lec14 DTMOS
Transcript of lec14 DTMOS
With advances in technology scaling, CMOS circuits are
increasingly more sensitive to transient pulses caused by
single event (SE) particles.
The reduced nodal capacitances, supply voltages and
transistor drive currents coupled with more dense chips are
increasing soft error rates and making them an important
design issue.
Effect of Power Optimizations on Soft Errors
Another important issue a circuit designer faces today is
the increasing power consumption. As technology scaling
continues, the leakage or static power consumption of a
circuit increases.
In order to reduce power consumption in chips, one
common approach has been the scaling of the
power supply voltage. This should be accompanied
by threshold voltage reduction in order to keep
circuit speed and high current drive to avoid
performance penalties.
However, the threshold voltage reduction results in
great amount of increase on the sub-threshold leakage sub-threshold leakage
current current
Subthreshold leakage is the current that flows between the
source and drain of a MOSFET when the transistor is in
subthreshold region, or weak-inversion region ( for gate-to-
source voltages below the threshold voltage)
The transistor in this region is not completely turned
OFFOFF. Instead, a small current still flows between the
source and drain terminals.
Dynamic power dissipation resulting from continuous
switching of transistors and charging and discharging of
capacitances, previously contributed the most to the total
chip’s power dissipation in CMOS digital circuits thus
most efforts were devoted towards reducing it.
LEAKAGE (STATIC) POWER
Load cap. Consists of diffusion capacitance of the drains of the inverter transistors, the total interconnect capacitance, and the input gate oxide capacitance of the driven gates that are connected to the inverter's output.
Dynamic power dissipation
The average dynamic power dissipation of the
inverter can be calculated from the energy
required to charge up the output node to VDD and
discharge the total output load capacitance to
ground (GND). The generalized expression for the
switching power dissipation of a CMOS logic gate
can be written
Pavg = αT Cload VDD 2 fCLK;
αT is the switching activity factor of the gate
But as technology advances into the sub-100nm
regime, leakage power dissipation, which is a static
power, increases at a much faster rate than dynamic
power and it is expected to dominate the chips’ total
power dissipation.
But as technology advances into the sub-100 nmregime, leakage power dissipation increases in a much faster rate than dynamic power.
we started to see these effects with DSM technology
Leakage power is static power dissipation and can be described as the power that is dissipated through transistors without producing any useful outcome.
For example, even if the gate voltage of a transistor is lower than the threshold voltage required to turn it ON, the transistor is not completely off and a small current still flows causing what is known as sub-threshold leakage current.
The second term corresponds to the static power
dissipation where Ileak corresponds to the
total static leakage current.
Total power dissipation then can be written as:
The active and leakage power
trends for process technologies
Since leakage power dissipation is approaching 40%
of today’s high performance microprocessors total
power dissipation, new techniques and new
technology innovations are urgently needed to
reduce leakage current components.
The continuous scaling down of technology minimum
feature size increases the Short Channel Effects.
A MOSFET device is considered to be short when
the channel length is the same order of magnitude
as the depletion-layer widths (xdD, xdS) of the
source and drain junction.
In short channel devices source and drain depletion
regions penetrate significantly into the channel and
control the potential and the field inside the channel.
This is known as the short channel effect
Due to the short channel effect, Vth reduces with
reduction in channel length (Vth roll off), and with an with an
increase in the drain biasincrease in the drain bias (Drain Induced Barrier
Lowering (DIBL))
If the drain voltage is increased, the potential barrier in
the channel decreases, leading to drain-induced barrier
lowering (DIBL).
The reduction of the potential barrier eventually allows
electron flow between the source and the drain, even if even if
the gate-to-source voltage is lower than the threshold
voltage..
The channel current that flows under this conditions
(VGS<VT) is called the sub-threshold currentsub-threshold current.
To effectively control channel, reducing gate oxide
thickness tox helps in controlling Short Channel Effects,
but the reduction in tox results also in higher gate leakage
current.
Large leakage power can lead to dramatic effects thatmay cause circuits to fail to function properly. Large leakage currents will:
■ Degrade noise immunity in dynamic circuits which may lead to circuit failure.
■ Increase the standby power dissipation to unacceptable levels.
■ Lead to excessive heating which may make components susceptible to failure and, as stated above, will consequently require complicated cooling and packaging techniques.
Even they affect IDDQ testing
Leakage current is also an important factor in setting total supply current (IDDQ) value, which is used as a
pass/fail threshold for testing chips.
IDDQ value is now determined by the sum of leakage
currents of those transistors that can leak.
Setting this value too high would cause some defective
chips to be considered non-defective; similarly if this
value is set too low, functional chips may be considered
defective.
Effect of Channel Length
In short channel transistors, the drain and source depletion
regions are in a near proximity of each other. They enter
more into the channel length and as a result part of the
channel gets depleted.
Consequently, less gate voltage is needed to turn the
transistor on. This means that the threshold voltage is lower
for the short channel transistor and sub-threshold current is
higher.
Equation shows the various parameters that affect the threshold voltage of a transistor such as doping densities, source to substrate bias, and some other process parameters:
Vth0 is the zero-bias threshold voltage and depends on
the manufacturing process,
γ is the body effect coefficient (typically equals to 0.4 V0.5) and it depends on the gate oxide capacitance, silicon permittivity, doping.
φF is the surface potential at threshold (typically |− 2φF | equals 0.6 V);
VSB is the source-to-body voltage;
ηVDS and the term represents the effect of Drain-
Induced Barrier Lowering (DIBL)
η is the DIBL coefficient and it is in the range of 0.02–0.1
Body bias:
Increasing the source-to-body biasing results in increasing the transistor’s threshold voltage as indicated by equation below and this effect is known as the Body Bias effect
Vth = Vth0 + γ |2φF| + VSB − |2φF | −
ηVDSand hence it reduces the sub-threshold leakage
current.
The increasing use of portable devices and higher demand
on battery life has made power consumption even more
important in today’s CMOS designs.
Subsequently, there have been several efforts spanning
from the circuit level to the architectural level at reducing the
consumed energy.
[1] L. Wei, K. Roy, and V. K. De. Low voltage low power CMOS design
techniques for deep submicron ICs. In VLSI Design, 2000. Thirteenth
International Conference on, pages 24-29, 2000.
[2] K. Roy, S. Mukhopadhyay, and H. Mahmoodi-Meimand, “Leakage current
mechanisms and leakage reduction techniques in deep-submicrometer CMOS
circuits”, Proceedings of the IEEE, 91(2):305-327, 2003.
Circuit mechanisms include adaptive substrate biasing,
dynamic supply scaling, dynamic frequency scaling, and
supply gating.
Circuit Level Leakage Reduction Techniques
Leakage reduction circuit techniques can be put into four categories :
A. Transistor stacking TechniquesB. Multi-Vth TechniquesC. Input Vector Control TechniquesD. Body Biasing Techniques
A.Transistor Stacking Techniques
All these techniques are based on the fact that when
there are two or more stacked transistors which are
switched OFF, we will have approximately an order of
magnitude lower subthreshold leakage.
Transistor stacking has a dual effect in reducing the
sub-threshold leakage current, it increases the source
bias of upper transistors in the stack and also lowers
the gate-to source voltages of these transistors,
A.Transistor Stacking Techniques
Reducing leakage through the use
of transistor stacks depends on the depends on the
choice of input pattern choice of input pattern during
standby periods since it determines
the number of OFF transistors in the
stack.
Serious connected ‘off’ Serious connected ‘off’ Transistors (Transistor Stacking) Transistors (Transistor Stacking) can reduce leakage current can reduce leakage current greatlygreatly
Leakage currents dependencies on circuit state canbe exploited and used to determine a low leakage state by searching the minimum leakage input vector, which is fed into the circuit during sleep mode
After this input vector is found, the circuit is evaluated and additional leakage control transistors are inserted in the non-critical paths where only one transistor is originally turned OFF.
B. Multi-Vth Techniques
This is one of the most common approaches to reduceleakage currents where two different types of transistorsare fabricated on the chip, a high-Vth to lower subthresholdleakage current and a low-Vth to enhance circuit performance by increasing its speed.
Obtaining two different sets of transistors each with either a low or a high is usually done by changing the channel doping profile, having two different oxide thicknesses, or by changing channel lengths
It is designed using an NMOS and a
PMOS transistor that have different
threshold voltages than the CMOS
logic circuit.
The CMOS logic circuit is designed
with low thresholds as usual to
ensure speed and performance.
The separate NMOS and PMOS transistors will have a high threshold voltage and will not suffer significant leakage currents.
These will be used to isolate the CMOS logic circuit by being placed in series with the CMOS structure and in essence will isolate the logic circuit in stand-by mode and prevent any significant leakage power.
the stand-by transistor design the stand-by transistor design creates a virtual power supply creates a virtual power supply and/or virtual ground.and/or virtual ground.
There are two families of the MTCMOS technique,the first family uses a high-Vth device to supplyvoltages TO a low-Vth logic block thus creating avirtual power rail instead of directly connecting the blockto the main power rail as shown.
The voltage continues scaling down with every generation, employing a high-a high-Vth Vth would not be practicalwould not be practical with supply voltages equal or lower than 0.7 V. Other techniques such as Super Cutoff CMOS (SCCMOS) provide a solution to this potential problem.
SCCMOS technique is another multi-Vth circuit technique used to reduce leakage currents.
In this technique, instead of using a high-Vth transistor to
provide the supply voltage, a low-Vth PMOS switch is
used whose gate is overdriven to VDD + 0.4 V through
the use of a MOS charge pump (to fully cut-off leakage)
The Super Cutoff CMOS (SCCMOS) technique
ALSO:
C. Input Vector Control Techniques
The basic idea behind input vector control techniques forleakage reduction is to force the circuits’ combinational logic into a low-leakage state during standby periods.
This low-leakage state forces the largest number possible of transistors to be turned OFF so as to reduce leakage and make use of multiple OFF transistors in stacks. Leakage current can be reduced to 10 times lower if input vector control is used
An alternative approach is to exploit transistor stacks
already embedded in the logic, by preferring to put each
gate in a low leakage state, for example, by inputting all
0’s for a NAND gate. Controlling each gate directly is
too expensive in terms of area and power, so typically
one wants to find an input vector (to a combinational
circuit) that minimizes total leakage current.
C. Input Vector Control Techniques contd. There are three main methods for selecting the required input vector:
(1) Analyzing the circuit and looking for a good inputvector;
(2) Employing an algorithm that searches forthe best input vector;
(3) Simulating the circuit with a large number of input test patterns and selecting the one that results in the lowest leakage power among these test patterns.
D. Body Biasing Techniques
A) Variable Threshold CMOS (VTCMOS)
To overcome subthreshold leakage, the threshold
voltage can be adjusted. This is accomplished by
changing the substrate-bias voltage as noticed in
Equation
In Variable Threshold CMOS (VTMOS)
technique, deep reverse body bias is applied during deep reverse body bias is applied during
standby, standby, a higher than VDD bias for PMOS transistors
and lower than GND bias for NMOS transistorslower than GND bias for NMOS transistors, to
further increase the threshold voltage and further push
the transistors in the OFF region to achieve lower
subthreshold leakage currents;
Body bias applied in the reverse bias regime, where threshold
voltage increases as body-to-source reverse bias is made
larger to reduce leakage.
1V
3V3V
-2V-2V
For controlling substrate
voltage, triple well technology
or SOI needed
-2V
3V
As mentioned earlier, whenever VDD is scaled to
reduce power consumption, it should be followed by
threshold voltage reduction in order to keep circuit
speed and high current drive to avoid performance
penalties.
However, the threshold voltage reduction results in
great amount of increase on the sub-threshold
leakage current
B) Dynamic Threshold MOS (DTMOS)B) Dynamic Threshold MOS (DTMOS)
.One solution to this problem is a dynamic threshold
(DTMOS) operation which applies an active body-bias to
MOSFETs
Technique applies a low threshold voltage during the
logic transition and high threshold voltage during the off-
state, the dynamic threshold circuit operates at high
speed with low power.
Dynamic threshold technique however can only be used
for low voltage (0.6 V and below) VLSI circuits.
In a Standard DTMOS logic gate, all transistors’ gates are
tied to their substrates
Namely, the body-source junction is “forward biased” (at less than
0.6 V), forcing the threshold voltage to drop.
0.6 V
Logic High
VSB=0-0.6= -0.6 V
NMOSNMOS threshold reduces while it
is switching which makes it faster
In comparison to VTCMOS, DTMOS operates in the
exact opposite regime.
0.6
0.6 V
Logic High
0 V
Logic LOW High threshold (normal
threshold applied to
OFF NMOS)
NMOS is ON. NMOS
threshold reduces via
body to source voltage.
negative
The high speed operation is provided by
forward bias to switching transistors, while
low leakage is obtained by applying zero
bias to other transistors
During switching, the body-source junction is “forward
biased” (at less than 0.6 V), forcing the threshold voltage
to drop.
PMOS gets 0 bias having low leakage while
NMOS gets a forward bias speeding up
All of these power optimizations AFFECT the susceptibility
of the circuits to soft errors. As circuit designers try to
address the excessive power consumption problem, they
need to be aware of the impact of the power optimizations
on circuit SE robustness.
This sort of analysis is important because it helps us in
making clever design choicesclever design choices that reduce static power
consumption and improve soft error reliability of the newer
designs.
There has been some work to examine the effect of
power optimizations on the soft errors.
Degalahal et al. have analyzed the effect of increasing effect of increasing
threshold voltagethreshold voltage (widely used for reducing static power
consumption) on circuit soft error rate (SER) and found
that increasing threshold voltages cause devices to devices to
slow downslow down which in turn increase CL circuit increase CL circuit
susceptibilitysusceptibility to Single Event transients.
V. Degalahal, R. Ramanarayanan, N. Vijaykrishnan, Y. Xie, M. J. Irwin, "The
Effect of Threshold Voltages on the Soft Error Rate", 5th International
Symposium on Quality Electronic Design (ISQED'04), pp. 503-508, 2004.
For increasing CL soft error tolerance, Dhillon et al.
have proposed a technique that used optimal
assignment of gate sizes, supply voltages, and
threshold voltages while meeting timing constraint.
Y. S. Dhillon, A. U. Diril, A. Chatterjee, and A. D. Singh, “Analysis
and optimization of nanometer CMOS circuits for soft-error
tolerance,” in IEEE Trans. on Very Large Scale Integration (VLSI)
Systems, Vol. 14, No. 5, pp. 514-524, 2006.
Later work has presented a power-aware methodology for
soft error hardening using dual supply voltages. In this
methodology, a higher supply voltage is assigned to the
gates that have large error impact and contribute most to
the overall SER.
Kai-Chiang Wu and Diana Marculescu, "Power-Aware Soft
Error Hardening via Selective Voltage Scaling", Proc. of Int'l
Conference on Computer Design (ICCD), pp. 301-306, 2008.
Recently, we have studied the radiation tolerance of Dynamic Recently, we have studied the radiation tolerance of Dynamic
Threshold based MOS Design (DTMOS) Technique.Threshold based MOS Design (DTMOS) Technique.
Various DTMOS inverter schemes have been proposed to
improve Standard DTMOS logic inverter. In order to
reduce standby leakage current, Chung et al. have
proposed a new scheme with minimum size small minimum size small
subsidiary transistors subsidiary transistors
The subsidiary transistors increase
the current drive by managing the
body bias
The input load of the inverter circuit
is reduced since the output charges
are used to raise the body potential
of the main transistors.
Later Gil et al. proposed another scheme with subsidiary
transistors that achieved better performance in terms of
speed
Another Dynamic Threshold Inverter Scheme
Below scheme is another DTMOS with subsidiary
transistors, but this time the gates of subsidiary devices
are tied to main transistor’s drain instead of gate
This configuration reported to perform best in terms of
power-delay product when compared to previous
schemes
Soleimani et al. [16] reported even better power
efficiency with the following
stability with respect to temperature was
an issue
In our initial analysis, we studied the effect of decreasing
threshold on SEU and soft delay tolerance. In order to
analyze the effect of decreasing the threshold voltage on
SE Transients and delay effects, we have used a string
of six inverters shown
ANALYSIS
For measuring SE Transients, the input of first inverter is
tied to logic 1, and for soft delay measurements, it is tied
to a switching input waveform. In our analysis, we have
considered 65 nm technology with parameters obtained
using Predictive Technology Model
A soft error occurs when the collected charge collected charge QQcolcol at a at a
particular node exceeds the critical charge particular node exceeds the critical charge QQcritcrit. When
this happens, the generated SE pulse can be latched,
resulting in bit flip in storage element or soft error.
The critical charge here can be defined as the minimum
charge collected due to a particle hit that can cause a
change in circuit output
In order to determine critical charge, the deposited
charge is slowly increased until the SET pulse appears at
the output of inverter string and hence cause a bit flip in
storage element.
Since our aim here is to examine the effect of decreasing
threshold on circuit SET, the threshold voltage value has
been modified via vth0 parameter using delvto option in
Hspice.
Results show that a 100 mV reduction from normal
threshold value increases Qcrit by 33%.
This indicates that if threshold voltage can be reduced,
the circuit robustness to SETs could be increased as the
critical charge value of a node becomes larger. This
confirms the results of Degalahal et al.
We have also studied the effect of threshold on soft delay
effect. This has not been done before to our knowledge.
For soft delay measurement, a pulse signal has been
connected to the input of the first inverter in the inverter
string
67
For soft delay measurement of a series of six inverters, a pulse input signal has been connected to the input of the first inverter. Since soft delay occurs when a particle strikes a sensitive node during signal transition, a radiation hit has been injected around halfway during falling transition at the output of first inverter using double exponential current source.
For soft delay calculation, the 50% delay at the output of
the last inverter gate is first recorded in the presence of
an SE charge. The delay measurement is then repeated
with SE current source removed (no SE charge). The
difference between the two is recorded as “soft delay” at
the output of the inverter string. In our simulations, we
have assumed the critical delay of the circuit is 200 ps
meaning any delay increase more than 200 ps results in
a timing violation and an SDE.
The minimum charge that creates this critical delay has
been recorded as critical charge Qcrit-delay. The result
shows that circuit robustness to soft delay effect also
increases with decreasing threshold voltage.
Fig. Critical Charge Qcrit vs. Vth for a soft delay of 200 pS.
Our initial analysis on threshold voltage has shown that
decreasing threshold voltages increase the critical
charge of logic circuits hence provide more robustness
to SE transients and Soft Delay effects. In normal
DTMOS scheme, the body-source junction is “forward
biased” (at less than 0.6 V), forcing the threshold
voltage to drop and hence this effect can be exploited
for Single Event mitigation. Therefore, in this work, we
study various Dynamic Threshold (DTMOS) schemes
for the purpose of mitigating SE transients and soft
delay errors.
DTMOS Schemes
Based on these results, we have examined various
dynamic threshold based schemes for their vulnerability
to SET and soft delay errors. These methods are widely
used for high speed and low power operations. Results
are then compared to that of Conventional configuration
where transistor body terminals connected to source
terminal.
Various Dynamic Threshold Inverter Schemes Reported
In the comparison, five different circuits including
benchmark circuits have been utilized. These circuits are
6-stage inverter chain, c17 ISCAS-85 and c432
ISCAS-85 benchmark circuit, Full Adder module in
ISCAS-85 c6288 and ALU module in AM2901 4-bit
microprocessor bit-slice.
Although it is dated, the ALU module alone contains 83
gates, 12 input and 10 outputs and 276 SE vulnerable
nodes.
Fig. ISCAS-85 benchmark Circuit
The sizes of all gates are taken as minimum size minimum size with power supply
voltage used in these configurations is 0.6V. The hit locations for
these circuits at nodes close to primary inputs. For example, in c17
benchmark circuit given, node “g7” was selected for hit location and
noise has been propagated to the output for observation. The
critical charge values Qcrit and Qcrit-delay for each circuit have
been determined using the procedure explained before.
ALU module in AM2901 4-bit microprocessorALU module in AM2901 4-bit microprocessor
Configuration
6-stage inverter c17 ISCAS-85 ISCAS-85 c432 ISCAS-85
c6288
AM2901ALU Module
Qcrit
(fC)
Qcrit-delay
(fC)
Qcrit
(fC)
Qcrit-delay
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Normal Body Tie 2.46 6.51 1.70 2.07 2.63 6.51 2.58 6.34 2.03 1.99
Standard DTMOS 3.72 11.05 2.64 4.56 3.76 10.30 3.79 10.86 3.01 3.74
Fig. 4.a [11] 3.27 7.96 2.31 2.65 3.42 8.14 3.48 7.82 2.87 2.53
Fig. 4.b [12] 3.53 8.33 2.50 2.55 3.59 7.94 3.64 7.98 3.44 2.56
Fig. 4.c [13] 2.66 10.00 1.80 3.55 2.84 9.42 2.84 9.46 2.16 2.84
Fig. 4.d [14] 3.22 9.39 2.12 3.24 3.34 9.26 3.36 8.81 2.56 3.02
CRITICAL CHARGE VALUES FOR VARIOUS DTMOS INVERTER SCHEMES
For all four example circuits considered, body tie design achieves the
smallest critical charge among all techniques and hence it is the most
vulnerable one to SETs and soft delay effects.
Configuration
6-stage inverter c17 ISCAS-85 ISCAS-85 c432 ISCAS-85 c6288AM2901
ALU Module
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Qcrit
(fC)
Qcrit-del
(fC)
Normal Body Tie
2.46 6.51 1.70 2.07 2.63 6.51 2.58 6.34 2.03 1.99
Standard DTMOS
3.72 11.05 2.64 4.56 3.76 10.30 3.79 10.86 3.01 3.74
Fig. 4.a [11] 3.27 7.96 2.31 2.65 3.42 8.14 3.48 7.82 2.87 2.53
Fig. 4.b [12] 3.53 8.33 2.50 2.55 3.59 7.94 3.64 7.98 3.44 2.56
Fig. 4.c [13] 2.66 10.00 1.80 3.55 2.84 9.42 2.84 9.46 2.16 2.84
Fig. 4.d [14] 3.22 9.39 2.12 3.24 3.34 9.26 3.36 8.81 2.56 3.02
CRITICAL CHARGE VALUES FOR VARIOUS DTMOS INVERTER SCHEMES
Standard DTMOS technique shows superior characteristics in terms of
SEU robustness due to highest critical charge for all cases. The critical
charge value needed for soft error in Standard DTMOS is 50% more than
what is required for normal body tie scheme, and hence more robust in
terms of SEU tolerance.
The reason can be explained as follows: Referring to Fig.
when input to DTMOS inverter is low, the body for PMOS
is low and hence PMOS has a forward bias. Since, this , this
increases the drivability of PMOS transistorincreases the drivability of PMOS transistor, a negative
SET pulse at the output can be easily dissipated due to
increased PMOS current drive.
A similar explanation can be made for the case that input is high and a positive SET pulse present at inverter output.
0.6 V
Logic High
For soft delay error, the critical charge values for Standard
DTMOS is more than 60% higher in all cases compared to
Normal body tie configuration. Let us assume the input
waveform is a rising from logic low to high. During the falling
waveform at the output of inverter, if a positive charge is
deposited by an SE particle, this may result in delay increase
or soft delay.
When input rises from Low to High, body of NMOS follows the input body of NMOS follows the input and as a result and as a result NMOS gets a NMOS gets a forward bias. forward bias. This causes output waveform to switch more rapidly and results in a reduced soft delay.
Hardening Against Combinational Circuit Related Soft Errors Hardening Against Combinational Circuit Related Soft Errors
Using DTMOSUsing DTMOS
Researchers have proposed many circuit level techniques
to mitigate SE Transients in CL:
The Spatial redundancy techniques such as Triple
Modular Redundancy (TMR) circuits triplicate the CL to
be protected and then use a voting circuit to filter out the
transient.
Temporal methods sample the data with different delays
and produce the output to a voting circuit to eliminate an
SET pulse.
The driver sizing technique proposed increases device
capacitance and drive current to decrease the device
vulnerability to SEUs
Larger drive strengths of NMOS and PMOS quickly
dissipates the collected charge, reduces the vulnerability
to Single Event particles.
Other techniques such as the one BELOW use different
circuit layouts at the gate and transistor levels to reduce
the chance of a SE to become a transient.
M. P. Baze, S. P. Buchner, D. McMorrow, A Digital CMOS Design
Technique for SEU Hardening, IEEE Trans. on Nuclear Science, vol. 43,
no.6, Dec. 2000.
Our proposed hardening technique is based on use of
Standard DTMOS scheme along with driver sizing. This
combine approach results in considerable area saving
compared to driver sizing alone. This is possible since a
Standard DTMOS gate is more SE robust compared to a
conventional one.
In sizing simulation, we consider the 6-stage inverter chain
consisting of conventional inverters and apply various
deposited charges in between 10-150 fC. For each charge
level, we determine the necessary hit inverter size to
eliminate the soft error effect. We size up the gate
transistors until the soft error is eliminated.
We then repeat the same for soft delay. In this case, the hit
driver should be sized up such that the delay should be
reduced to less than 200 ps. This process is then applied
for the Standard DTMOS inverters.
Above procedure has been repeated for all four example
circuits, and necessary gate sizes have been determined
to eliminate soft error and soft delay error effects.
Again, these circuits are 6-stage inverter chain, c17
ISCAS-85 and ISCAS-85 c432 benchmark circuit, Full
Adder module in ISCAS-85
Deposited Charge
6-stage inverter c17 ISCAS-85 ISCAS-85 c432 ISCAS-85 c6288AM2901
ALU Module
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
10 fC10 fC5X 3X
2X1X
7X5X
4X2X
4X3X
2X1X
4X3X
2X1X
5X4X
4X3X
25 fC25 fC11X7X
3X2X
17X11X
9X4X
10X7X
4X2X
9X6X
4X3X
11X8X
11X6X
50 fC50 fC21X15X
6X4X
34X22X
17X8X
19X14X
7X4X
17X11X
7X5X
20X16X
23X11X
100 fC100 fC42X29X
12X7X
67X44X
34X15X
38X28X
13X9X
33X17X
13X9X
36X29X
47X23X
150 fC150 fC62X43X
18X11X
100X66X
50X22X
57X42X
19X13X
50X21X
19X14X
51X42X
83X33X
GATE SIZES REQUIRED FOR SOFT ERROR AND SOFT DELAY
MITIGATION AT VARIOUS DEPOSITED CHARGES
Deposited Charge
6-stage inverter c17 ISCAS-85 ISCAS-85 c432 ISCAS-85 c6288AM2901
ALU Module
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
SEU Normal /DTMOS
SDE Normal /DTMOS
10 fC10 fC5X 3X
2X1X
7X5X
4X2X
4X3X
2X1X
4X3X
2X1X
5X4X
4X3X
25 fC25 fC11X7X
3X2X
17X11X
9X4X
10X7X
4X2X
9X6X
4X3X
11X8X
11X6X
50 fC50 fC21X15X
6X4X
34X22X
17X8X
19X14X
7X4X
17X11X
7X5X
20X16X
23X11X
100 fC100 fC42X29X
12X7X
67X44X
34X15X
38X28X
13X9X
33X17X
13X9X
36X29X
47X23X
150 fC150 fC62X43X
18X11X
100X66X
50X22X
57X42X
19X13X
50X21X
19X14X
51X42X
83X33X
Results show that Standard DTMOS technique can be used
along with gate sizing in mitigating the Single Event Transients
and Soft Delays with considerably less area overhead.
CONCLUSIONCONCLUSION
As designers address the static power consumption via
optimizations, they need to be aware of the impact on
SEU robustness.
We have analyzed various DTMOS schemes for their soft
error tolerance.
We have found that DTMOS increase circuit robustness
due to increased current drive and hence can be
combined with driver sizing technique to mitigate SET
transients and soft delay effects. This approach results in
considerable area saving compared to driver sizing alone.