Presentation is loading. Please wait.

Presentation is loading. Please wait.

Vishwani D. Agrawal Auburn University, Dept. of Elec. & Comp. Engg. Auburn, AL 36849, U.S.A. Nitin Yogi NVIDIA Corporation, Santa Clara, CA 95050 20 th.

Published byEugenia Prudence Dalton Modified over 9 years ago

Similar presentations

Presentation on theme: "Vishwani D. Agrawal Auburn University, Dept. of Elec. & Comp. Engg. Auburn, AL 36849, U.S.A. Nitin Yogi NVIDIA Corporation, Santa Clara, CA 95050 20 th."— Presentation transcript:

1 Vishwani D. Agrawal Auburn University, Dept. of Elec. & Comp. Engg. Auburn, AL 36849, U.S.A. Nitin Yogi NVIDIA Corporation, Santa Clara, CA 95050 20 th April 2010 28 th IEEE VLSI Test Symposium Santa Cruz, CA

2 Introduction Propose an information and noise analysis method for digital signals using Hadamard transform. Analysis identifies information (spectrally structured) and noise (random) contents of the signal in relative measures. The analysis is useful in a variety of applications like test generation, BIST, test compression, etc. We illustrate its application to test generation. 4/20/20102VTS'10, Santa Cruz, CA

3 June 12, 2009Nitin Yogi - Doctoral Defense3 Test Vectors and Bit-streams Circuit Under Test (CUT) Input 1 Input 2 Input 3 Input 4 Input 5 Input J Vector 1 → Vector 2 → Vector 3 → Vector 4 → Vector 5 → Vector 6 → Vector 7 → Vector 8 → Outputs Time in clocks A digital binary bit- stream signal vector 11010..1 00101..1 10011..0 11010..1 00110..0 01101..0 10111..1 10001..0

4 Hadamard Transform 11111111 11 1 1 11 11 1 11 1 1111 1 1 1 1 11 11 1 1 11 H(8) = Hadamard transform transforms a digital signal from time domain to frequency-related domain. Uses Walsh functions, which are a complete orthogonal set of basis functions that can represent any arbitrary bit-stream. Can be used for binary signals by using the representation {0,1} -> {-1,1} Example of Hadamard matrix of order 8 w0w0 w1w1 w2w2 w3w3 w4w4 w5w5 w6w6 w7w7 Walsh functions (order 3) time 4/20/20104VTS'10, Santa Cruz, CA

5 Hadamard transformation where: X Time domain digital signal vector of stream of N bits H(N) Hadamard transform matrix of order N S Hadamard transform (Walsh spectrum) of X 4/20/20105VTS'10, Santa Cruz, CA Forward transformation (time domain to spectral domain): Reverse transformation (spectral domain to time domain):

6 Properties of Hadamard Transform Orthogonality and symmetry Energy conservation where:  H(N) Hadamard transform matrix of order N  X Binary bit-stream vector in time domain  S Hadamard transform in spectral domain (Walsh spectrum) 4/20/20106VTS'10, Santa Cruz, CA

7 Energy Analysis Total energy in a {-1,+1} binary signal of length N clocks in time domain: Total energy in spectrum: 4/20/20107VTS'10, Santa Cruz, CA

8 Analysis of random binary bit-streams Values of spectral components of random binary bit- streams can be approximated as Gaussian distribution  Mean (µ) = 0  Standard deviation (σ) where:  S r (j) j th spectral component of a random binary bit-stream of length N  Square of the mean of S r (j) 4/20/20108VTS'10, Santa Cruz, CA Equal to N (by energy conservation) Equal to 0 (since mean = 0)

9 Spectral coefficients of random bit-stream 500 samples of random binary bit-streams of length 64 were generated Distribution of values of spectral components analyzed Mean = 0.0035 ≈ 0 Standard deviation = 1.000425 ≈ 1 1σ (68.27%) 2σ (95.45%) 3σ (99.73%) 4/20/20109VTS'10, Santa Cruz, CA Spectral components below a magnitude of 2σ or 3σ can be treated as noise components

10 Generating spectral bit-streams 1. Perform Hadamard transform on binary bit-stream. 2. Filter out noise-like spectral components having magnitudes less than a spectral threshold (Energy conservation of the transfom transfers the energy of filtered components to noise). 3. Perform reverse Hadamard transform to obtain time- domain values in the range (-1,+1) for bits in the bit- stream. 4. Time-domain values are normalized to range (0,1) and used as probabilities of logic 1 in new random bit-streams. 4/20/201010VTS'10, Santa Cruz, CA

11 Generating spectral bit-streams 2 6 -2 2 2 2 1 1 1 1 1 11111111 1 1 1 1 11 11 1 11 1 1111 1 1 1 1 11 11 1 1 11 Hadamard Matrix H(3) {-1,+1} binary bit stream (X) Hadamard transform (S) = Example {0,1} binary bit-stream 1 √8 1 1 0 1 1 1 0 1 0 {0,1} converted to {-1,+1} 4/20/201011VTS'10, Santa Cruz, CA

12 Example of generating spectral bit-streams 0 6/√8 0 0 0 0 0 0 0.75 -0.75 0.75 -0.75 0.75 -0.75 0.75 -0.75 11111111 11 1 1 11 11 1 11 1 1111 1 1 1 1 11 11 1 1 11 2 6 -2 2 2 2 Hadamard transform (S) of 8-bit binary bit- stream Spectral threshold = 2 σ = 2 0 6/√8 0 0 0 0 0 0 0.875 0.125 0.875 0.125 0.875 0.125 0.875 0.125 Probabilities for generating bit-streams 4/20/201012VTS'10, Santa Cruz, CA 0’s are filtered spectral coefficients Time-domain Reverse transformation Unfiltered spectral component 1 √8 1 Normalization for probabilities [X(k)+1] 2

13 Generated spectral bit-streams Original {0,1} binary bit- stream 1 1 1 1 1 Unfiltered spectral component 1 1 1 1 0.875 0.125 0.875 0.125 0.875 0.125 0.875 0.125 Probabilities for generating bit-stream 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 Randomly generated bit-streams from probabilities 1 bit difference between original bit- stream & unfiltered spectral component Generated bit-streams exhibit similar correlation with the unfiltered spectral component as the original bit-stream. Few bits changed by noise are shown in red. 4/20/201013VTS'10, Santa Cruz, CA

14 Application of analysis Application of spectral information analysis: Test generation, BIST, Test data compression, etc. Illustration of effectiveness of analysis using test generation Test vectors generated for RTL faults (PIOs & flip- flops) Generate spectral vectors & fault grade on circuit Compare with random, weighted random & randomly perturbed vectors Analysis applied to ISCAS’89 benchmark circuits s1488, s5378 & s38417 4/20/201014VTS'10, Santa Cruz, CA

15 Spectral coefficients & power analysis for s1488 InputsSpectral CoefficientAmplitudePowerNoise Power Input 1w0w0 0.750.560.44 Input 2 w1w1 0.50.25 0.61 w 13 -0.380.14 Input 3 w1w1 0.560.32 0.49 w 19 0.440.19 Input 4 w0w0 0.630.39 0.47 w 22 -0.380.14 Input 5 w1w1 -0.50.25 0 w5w5 0.50.25 w 19 0.50.25 w 23 0.50.25 Input 6 w4w4 0.50.25 0.5 w 22 0.50.25 Input 7 w1w1 0.50.25 0.47 w5w5 -0.380.14 w 30 0.380.14 Input 8 w4w4 0.380.14 0.58 w 12 0.380.14 w 22 0.380.14 4/20/201015VTS'10, Santa Cruz, CA 32 vectors were generated to detect RTL faults (PIOs & FFs) & analyzed using H(32)

16 Gate level coverage for s1488 4/20/201016VTS'10, Santa Cruz, CA

17 Gate level coverage for s5378 4/20/201017VTS'10, Santa Cruz, CA

18 Gate level coverage for for s38417 4/20/201018VTS'10, Santa Cruz, CA

19 Conclusion Proposed an information analysis framework to distinguish noise from signal content Illustrated effectiveness of method for the application of test generation The method can easily be extended to other applications like BIST and test compression. See, N. Yogi and V. D. Agrawal, “BIST/Test-Compressor Design using Combinational Test Spectrum,” Proc. 13 th IEEE VLSI Design & Test Symp. (VDAT), July 2009, pp. 443-454. N. Yogi and V. D. Agrawal, “Sequential Circuit BIST Synthesis using Spectrum and Noise from ATPG Patterns” Proc. 17 th IEEE Asian Test Symp. (ATS), Nov 2008, pp. 69-74. There is potential for further applications. 4/20/201019VTS'10, Santa Cruz, CA

20 Thank you 4/20/2010VTS'10, Santa Cruz, CA20 Questions please?

21 Test generation comparison with commercial tool 4/20/2010VTS'10, Santa Cruz, CA21 Circuit name RTL-ATPG spectral testsFlexTest gate-level ATPG Coverage (%) No. of vectors CPU* (secs) Coverage (%) No. of vectors CPU* (secs) s148895.6551210398.42470131 s537876.492432208876.798354439 N. Yogi and V. D. Agrawal, “Spectral RTL Test Generation for Gate-Level Stuck-at Faults,” Proc. 15 th IEEE Asian Test Symp. (ATS), Nov 2006, pp. 83-88. * Sun Ultra 5, 256MB RAM Circuit name FlexTest gate-level ATPGBIST gate-level fault coverage (%) Coverage (%) No. of vectors 64k random vectors 64k weighted random vectors Spectral BIST (64k vectors) s1488 97.3173692.1397.11 s5378 77.0673974.3976.8478.28 s38417 49.625511013.4215.8754.59 N. Yogi and V. D. Agrawal, “Sequential Circuit BIST Synthesis using Spectrum and Noise from ATPG Patterns,” Proc. 17 th IEEE Asian Test Symp. (ATS), Nov 2008, pp. 69-74. Test generation: BIST:

Download ppt "Vishwani D. Agrawal Auburn University, Dept. of Elec. & Comp. Engg. Auburn, AL 36849, U.S.A. Nitin Yogi NVIDIA Corporation, Santa Clara, CA 95050 20 th."

Similar presentations

Retiming Scan Circuit To Eliminate Timing Penalty

Retiming Scan Circuit To Eliminate Timing Penalty
Weighted Random and Transition Density Patterns for Scan-BIST Farhana Rashid* Vishwani D. Agrawal Auburn University ECE Department, Auburn, Alabama 36849.

Weighted Random and Transition Density Patterns for Scan-BIST Farhana Rashid* Vishwani D. Agrawal Auburn University ECE Department, Auburn, Alabama
An Algorithm for Diagnostic Fault Simulation Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama 36849 USA 13/29/2010IEEE LATW 10.

An Algorithm for Diagnostic Fault Simulation Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama USA 13/29/2010IEEE LATW 10.
Committee: Vishwani D. Agrawal Adit Singh Victor P. Nelson

Committee: Vishwani D. Agrawal Adit Singh Victor P. Nelson
10/28/2009VLSI Design & Test Seminar1 Diagnostic Tests and Full- Response Fault Dictionary Vishwani D. Agrawal ECE Dept., Auburn University Auburn, AL.

10/28/2009VLSI Design & Test Seminar1 Diagnostic Tests and Full- Response Fault Dictionary Vishwani D. Agrawal ECE Dept., Auburn University Auburn, AL.
Nov. 21, 2006ATS'06 1 Spectral RTL Test Generation for Gate-Level Stuck-at Faults Nitin Yogi and Vishwani D. Agrawal Auburn University, Department of ECE,

Nov. 21, 2006ATS'06 1 Spectral RTL Test Generation for Gate-Level Stuck-at Faults Nitin Yogi and Vishwani D. Agrawal Auburn University, Department of ECE,
3/30/05Agrawal: Implication Graphs1 Implication Graphs and Logic Testing Vishwani D. Agrawal James J. Danaher Professor Dept. of ECE, Auburn University.

3/30/05Agrawal: Implication Graphs1 Implication Graphs and Logic Testing Vishwani D. Agrawal James J. Danaher Professor Dept. of ECE, Auburn University.
Aiman El-Maleh, Ali Alsuwaiyan King Fahd University of Petroleum & Minerals, Dept. of Computer Eng., Saudi Arabia Aiman El-Maleh, Ali Alsuwaiyan King Fahd.

Aiman El-Maleh, Ali Alsuwaiyan King Fahd University of Petroleum & Minerals, Dept. of Computer Eng., Saudi Arabia Aiman El-Maleh, Ali Alsuwaiyan King Fahd.
Nitin Yogi and Vishwani D. Agrawal Auburn University Auburn, AL 36849

Nitin Yogi and Vishwani D. Agrawal Auburn University Auburn, AL 36849
Dynamic Scan Clock Control In BIST Circuits Priyadharshini Shanmugasundaram Vishwani D. Agrawal

Dynamic Scan Clock Control In BIST Circuits Priyadharshini Shanmugasundaram Vishwani D. Agrawal
Diagnostic Test Generation and Fault Simulation Algorithms for Transition Faults Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama 36849.

Diagnostic Test Generation and Fault Simulation Algorithms for Transition Faults Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama
Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama 36849 5/13/2010 NATW 10 1 A Diagnostic Test Generation System.

Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama /13/2010 NATW 10 1 A Diagnostic Test Generation System.
A Diagnostic Test Generation System Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama 36849 USA Nov. 3rdITC 20101.

A Diagnostic Test Generation System Yu Zhang Vishwani D. Agrawal Auburn University, Auburn, Alabama USA Nov. 3rdITC
May 11, 2006High-Level Spectral ATPG1 High-Level Test Generation for Gate-level Fault Coverage Nitin Yogi and Vishwani D. Agrawal Auburn University Department.

May 11, 2006High-Level Spectral ATPG1 High-Level Test Generation for Gate-level Fault Coverage Nitin Yogi and Vishwani D. Agrawal Auburn University Department.
May 17, 2007North Atlantic Test Workshop (NATW) 2007, May 16-18, Boxborough, Massachusetts 1 Nitin Yogi and Vishwani D. Agrawal Auburn University Department.

May 17, 2007North Atlantic Test Workshop (NATW) 2007, May 16-18, Boxborough, Massachusetts 1 Nitin Yogi and Vishwani D. Agrawal Auburn University Department.
Externally Tested Scan Circuit with Built-In Activity Monitor and Adaptive Test Clock Priyadharshini Shanmugasundaram Vishwani D. Agrawal.

Externally Tested Scan Circuit with Built-In Activity Monitor and Adaptive Test Clock Priyadharshini Shanmugasundaram Vishwani D. Agrawal.
Parallel Pattern Single Fault Propagation for Combinational Circuits VLSI Testing (ELEC 7250) Submitted by Blessil George, Jyothi Chimakurthy and Malinky.

Parallel Pattern Single Fault Propagation for Combinational Circuits VLSI Testing (ELEC 7250) Submitted by Blessil George, Jyothi Chimakurthy and Malinky.
11/17/05ELEC 5970-001/6970-001 Lecture 201 ELEC 5970-001/6970-001(Fall 2005) Special Topics in Electrical Engineering Low-Power Design of Electronic Circuits.

11/17/05ELEC / Lecture 201 ELEC / (Fall 2005) Special Topics in Electrical Engineering Low-Power Design of Electronic Circuits.
5/1/2006VTS'061 Upper Bounding Fault Coverage by Structural Analysis and Signal Monitoring Vishwani D. Agrawal Auburn University, Dept. of ECE, Auburn,

5/1/2006VTS'061 Upper Bounding Fault Coverage by Structural Analysis and Signal Monitoring Vishwani D. Agrawal Auburn University, Dept. of ECE, Auburn,
May. 04, 2007General Oral Examination1 Gate-Level Test Generation Using Spectral Methods at Register-Transfer Level Committee Members: Prof. Victor P.

May. 04, 2007General Oral Examination1 Gate-Level Test Generation Using Spectral Methods at Register-Transfer Level Committee Members: Prof. Victor P.

Similar presentations

Ads by Google