Lecture 080 - Latchup and Esd: CMOS Analog Circuit Design, 2

Lecture 080 Latchup and ESD (3/24/10)
Page 080-1
LECTURE 080 LATCHUP AND ESD

LECTURE ORGANIZATION
Outline
Latchup
ESD
Summary
CMOS Analog Circuit Design, 2nd Edition Reference
Pages 48-52 and new material
CMOS Analog Circuit Design
P.E. Allen - 2010
Page 080-2
LATCHUP
nt
What is Latchup?
Latchup is the creation of a low impedance path
essive Curre
xc
between the power supply rails.
Latchup is caused by the triggering of parasitic VDD
bipolar structures within an integrated circuit
when applying a current or voltage stimulus on
an input, output, or I/O pin or by an over-voltage
on the power supply pin.
070221-01
Temporary versus true latchup:
A temporary or transient latchup occurs only while the pulse stimulus is connected
to the integrated circuit and returns to normal levels once the stimulus is removed.
A true latchup remains after the stimulus has been removed and requires a power
supply shut down to remove the low impedance path between the power supply rails
P.E. Allen - 2010
Page 080-3
Latchup Testing
The test for latchup defines how the designer must think about latchup.
For latchup prevention, you must consider where a current limited (100mA), 10ms
pulse is going to go when applied to a pad when the voltage compliance of the pad is
constrained to 50% above maximum power supply and to 2V below ground. (Higher
temperatures, 85Cand 125C, are more demanding, since VBE is lower.)
100m
A
10m
VDD
050727-06
Latchup is sensitive to layout and is most often solved at the physical layout level.
P.E. Allen - 2010
Page 080-4
How Does Latchup Occur?

Latchup is the regenerative process that can occur in a pnpn structure (SCR-silicon
controlled rectifier) formed by a parasitic npn and a parasitic pnp transistor.
Anode
Anode
iPNPN
DD
iPNPN
p
n
p
vPNPN
1/Slope =
Limiting To avoid latchup
Resistance vPNPN < VS
Triggering by
increasing V
Hold Current, IH
DD
n
Cathode
Cathode
Avalanche
Breakdown
Hold
Voltage, VH
Body diode
(CMOS)
Sustaining
voltage, VS
vPNPN
050414-01
Important concepts:
To avoid latchup, vPNPN VS
Once the pnpn structure has latched up, the large current required by the above i-v
characteristics must be provided externally to sustain latchup
To remove latchup, the current must be reduced below the holding current
P.E. Allen - 2010
Page 080-5
Latchup Triggering
Latchup of the SCR can be triggered by two different mechanisms.
1.) Allowing vPNPN to exceed the sustaining voltage, VS.
2.) Injection of current by a triggering device (gate triggered)
Anode
Pad
VDD
Gate
Current
pnp
Gate
Injector
npn
Gate
Injector
Gate
Current
SCR
SCR
Pad
Cathode
050414-03
Note: The gates mentioned above are SCR junction gates, not MOSFET gates.
From the above considerations, latchup requires the following components:
1.) A four-layer structure (SCR) connected between VDD and ground.
2.) An injector.
3.) A stimulus.
P.E. Allen - 2010
Page 080-6
Necessary Conditions for Latchup

1.) The loop gain of the relevant BJT configuration must exceed unity.
+fb
loop
VDD
ii
io
050414-04
Loop gain:
io
i i p n
2.) A bias condition must exist such that both bipolars are turned on long enough for
current through the SCR to exceed its switching current.
3.) The bias supply and associated circuits must be capable of supplying the current at
least equal to the switching current and at least equal to the holding current to maintain
the latched state.
P.E. Allen - 2010
Page 080-7
Latchup Trigger Modes

Current mode (Positive Injection Example):
When a current is applied to a pad, it can flow through
an injector and trigger latchup of an SCR formed from
parasitic bipolar transistors.
SCR gate current injection parasitic can occur in pwell or n-well technology.
Voltage mode:
When the power supply is increased
above the nominal value, the SCR formed from
parasitic bipolar transistors can be triggered.
Pad
Injector
VDD
Gate
Current
050414-05
SCR
VDD
VAnode
VDD < VAnode <Vabs,max
SCR
050414-06
P.E. Allen - 2010
Page 080-8
How does Latchup Occur in an IC?

Consider an output driver in CMOS technology:
vIN
vOUT
VDD
VDD
vIN
vOUT
p+
p-
n-
n+
Oxide Poly 1 Poly 2 Nitride Salicide Metal
050416-02
Assume that the output is connected to a pad.
P.E. Allen - 2010
Page 080-9
Parasitic Bipolar Transistors for the n-well CMOS Inverter

vIN
vOUT
VDD
Rw3
LT2
Rs2
Rs1 LT1
Rw4
VT2
Rs3
p+
n-
p-
n+
Rw2
VT1 Rw1
Rs4
050416-03
Parasitic components:
Lateral BJTs LT1 and LT2
Vertical BJTs VT1 and VT2
Bulk substrate resistances Rs1, Rs2, Rs3, and Rs4
Bulk well resistances Rw1, Rw2, Rw3, and Rw4
P.E. Allen - 2010
Page 080-10
Current Source Injection

Apply a voltage compliant current source to the output pad (vOUT > VDD).
vIN vOUT
Voltage Compliant
Current Source
VDD
LT2
Rs
p+
LT1
p-
n-
VT1 Rw
VT2
n
n+
050416-04
Loop gain:

iout
Rw
Rs

iin = P1Rw+rP1N1Rs+rN1
Rw
R s

V
R + P1 tR + N1Vt
w IP1 s IP2
= P1N1
P.E. Allen - 2010
Page 080-11
Current Sink Injection

Apply a voltage compliant current sink to the output pad (vOUT < 0).
vIN vOUT
Voltage Compliant
Current Sink
VDD
Rw3
LT2
Rs
p+
LT1
p-
VT1 Rw
VT2
n-
n+
050416-07
Loop gain:

iout
Rw
Rs

=

P1
N1
iin
Rw+rP1
Rs+rN1
Rw
R s

R + P1VtR + N1Vt
w IP1 s IP2
= P1N1
P.E. Allen - 2010
Page 080-12
Latchup from a Transmission Gate

The classical push-pull output stage is only one of the many configurations that can lead
to latchup. Here is another configuration:
Transmission
Gate
Internal
Core
Circuits
Internal Core
Circuitry
Clk
Pad
VDD
Pad
VDD
VDD
Clk
Injectors
Driver
Receiver
Transmission Gate
050416-09
p+
p-
n-
n+
Clock Driver
The two bold solid bipolar transistors in the transmission gate act as injectors to the npnpnp parasitic bipolars of the clock driver and cause these transistors to latchup. The
injector sites are the diffusions connected to the pad.
P.E. Allen - 2010
Page 080-13
The Influence of Shallow Trench Isolation on Latchup

As seen below, the STI causes the parasitic betas to be smaller.
Protective Insulator Layer
VDD
Intermediate
Oxide
Layers
OUTPUT
Metal Vias
GRD
Tungsten
Plugs
p+
Metal Via
Tungsten Plugs
Polycide
Sidewall
Spacers
Tungsten
Plugs
Salicide
GRD
Salicide
Salicide
n+
p+
p+
Shallow
Trench
Isolation
Tungsten
Plug
Salicide
n+
Top
Metal
Second
Level
Metal
First
Level
Metal
p+
n+
Shallow
Trench
Isolation
Shallow
Trench
Isolation
p-well
n-well
Substrate
Gate Ox
p+
Oxide
p-
n-
n+
Poly
Salicide Polycide
Metal
060406-01
P.E. Allen - 2010
Page 080-14
Preventing Latch-Up
1.) Keep the source/drain of the MOS device not in the well as far away from the well as
possible. This will lower the value of the BJT betas.
2.) Reduce the values of RN- and RP-. This requires more current before latch-up can
occur.
3.) Surround the transistors with guard rings. Guard rings reduce transistor betas and
divert collector current from the base of SCR transistors.
p-channel transistor
n+ guard bars
VDD
FOX
VSS
FOX
n- substrate
n-channel transistor
p+ guard bars
FOX
FOX
FOX
p-well
FOX
FOX
Figure 190-10
P.E. Allen - 2010
Page 080-15
What are Guard Rings?

Guard rings are used to collect carriers flowing in the silicon. They can be designed to
collect either majority or minority carriers.
Guard rings in n-material:
n+ guard ring
Collects
majority
carriers
Guard rings in p-material:
p+ guard ring
Collects
minority
carriers
VDD
n+ guard ring
Collects
minority
carriers
p+ guard ring
Collects majority
carriers
Decreased bulk
resistance
Decreased bulk
resistance
p+
p-
n-
p+
n+
VDD
p-
n-
n+
051201-01
051201-02
Also, the increased doping level of the n+ (p+)guard ring in n (p) material decreases the
resistance in the area of the guard ring.
P.E. Allen - 2010
Page 080-16
Example of Reducing the Sensitivity to Latchup

Start with an inverter with no attempt to minimize latchup and minimum spacing between
the NMOS and PMOS transistors.
vIN
vOUT
VDD
Rw
Rs
Note minimum separation

p+
p-
n-
n+
050427-03
P.E. Allen - 2010
Page 080-17
Example of Reducing the Sensitivity to Latchup by using Guard Rings

Next, place guard rings around the NMOS and PMOS transistors (both I/O and logic) to
collect most of the parasitic NPN and PNP currents locally and prevent turn-on of
adjacent devices.
vIN
p+ guard
ring
vOUT
VDD
n+ guard
ring
VDD
Rw
Rs
Note increased separation

p+
p-
n-
n+
050427-04
The guard rings also help to reduce the effective well and substrate resistance.
The guard rings reduce the lateral beta
Key: The guard rings should act like collectors
P.E. Allen - 2010
Page 080-18
Example of Reducing the Sensitivity to Latchup by using Butted Contacts

Finally, use butted source contacts to further reduce the well resistance and reduce the
substrate resistance.
vIN
p+ guard
ring
vOUT
VDD
n+ guard
ring
Rw
Rs
p+
VDD
p-
n-
n+
050427-05
P.E. Allen - 2010
Page 080-19
Guidelines for Guard Rings

Guard rings should be low resistance paths.
Guard rings should utilize continuous diffusion areas.
More than one transistor of the same type can be placed inside the same well inside the
same guard ring as long as the design rules for spacing are followed.
Only 2 guard rings are required between adjacent PMOS and NMOS transistors
The well taps and/or the guard ring should be laid out as close to the MOSFET source
as possible.
I/O output NMOSFET should use butted composite for source to bulk connections
when the source is electrically connected to the p-well tap. If separate well tap and
source connections are required due to substrate noise injection problems, minimize the
source-well tap spacing. This will minimize latch up and early snapback of the output
MOSFETs with the drain diffusion tied directly (in metal) to the bond pad.
P.E. Allen - 2010
Page 080-20
ESD IN CMOS TECHNOLOGY

What is Electrostatic Discharge?
Triboelectric charging happens when 2 materials come in contact and then are separated.
An ESD event occurs when the stored charge is discharged.
P.E. Allen - 2010
Page 080-21
ESD and Integrated Circuits

ICs consist of components that are very sensitive to excess current and voltage above
the nominal power supply.
Any path to the outside world is susceptible to ESD
ESD damage can occur at any point in the IC assembly and packaging, the packaged
part handling or the system assembly process.
Note that power is normally not on during an ESD event
050727-01
P.E. Allen - 2010
Page 080-22
ESD Models and Standards

Standard tests give an indication of the ICs robustness to withstand ESD stress.
Increased robustness:
- Reduces field failures due to ESD
- Demanded by customers
t=0 RLim
Simple ESD model:
+
- VSE = Charging Voltage
i(t)
VSE
IC
C
- Key parameters of the model:
o Maximum current flow
o Time constant or how fast the ESD
event
Current
discharges
o Risetime of the pulse
Imax
Time constant ()
RLimC
Risetime
0
t
070210-01
P.E. Allen - 2010
Page 080-23
ESD Models
Human body model (HBM): Representative of an ESD
event between a human and an electronic component.
050423-02
Machine model (MM): Simulates the ESD event when a

charged machine discharges through a component.
040929-03
Charge device model (CDM): Simulates the

ESD event when the component is charged
and then discharges through a pin. The
substrate of the chip becomes charged and
discharges through a pin.
P.E. Allen - 2010
Page 080-24
ESD Influence on Components

An ESD event typically creates very high values of current (1-10A) for very short periods
of time (150 ns) with very rapid rise times (1ns).
Therefore, components experience extremely high values of current with very little power
dissipation or thermal effects.
Resistors become nonlinear at high currents and will breakdown
Capacitors become shorts and can breakdown from overvoltage (pad to substrate)
Diodes current no longer flows uniformly (the connections to the diodes represent the
ohmic resistance limit)
Transistors ESD event is only a two terminal event, the third terminal is influenced by
parasitics and many of the transistor parameters are poorly controlled.
MOSFETs the parasitic bipolar experiences snapback under an ESD event
BJTs will experience snapback under ESD event
P.E. Allen - 2010
Page 080-25
Objective of ESD Protection

There must be a safe low impedance path between every combination of pins to sink the
ESD current (i.e. 1.5A for 2kV HBM)
The ESD device should clamp the voltage below the breakdown voltage of the internal
circuitry
The metal busses must be designed to survive 1.5A (fast transient) without building up
excessive voltage drop
ESD current must be steered away from sensitive
VDD
circuits
Limiting
Resistor
Sensitive
Circuits
ESD protection will require area on the chip (busses

and timing components)
VSS
ESD
Power
Rail
Clamp
041008-01
P.E. Allen - 2010
Page 080-26
ESD Protection Architecture

Rail based protection
VDD
Local
Clamp
Local
Clamp
Internal
Circuits
Input
Pad
Output
Pad
Local
Clamp
Local clamp based protection
Local
Clamp
VSS
ESD
Power
Rail
Clamp
040929-06
Local clamps Conducts ESD current without loading the internal (core) circuits
ESD power rail clamps Conducts a large amount of current with a small voltage drop
ESD Events:
Pad-to-rail (uses local clamps only)
Pad-to-pad (uses either local or local and ESD power rail clamps)
P.E. Allen - 2010
Page 080-27
Example of an ESD Breakdown Clamp

A normal MOSFET that uses the parasitic lateral BJT to achieve a snapback clamp.
Normally, the MOSFET has the gate shorted to the source so that drain current is zero.
S
G
D
- +
vDS
n+
Shallow
Trench
Isolation
p-substrate
iC
Rsub
Device destruction
iDS
Negative TC
p+
n+
iSub
iDS
Shallow
Trench
Isolation
Second Breakdown
Snapback Region
First Breakdown
Positive TC
Avalanche
Region
Linear Region
Saturation Region
Vt2
041217-04
vDS
Vt1
Issues:
If the drain voltage becomes too large, the gate oxide may breakdown
If the transistor has multiple fingers, the layout should ensure that the current is
distributed evenly.
P.E. Allen - 2010
Page 080-28
Example of a Non-Breakdown Clamp

NMOS Clamp:
VDD
Speed-up
Capacitor
R
Trigger
Circuit
C
Inverter
Driver
NMOS
Clamp
Operation:
041001-03
VSS
Normally, the input to
the driver is
high, the output low and the NMOS clamp off
For a positive ESD event, the voltage increases across R causing the inverter to turn on
the NMOS clamp providing a low impedance path between the rails
Cannot be used for pads that go above power supply or are active when powered up
For power supply turn-on, the circuit should not trigger (C holds the clamp off during
turn-on)
Also, forward biased diodes serve as non-breakdown clamps.
P.E. Allen - 2010
Page 080-29
Current
Current
IV Characteristics of Good ESD Protection

Goal: Sink the ESD current and clamp the voltage.
ITarget
ITarget
ESD Clamp
ESD Clamp
Protected
Device
Voltage
Case 2 - Protected Device Fails
Voltage
ESD
Clamp
Current
Case 1 - Okay
Current
Protected
Device
Protected
Device
ITarget
ESD Clamp
ITarget
ESD Clamp
Protected
Device
Protected
Device
Voltage
Case 3 - Okay
Voltage
Case 4 - Protected Device Fails
070221-02
P.E. Allen - 2010
Page 080-30
Current
Target
Iesd
Increasing
NMOS W
NMOS Vt
Max operating voltage
Comparison Between the NMOS Clamp and the Snapback Clamp

Increasing the width of theNMOS clamp will reduce the clamp voltage.
Increasing
snapback W
Vc
Holding
voltage
Vc
Vc
Voltage
Trigger
voltage
Note that the NMOS clamp does not normally exceed the absolute maximum voltage.
NMOS clamps should be used with EPROMs to avoid reprogramming during an ESD
event.
P.E. Allen - 2010
Page 080-31
ESD Practice
General Guidelines:
Understand the current flow requirements for an ESD event
Make sure the current flows where desired and is uniformly distributed
Series resistance is used to limit the current in the protected devices
Minimize the resistance in protecting devices
Use distributed (smaller) active clamps to minimize the effect of bus resistance
Understand the influence of packaging on ESD
Use guard rings to prevent latchup
Check list:
Check the ESD path between every pair of pads
Check for ESD protection between the pad and internal circuitry
Check for low bus resistance
- Current: Minimum metal for ESD 40 x Electromigration limit
- Voltage: 1.5A in a metal bus of 0.03/square of 1000m long and 30m wide gives
a voltage drop of 1.5V
Check for sufficient contacts and vias in the ESD path (uniform current distribution)
P.E. Allen - 2010
Page 080-32
SUMMARY
Latchup is the creation of a low impedance path between the power supply rails
resulting in excessive current.
The conditions for latchup are:
- A four-layer, pnpn structure connected between power supply rails
- An injector (any diffusion connected to a pad)
- A stimulus
Latchup is prevented by:
- Keeping the NMOS and PMOS transistors separated
- Reducing the well resistance with appropriate well ties
- Surrounding the transistors with guard rings
ESD is caused by triobelectric charging which discharges through the IC when the
power is off
The current produced by an ESD event must be controlled uniform current flow,
minimum voltage drop, and must not flow through sensitive circuitry
An ESD event turns on very quickly (<1ns), has a high peak current (1A), and lasts for
approximately 100 ns.
ESD clamps consist of breakdown clamps (snapback) and non-breakdown clamps.
P.E. Allen - 2010

Lecture 080 - Latchup and Esd: CMOS Analog Circuit Design, 2

Uploaded by

Copyright:

Available Formats

Lecture 080 - Latchup and Esd: CMOS Analog Circuit Design, 2

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture 080 - Latchup and Esd: CMOS Analog Circuit Design, 2

Uploaded by

Copyright:

Available Formats

Lecture 080 Latchup and ESD (3/24/10)