Low Power-Area Design of Full Adder Using Self Resetting Logic with GDI Techn...
Thesis slides
1. A COMPARISON OF RADIATION
TOLERANCE OF DIFFERENT LOGIC STYLES
Thesis Defense
May 2011
Parthivkumar Prajapati
Thesis Advisor: Dr. Selahattin Sayil
The Phillip M. Drayer Department of Electrical Engineering
Lamar University
Beaumont, TX
12/25/2012 Prajapati 1
3. Introduction
Motivation
• With the continuous scaling of CMOS technologies,
the device reliability and power dissipation become
major issues.
• Designers are now selecting different logic gate
implementations to save the complexity of design
and to minimize power issues.
• The sensitivity of semiconductor devices can
become a major cause of temporal or soft errors.
12/25/2012 Prajapati 3
4. Introduction
Motivation
• Soft error susceptibility can become cause of a
system failure in deep submicron ICs.
• Need to study the radiation sensitivity of different
logic implementations and choose the better
radiation tolerant design.
• Need to find radiation sensitive nodes of different
logic styles for reliable system design.
12/25/2012 Prajapati 4
5. Background
Continues Downscaling Technologies
Transistor Density increases, total power
increases
To reduce power consumption VDD are scaled
down
Charge to make a logic value at a circuit node
decreases, noise margin decreases
The systems are now susceptible to noise sources
coming from radiation, cross talk, Power supply
noise etc.
12/25/2012 Prajapati 5
6. Single Event Effects
Single Event Hit on a Semiconductor
Reference: Baumann R. “Soft Errors in Commercial Integration Integrated Circuits”
2004
12/25/2012 Prajapati 6
7. Single Event Effects
SEE is generated due to a highly energetic particle strike to a sensitive node on
a Semiconductor device.
Main Sources:
• High energy Neutrons
• Alpha particles
Higher SEEs as Technology advances due to smaller device sizes and lower
noise margins with decreasing power supply voltage VDD.
Single event phenomena can be classified into three effects
• Single event upset (soft error)
• Single event latchup (soft or hard error)
• Single event burnout (hard failure)
12/25/2012 Prajapati 7
8. Single Event Effects
Modeling of SE hit
(The SE Current Pulse)
Q t / t /
I (t ) (e e )
• Q is the charge deposited by a particle strike,
• τα is the charge collection time constant of the p-n
junction, and
• τβ is the ion-track establishment time constant.
Reference: Baumann
Double Exponential Current Pulse Model is R. “Soft Errors in
Commercial
Commonly used for Simulating Radiation Integration Integrated
Effects Circuits” 2004
12/25/2012 Prajapati 8
9. Single Event Transient
SET Generation in a CMOS Inverter
• SET is a voltage glitch generated due to a particle strike at a
semiconductor device.
• It may be positive or negative depending on the input conditions.
12/25/2012 Prajapati 9
10. Single Event Soft Errors
SET and SEU in a 6-stage Inverter
12/25/2012 Prajapati 10
11. Single Event Soft Errors
SE Induced Soft Delay
12/25/2012 Prajapati 11
12. Single Event Soft Errors
SE Induced Clock Jitter and Race
12/25/2012 Prajapati 12
13. Single Event Soft Errors
SE Induced Crosstalk Noise (SECN)
With the continuous shrinking in device sizes:
• Device density is increasing rapidly in advanced ICs
• Spacing between interconnecting metal wires is also constantly being
reduced
• Coupling capacitance between wires increases.
The noise generated on the neighbor (victim) line from the affecting
(aggressor) line switching is called Crosstalk Noise.
12/25/2012 Prajapati 13
14. Single Event Soft Errors
SE Induced Crosstalk Delay (SECD)
The delay generated in the victim signal
caused by an SET at the aggressor line is
called Single Event Crosstalk Delay (SECD).
12/25/2012 Prajapati 14
15. Different Logic Styles
The high acceptance of low power VLSI integration added a
crucial role to various design levels such as layout, circuit
design, architecture, and fabrication technology.
Designers are now selecting different logic gate
implementations to save the complexity of design and to
minimize power issues.
Various Proposed logic styles are:
• Complementary CMOS
• Pass Transistor Logic
• Transmission Gate
• Gate Diffusion Input
12/25/2012 Prajapati 15
16. Complementary CMOS
A complementary CMOS gate is a combination
of two networks, called the pull-up network
(PUN) and the pull-down network (PDN).
Two networks of opposite type, in which the two
conduction functions are complementary.
Any logic function can be fully realized using
NMOS as well as PMOS connecting between the
Reference: Rabaey, J.M. Digital supply and the gate output.
Integrated Circuits: A Design
Perspective. 1994
12/25/2012 Prajapati 16
17. Complementary CMOS
Advantages
• Robustness against voltage and transistor scaling
• High noise margin
• Sufficient speed
• Ease of Design
Disadvantages
• High input load
• Weak output driving skill
12/25/2012 Prajapati 17
18. Pass Transistor Logic
The basic difference of PTL compared to the
CMOS logic style is that the source side of
the logic transistor networks is connected to
some input signals instead of the power
lines.
It consists of NMOS or PMOS pass transistor
logic with CMOS output inverters
Input inverters that buffer inputs and
generates all signals for the pass transistor
network
12/25/2012 Prajapati 18
19. Pass Transistor Logic
The output buffers for speed
improvement and voltage level
restoration
Many different pass-transistor logic
styles have been proposed such as:
• CPL
• SRPL
• DPL
• LEAP
• DPTL
Reference: Reto Zimmermann and • EEPL
Wolfgang Fichtner “Low-Power Logic • PPL
Styles: CMOS Versus Pass-Transistor
Logic” 1997
12/25/2012 Prajapati 19
20. Pass Transistor Logic
Advantages
• Compact layout
• Reduces number of transistor count
Disadvantages
• Require inverter at the output for level restoration
• Static power consumption
12/25/2012 Prajapati 20
21. Transmission Gate
It builds on the complementary properties
of NMOS and PMOS transistors: NMOS
devices pass a strong 0 but a weak 1, while
PMOS transistors pass a strong 1 but a weak
0.
NMOS and PMOS logic networks are used in
parallel and applied the complementary signals at
the gate of each transistor.
The control signal to the transmission gate is
complementary
12/25/2012 Prajapati 21
22. Transmission Gate
The transmission gate acts as a
bidirectional switch controlled by the
gate signal.
Transmission gates can be used to build
some complex gates very efficiently.
12/25/2012 Prajapati 22
23. Transmission Gate
Advantages
• Simple and efficient operation
• Reduce voltage drop
Disadvantages
• Require more area
• Require complement control signal
• Increase transistor count
12/25/2012 Prajapati 23
24. Gate Diffusion Input
The GDI basic cell seems like a CMOS inverter
built in SOI or dual well CMOS process.
The GDI cell contains three inputs:
• G(common gate input of nMOS and pMOS)
• P (input to the source/drain of pMOS)
• N (input to the source/drain of nMOS)
Bulks of both nMOS and pMOS are connected to
N or P (respectively), so it can be arbitrarily
biased at contrast with a CMOS inverter.
12/25/2012 Prajapati 24
25. Gate Diffusion Input
Advantages
• Reduce number of transistor count
• Reduce supply voltage
• Reduce Area, Power and Delay
Disadvantages
• Limitation with CMOS process
• Increase manufacturing cost
12/25/2012 Prajapati 25
26. Gate Diffusion Input
The all basic GDI logic gate functions cannot implement using the standard
CMOS process.
All GDI logic gate functions only possible in SOI or Twin well CMOS process
12/25/2012 Prajapati 26
28. Simulations and Results
Circuits used for Experiments:
• NAND String
• Random circuit
• ISCAS-85 c17
• M2 module in ISCAS-85 c432
• Full Adder module in ISCAS-85 c6288
All these circuits were constructed using:
• Standard CMOS, VDD =1.2V
Software used:
• Hspice, Custom Waveview
Device Parameters:
• 45nm,65 nm, and 90nm BSIM 4.0 model card, University of California Berkeley
12/25/2012 Prajapati 28
29. Simulations
Radiation Sensitive nodes in TG NAND string
When A=1 and B=1, Sensitive nodes are N1, N2 and OUT1
When A=1 and B=0, Sensitive nodes are N2 and OUT1
12/25/2012 Prajapati 29
30. Simulations
1.An example circuit (NAND String) is first
constructed using a 45nm CMOS logic style
2. The value of A=1 B=1, and A=1 B=0 is considered,
respectively for positive and negative critical charge
calculation.
3. A energetic particle hit is applied at primary NAND
gate sensitive node, O1.
4. The results of critical charges are obtained for a 45nm CMOS logic style.
5. Now, the same circuit is constructed using 65nm and 90nm CMOS logic style and results are
obtained.
6. Steps 1-5 are repeated for logic style PTL and TG.
12/25/2012 Prajapati 30
33. Analysis
Analysis Considerations are:
• positive critical charge is assumed for SET
generation in logic styles.
• Lower critical charge node among the multiple
sensitive nodes of logic style is considered.
12/25/2012 Prajapati 33
34. Critical Charge Comparisons in 90nm, 65 nm, and
45nm CMOS technologies
25
Critical Charge (fc)
20
15
10 CMOS
5 TG
0 PTL
NAND Random c17 c432 M2 c6288 FA
String Circuit
90nm
20
Critical Charge (fc)
15
10
5 CMOS
0 TG
NAND Random c17 c432 M2 c6288 FA PTL
String Circuit
65nm
12/25/2012 Prajapati 34
35. Critical Charge Comparisons in 90nm, 65 nm, and
45nm CMOS technologies
15
Critical Charge (fc)
10
5 CMOS
0 TG
NAND Random c17 c432 M2 c6288 FA PTL
String Circuit
45nm
The value of critical charge (Qcrit) is least in CMOS logic style
and highest in PTL logic style.
12/25/2012 Prajapati 35
36. Collected Charge comparisons for SD 200ps in 90nm, 65nm and
45nm CMOS technologies
50
Charge (fc)
40
30
20 CMOS
10
PTL
0
TG
NAND Random c17 c432 M2 c6288 FA
String Circuit
90nm
40
Charge (fc)
30
20
CMOS
10
PTL
0
TG
NAND Random c17 c432 M2 c6288 FA
String Circuit
65nm
12/25/2012 Prajapati 36
37. Collected Charge comparisons for SD 200ps in 90nm, 65nm and
45nm CMOS technologies
30
Charge (fc) 25
20
15
10 CMOS
5 PTL
0
TG
NAND Random c17 c432 M2 c6288 FA
String Circuit
45nm
TG logic style requires more collected charges to generate
the same value of SD than other logic styles.
12/25/2012 Prajapati 37
38. Conclusion
PTL logic style is least sensitive to soft errors in comparison to
Complementary CMOS and TG.
TG implementation is the most radiation intolerant design
among all other techniques considered.
The logic gate implementations other than static CMOS
technique have more than one sensitive node.
12/25/2012 Prajapati 38
39. Conclusion
As transistor size decreases, combinational circuit becomes
more susceptible to an energetic particle hit.
TG logic style is less susceptible to Soft Delay Error.
12/25/2012 Prajapati 39
40. References
• Rabaey, J.M. Digital Integrated Circuits: A Design
Perspective. 1994
• Baumann R. “Soft Errors in Commercial Integration
Integrated Circuits” 2004
• Reto Zimmermann and Wolfgang Fichtner “Low-Power Logic
Styles: CMOS Versus Pass-Transistor Logic” 1997
• Morgenshtein, A., A. Fish, and I.A. Wagner "Gate-Diffusion
Input (GDI): A Power-Efficient Method for Digital
Combinatorial Circuits." 2002.
• http://radhome.gsfc.nasa.gov/radhome/see.htm
12/25/2012 Prajapati 40
42. Extras
(a) (b)
(c)
(a) Electron hole path created by charged particle (b) Example of three different particle strike
paths and (c) Experimental result of collected charges at different particle strike paths (Karnik,
Hazucha and Patel 2004)
12/25/2012 Prajapati 42
45. A COMPARISON OF
RADIATION TOLERANCE OF
DIFFERENT LOGIC STYLES
12/25/2012 Prajapati 45
Notas do Editor
After the generation of a dense track of free electron-hole pairs in the semiconductor due to a particle strike (a), carriers are rapidly collected by the electric field. Thus, the collected charge compensates the charge stored at the junction. The non-equilibrium charge distribution outside the depletion region produces a temporary funnel-shaped potential distortion along the track, which further increases charge collection by the drift (b). The funnel starts collapsing and charge collection continues due to the diffusion (c) until all excess carriers have been collected, recombined or diffused away from the junction area. Hence, the transient charge collected from a particle hit produces a current pulse at the junction area