02 Simulating Electrostatic Discharge.pdf

6
Simulating Electrostatic Discharge
Dejan Maksimović, Guido Notermans
Abstract - In this paper we justify the necessity of the
electrostatic discharge simulation. We give an overview of the
ESD stress standards and the ESD protection devices. We further
describe the modelling of the ESD devices and give a case study
which shows the importance of timely ESD simulation for the
design success.
Keywords – ESD, simulation, modelling, HBM, MM, CDM,
HMM.
I. INTRODUCTION
Electrostatic discharge (ESD) occurs between two
bodies at different electrostatic potentials. The charging of
these bodies can occur either by triboelectricity or by
induction. ESD is characterized by a short duration (0.1ns
to 100ns), high current (1A to 30A) pulse. Such high
current can damage the semiconductor device. Failures can
be thermally induced due to the high power dissipated
during the ESD event: silicon melting can be observed as
well as metal or polysilicon resistance blow-up if the
metal/poly line is not designed wide enough. Gate oxide
breakdown can also be observed due to the large voltage
drop bult-up by the ESD current.
ESD can occur any time in the life of the product:
during manufacturing, assembly, testing, shipment and in
the final application and it is a key issue for the reliability
of the integrated circuits (ICs).
Two approaches are used together to fight against the
ESD. The first one is to prevent the ESD events. Special
dissipative materials are used in clean rooms and labs,
ionizers, proper grounding of the equipment, wearing of a
wrist strap during the tests etc. The second approach is to
implement efficient ESD protection on the IC. The ideal
ESD protection circuit is similar to a switch: it is highly
resistive during the normal operation of the IC, but it is
able to detect an ESD event and to become low resistive
when it occurs. In such a way the ESD device shunts the
ESD current with the lowest possible voltage drop.
II. MODELING THE ESD EVENTS
There are many ESD models, three of them being the
most widely used: human body model (HBM), machine
model (MM) and charged device model (CDM).
2A. Human Body Model (HBM)
This model corresponds to the discharge of a charged
human being into the IC. The capacitance of the average
human body (to ground) is 100pF. The average skin
resistance is 1.5kOhm. The body capacitance is charged to
a certain voltage level and discharged throuth the skin
resistance and the device under test (DUT) to the ground.
The electronic circuit representing this event is shown in
Figure. 2.1.
Fig. 2.1. Human Body Model (HBM).
This ESD model is defined by the JEDEC standard
JESD22-A114F [1].
2B. Machine Model (MM)
This model emulates the discharge that can occur in
automatic assembly lines between a machine and the IC.
The charged machine has a higher capacitance of 200pF,
while the contact resistance is very low, almost zero, often
considered as a few ohms. Due to the low resistance, this
model is strongly dependent on the parasitic inductance
which has to be fixed and is in the order of 0.5uH. The
electronic circuit representing this kind of ESD event is
shown in Figure. 2.2.
Dejan Maksimović and Guido Notermans are with the
ST-Ericsson, Binzstrasse 44, 8045 Zurich, Switzerland, E-mail:
dejan.maksimovic@stericsson.com,
guido.notermans@stericsson.com .

Proceedings of Small Systems Simulation Symposium 2010, Niš, Serbia, 12-14 February 2010
7
: 0
Fig. 2.2. Machine Model (MM).
This ESD model is defined by the IEA/JEDEC standard
IEA/JESD22-A115-A [2].
2C. Charged Device Model (CDM)
This model emulates the discharge of a charged IC to
the ground which occurs when one pin of the IC touches a
grounded surface. The whole IC is charged and the
discharge is determined by many device parameters such as
the package type and the die size. The CDM event is very
short (rise time is less than 0.5ns) high current pulse in
order of tens of amperes. It mostly causes the gate oxide
failures due to the overvoltage caused by such a high
current. A typical CDM current waveform is shown in
Figure 2.3.
Fig. 2.3. Charged Device Model (CDM).
During the CDM test the device is placed in a "dead
bug" position on a charging plane connected to a high
voltage source. Above the device there is a ground plane.
The discharge occurs when the pogo pin connected to the
ground plane touches one pin of the IC. This ESD model is
defined by the JEDEC standard JESD22-C101C [3].
While the HBM and MM events occur between two
pins of the IC and the circuit can be designed in such a way
that the ESD current path is predictable, during the CDM
stress the current comes from the silicon substrate and
distributes in an unpredictable way through the metal lines
and devices towards the stressed pin.
2D. System Level ESD Stress (Gun Test)
This model corresponds to the "real-world" discharge
that happens when the final user handles the product that
contains the IC. The stress levels are much higher and the
test is performed using the ESD gun. The discharge is
applied to every possible exposed surface of the product,
such as metal connectors, displays, case, etc. The ESD
current flows from the stressed point to the system ground
(and another way around), which is similar to the CDM
discharge. The current waveform is shown in Fig. 2.4. It
consists of a very short CDM-like first pulse of very high
amplitude and a HBM-like second pulse of the amplitude
higher than that of the HBM pulse for the same stress level.
The gun pulse parameters for different stress levels are
given in Table 2.1.
Fig. 2.4. Gun test current waveform.
TABLE 2.1
GUN TEST CURRENT WAVEFORM PARAMETERS FOR
DIFFERENT STRESS LEVELS
TABLE 2.2
HBM PEAK CURRENT VERSUS THE FIRST PEAK AMPLITUDE
IN THE GUN TEST
Applied
voltage [kV]
HBM peak
current [A]
Gun test first peak
current [A]
2 1.33 7.5
4 2.67 15.0
6 4.00 22.5
8 5.33 30.0
10 6.67 37.5
This ESD model is defined by the IEC standard 61000-
4-2 [4]. Table 2.2 compares the maximum peak current
during the HBM and gun test for the same voltage levels
[5].
2E. Transmission Line Pulse (TLP) Measurement
The ESD tests described so far are pass/fail
measurements. They do not give any information on the
behaviour of the IC during the ESD event. To obtain I(V)
characteristics of the ESD protection circuits and devices a
special tool called TLP is used [6].
During the TLP measurement, the DUT is subjected to
a trapezoidal positive current pulse. Once the transients in

8
the device are over, the current through and the voltage
over the DUT are measured. This produces one I/V data
point. To check if the device is still not damaged, the DC
leakage through the DUT is measured after each I/V data
point extraction. If no degradation is observed, the
amplitude of the current pulse is increased and the next
data point is measured. In this way the I(V) curve of the
device can be constructed starting from the low ESD
currents and finishing after the device is damaged.
The TLP can vary the rise time and/or the width of the
current pulse. The most commonly used parameter values
are Tr=10ns, Tw=100ns. The I/V data point is usually
taken at 90% of the Tw, i.e. after 90ns.
The current and voltage waveforms during the TLP
measurement look like those shown in Fig. 2.5. The
waveforms in Fig. 2.5 were obtained by the electrical
simulation though.
Fig. 2.5. The TLP waveforms obtained by the simulation. The
measurement point at 90% of pulse width is shown.
A typical measured TLP curve of an ESD device is
shown in Fig. 2.6.
2F. Definition of ESD Parameters
The most important ESD parameters are shown on the
TLP curve in Fig. 2.6. They are as follows:
Vt1: Trigger voltage. This voltage must not be higher
than the (gate oxide or PN junction) breakdown voltage of
the circuitry connected in parallel to the ESD device. In
case the device does not exhibit the snap-back (like in the
case of the diode), this parameter is sometimes called Von.
Vh: Hold voltage. Minimum voltage on the ESD device
is important since if it is below the supply voltage of the
IC, the ESD event can cause the latch-up (large leakage
from the supply to the ground that can be stopped only if
the supply is turned off).
It2 – Failure current. This is the maximum ESD current
that the ESD device can conduct without being damaged.
Vt2 – Voltage at failure level. This parameter is
important only if Vt2>Vt1.
Vh Vt1 Vt2
It2
Fig. 2.6. Typical TLP characteristic of a ggNMOST with the
ESD parameters marked.
2G. Correlation between different ESD stresses and TLP
A 100ns current pulse delivers a thermal stress
equivalent to the HBM stress. A good correlation has been
reported between the TLP It2 value and the HBM fail level:
the device which fails at 2kV HBM level will also fail at
1.33A TLP current if 100ns wide pulse is applied. Hence,
there is a correlation factor of approximately
2kV/1.33A=1.5kVHBM/ATLP between the It2 in TLP curve
and the HBM fail level.
The HBM and MM produce similar IC failures and
there is a relation between them: typically 2kV HBM level
corresponds to 100V MM level (ratio 20 times).
Consequently, the correlation factor between the MM fail
level and the It2 is approximately 75VMM/ATLP.
The relation between IEC and TLP stress is: 1A TLP
corresponds to 600V IEC gun test [7]. This results in 13.3A
TLP current for 8kV IEC.
III. ESD PROTECTION DEVICES
Early in the design process, the ESD protection concept
is being discussed and planned. There are two main
approaches: local ESD protection and rail-based ESD
protection. If the ESD protection is implemented locally,
then each pin of the IC has its own ESD protection element

9
(also known as "local ESD clamp"). In rail-based concept,
the protection elements are connected only between the
power and ground rails (rail clamps). The ESD current is
diverted from each input/output (I/O) pin using the ESD
diodes connected between the I/O pin and the power rail
and between the I/O pin and the ground rail.
There are many different types of ESD protection
devices. Some of them will be shortly described here.
3A. Diodes
Diodes are used in forward bias to conduct large ESD
currents. The forward biased diode can conduct between
5mA and 30mA per micrometer of the junction perimeter
(typically 10ma/um). In the inverse (Zener) breakdown the
diode can however conduct very small ammount of ESD
current.
The problem that often arises when one wants to
simulate the ESD is that the diodes are not properly
modelled in forward bias. This because the diodes are very
rarely used in functional part of the IC – they are only the
parasitic junctions and are supposed to always remain
inverse biased. ESD diodes are however an exception to
this rule.
3B. Snap-back devices
The most popular snap-back devices are ggNMOST and
thyristor (also known as silicon-controlled rectifier – SCR).
They have the I(V) characteristic similar to that in Fig. 2.6.
Due to the negative resistance region in the curve, they
cannot be simulated. SCRs have much deeper snap-back
(lower Vh) than the ggNMOSTs.
It is important to say that each NMOST (or PMOST)
can be driven into the bipolar mode and experience the
snap-back. For this to happen, the voltage between their
drain and source must reach the inverse breakdown of the
drain/bulk junction. Of course, this is an unwanted event
and it is sometimes possible to simulate the surrounding
circuitry to discover if the critical voltage is reached on the
MOS transistor.
3C. BigFETs
If a large NMOS transistor is conducting in parallel to
the protected circuitry during the ESD pulse, it can take
over all the ESD current and such protect the rest of the
circuit. This ESD device is known as bigFET (or RC
triggered FET) and is used as a rail clamp in most of the
modern CMOS technologies. BigFET triggers at around 1V
(slightly higher than a diode) and has a special RC circuit
which is responsible to switch it off after the ESD pulse is
over. Since all the circuitry in this clamp is working in
normal MOS operation mode, the device can be simulated.
IV. ESD SIMULATION
To be able to simulate the ESD event two models are
necessary:
- the model of the ESD current pulse
- the model of all the ESD (and other) components in
the circuitry connected to the stressed IC pin.
Both models can be easily developed for the analog
simulator. The stimulus generator is made using a circuit
such as that in Fig. 2.1 and 2.2. The waveform is adjusted
to that prescribed by the corresponding ESD test standard.
The ESD devices are usually produced on a test wafer
and TLP measured to extract their electrical parameters.
Then the electrical model is built using these parameters
and used in ESD simulation.
A few examples of ESD simulations on the device level
will be given in following paragraphs.
4A. Generating the TLP characteristic of the bigFET clamp
The TLP tester is used for on-wafer measurements. The
test circuit often contains additional resistance of metal
connections from the ESD device to the measurement pads.
Therefore the measured TLP curve has higher resistance
than the ESD device itself.
If the ESD device is of the kind that can be simulated,
such as a bigFET clamp, the simulation can show us the
real resistance. The Fig. 4.1 shows a simple simulation
testbench with such a bigFET rail clamp.
Fig. 4.1. Testbench for generating the TLP characteristic of the
clamp by the simulation.
The trapezoidal current generator and the 50ohm
resistor model the TLP system. The current and voltage
waveforms obtained by the simulation are shown in Fig.
2.5. The parametric analysis is used to change the
amplitude of the current pulse with constant step. Same as
in TLP system, the I/V points are measured at 90ns
simulation time.

10
The resulting TLP curves of a few different types of
clamps are shown in Fig. 4.2. It must be stressed that the
transistor models are valid only in the nominal supply
range (up to 5V in this case). The simulations we did here
by far exceed this range. This means the curves can be
trusted only up to 5V. Nevertheless, the TLP measurements
and long experience in this technology process show us
that the clamp does not change the resistance until very
close to the destruction level. The failure level can only be
determined by the TLP measurement on the silicon. The
simulation cannot give us this information. However, even
the TLP can measure only up to 10A and above this point
we cannot judge the behaviour of the devices which are
designed for higher currents.
Fig. 4.2. Simulation waveforms.
To eliminate the simulation artefacts (invalid simulation
results due to too high voltages), we can interpolate the
clamp curves obtained by the simulation in the valid range
as shown in Fig. 4.3. This enables us to estimate the
voltage drop on the clamp in case of different ESD stresses
and help us decide if we need to use more than one clamp
in parallel to reduce the voltage.
Fig. 4.3. Linear interpolation of the TLP curves generated by the
simulation outside of the valid range of the circuit models.
B. Simulating a bigFET clamp under the CDM stress
The testbench schematic for the simulation of the CDM
stress influence on the clamp is shown in Fig. 4.4. The
sinewave current generator emulates the CDM pulse. The
CDM pulse is approximately 2ns long positive pulse of
certain amplitude. The negative half-period of the used sine
generator is not important – the reverse diode of the clamp
will conduct during it. The results of the simulation are
shown in Fig. 4.5.
First of all, during the negative half-period of the
current pulse, the simulation shows only 1V over the
reverse diode of the clamp. This is definitely an
underestimation – the diode model is obviously not valid in
forward bias.
Fig. 4.4. A simple simulation testbench for the CDM stress.
Fig. 4.5. CDM simulation waveforms.
During the positive half-period of the current pulse the
voltage on the clamp shows a short, very high peak. This
peak is the consequence of the bad design of the clamp.
The clamp cannot trigger fast enough to conduct the CDM
pulse. Some time later, the clamp triggers and this pulse
falls down (clamp resistance is reduced) to 2V. This
voltage is realistic to expect for this kind of the clamp
when conducting a 4A current.
From this simulation we can conclude that the clamp
triggering is not fast enough if 4A CDM peak current is to
be expected from the IC.
V. A CASE STUDY
A simplified schematic of the output stage of the radio
antenna in an IC is shown in Fig. 5.1. The power amplifier
has a PMOST connected between the VDD_PA and the
antenna pin Int_Ant. The bulk diode of the PMOST
(Dpmost) is also shown in the schematic. The wire
resistance between the PMOST drain and the Int_Ant pad
is around 2ohm. The voltage at the antenna pin can vary
between -2V and +2V. Therefore, the four-diode string
(D11-D14 and D21-D24) is used as the ESD protection on
this pin. The rest of the circuit is protected with a standard

11
rail-based ESD concept, which assumes the rail clamps
(bigFETs) between the power (VDD_PA) and ground
(VSS_PA) rail. In the shown voltage domain two such
clamps are connected (C1 and C2). Each rail clamp has the
reverse diode (DrC1 and DrC2).
VDD_PA
Int_Ant
VSS_PA
DrC1
C1
PMOST
Dpmost
D11
C2
DrC2
D12
D13
D14
D21
D22
D23
D24
2ohm
Fig. 5.1. ESD protection and functional circuitry at the antenna
pin of an FM radio IC.
The Int_Ant pin is supposed to withstand 8kV system
level stress (gun test) without adding any additional
external ESD protection on the application board. The
intended ESD current path for the current between Int_Ant
and VSS_PA is through the 4-diode string (depicted with
red line in Fig. 5.1). There is however the alternative path
through Dpmost and clamps C1/C2, as shown by the
dashed green line in Fig. 5.1. This unwanted current could
damage some of these devices. To prevent this, the circuit
was modeled and simulated before the production.
The 4-diode strings are designed to be able to conduct
14A of the 100ns TLP current. Their TLP characteristics
measured on a separate test wafer are shown in Fig. 5.2.
The TLP tester could produce maximum current of 10A,
hence the diodes could not be damaged. The diode curve
was estimated up to 14A by removing the connection
resistance (depicted red in Fig. 5.2). The simulation model
was generated with Von=4V and Ron=0.25ohm, same in
both directions.
Fig. 5.2. Measured TLP characteristics of the 4-diode string, two
samples measured in both directions on a test silicon.
In a similar way the rail clamps C1 and C2 were
measured and their electrical model was created. In the
forward direction the clamps have Von=0.35V and
Ron=1ohm. In the reverse direction (DrC1 and DrC2) the
equivalent circuit contains Von=0.6V and Ron=1ohm. The
bulk diode of the PMOST was also modeled by Von=1ohm
and Ron=1ohm. The resulting simulation model of the
circuit is shown in Fig. 5.3.
Fig. 5.3. Simulation model of the system for the positive IEC
pulse at Int_Ant pin.
The 8kV IEC stress consists of the first short 30A
current peak which can generate the overvoltages in the
circuit and the second 16A current peak which can damage
the components by inducing thermal damage. One simple
approach to the simulation is to use a DC current source of
a maximum current value (30A and 16A) and check the DC
currents that flow through all the circuit elements.
The result of such a simulation is shown in Fig. 5.4.
The second 16A current pulse was brought to the circuit.
The simulation shows that there are no large overvoltages
in the circuit nodes. The maximum node voltage is
V(Int_Ant)=7.44V, which is lower than the breakdown
voltage (8.2V) of the gate oxide connected to this node.
VDD_PA
Int_Ant
VSS_PA
DrC1
C1
PMOST
Dpmost
D11
C2
DrC2
D12
D13
D14
D21
D22
D23
D24
2ohm
16A 13.76A
2.24A
7.44V
1.72V
Fig. 5.4. Simulation model of the system for the positive IEC
pulse at Int_Ant pin.
The maximum of 13.76A flows through the diode string
D21-D41. This is safe since the diodes were designed for
14A. The simulation also shows that 2.24A flows through
the unwanted path – through the Dpmost and two clamps
C1/C2. Based on the size of the PMOST (720um) it can be

12
estimated that Dpmost can stand as much as
720um*10mA/um=7.2A. Hence, Dpmost will not be
damaged. Two rail clamps C1/C2 are designed to be able to
survive as much as 4A of 100ns TLP current (to stand
200V MM stress). Obviously, they will also be able to
conduct 2.24A current.
The 2ohm resistor represents the estimated resistance of
the relatively long and thin metal connection between the
Int_Ant pad and the PMOST. Unfortunately, this
connection was not carefully reviewed during the ESD
review of the circuit before the production. As a result, the
gun test on the produced IC showed circuit failure at 6.5kV
stress level instead of targeted 8kV. The failure analysis
(FA) of the failing samples showed that the connection
towards the PMOST was lost. The weakest point on the
metal connection were three 1.9um wide lines in M3 which
melted, as visible in Fig. 5.5.
PMOST
to pad
Fig. 5.5. The failure analysis result, three melted M3 lines at
PMOST's drain.
A M3 line in the used 45nm process can conduct
maximum 225mA of the 100ns TLP current per
micrometer of the line width. That means that these three
parallel M3 wires could conduct maximum
3*1.9um*225mA/um=1.28A, which confirms the root
cause of the fail.
VI. CONCLUSION
In this paper we defined the pre-requisites for ESD
simulation. The modelling of the ESD devices is explained
and a case study is given that proves the usefulness of the
ESD simulations. In this case, we show that quite simple
DC simulation of the circuit is able to explain the root
cause of the gun test fail.
One should be careful when simulating ESD since the
currents and/or voltages often exeed the valid range of the
component models. Therefore it is always recommended to
calibrate the simulation results using the measured TLP
characteristics of the silicon test structures.
The bigFETs (which are the most popular ESD devices
in technologies below 90nm) can be accurately simulated.
The simulation of the pad-ring is obligatory when defining
the ESD rules in the rail-based ESD concept.
Only the snap-back ESD devices cannot be simulated
due to the negative resistance region in their I/V curve.
They can be modelled by the Vt1 and Ron, but the designer
must be aware that the simulation result becomes invalid as
soon as the voltage on the snap-back device exceeds Vt1.
REFERENCES
[1] Electrostatic Discharge (ESD) Sensitivity Testing
Human Body Model (HBM), JEDEC standard JESD22-
A114F, December 2008.
[2] Electrostatic Discharge (ESD) Sensitivity Testing
Machine Model (MM), IEA/JEDEC standard
IEA/JESD22-A115-A, October 1997.
[3] Field-Induced Charged-Device Model: Test Method
for Electrostatic-Discharge-Withstand Thresholds of
Microelectronic Components, JEDEC standard
JESD22-C101C, December 2004.
[4] Electromagnetic compatibility (EMC) – Part 4-2:
Testing and measurement techniques – Electrostatic
discharge immunity test, IEC standard 61000-4-2,
edition 2.0, December 2008.
[5] Human Body Model (HBM) vs. IEC IEC61000-4-2,
California micro Devices, white paper, January 2008.
[6] J. Barth, K. Verhaege, L. G. Henry, J. Richner, "TLP
calibrarion, correlation, standards and new
techniques", Proc. EOS/ESD Symposium, pp. 85-96,
2000.
[7] T. Smedes, J. van Zwol, G. de Raad, T. Brodbeck, H.
Wolf, “Relations between system level ESD and
(vf-)TLP", EOS/ESD Symposium Proc., 3A1, pp. 136-
143, 2006.

02 Simulating Electrostatic Discharge.pdf

More Related Content

Similar to 02 Simulating Electrostatic Discharge.pdf (20)

Recently uploaded (20)

02 Simulating Electrostatic Discharge.pdf