong Chen, 258
VLSI technology, 234
Voelker, Matthias, 306 Z
Voltage-controlled oscillator (VCOs), 273–274
comparison of AI-based, 321 Zeus, 95
discrete Fourier transform for, 319–320 Zigzag-edge GNR (ZGNR), 138–139
ring, 314
used for frequency translation, 274
using active inductor, 315–319
Taylor & Francis eBooks
www.taylorfrancis.com
Improved
A streamlined A single point search and
experience for of discovery discovery of
our library for all of our content at both
customers eBook content book and
chapter level