Title: A novel approach to reduce test power consumption
1A novel approach to reduce test power consumption
- Santanu Chattopadhyay
- Shantanu Gupta
- Tarang Vaish
- Department of Computer Science and Engineering,
- Indian Institute of Technology Guwahati
2Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
3Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
4Motivation
- Cost constraints on packaging require a tight
limitation on power dissipation. - Power consumed during test is as much as twice
the normal mode operation 10 - Power consumption is attributed to the high
switching activity during test. - Thus reducing test power engenders lower
production costs.
5Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
6Scan based DFT
- Most common Design for test (DFT) method
- Sequential elements modified to scan cells and
form a serial shift register - In scan mode, cell shifts in value from previous
scan cell - In normal mode, data loaded in parallel from
combinational part
7Test Scan Architecture
PO
PI
SFF
SCANOUT
Combinational logic
SFF
SFF
TC
Not shown CK or MCK/SCK feed all SFFs.
SCANIN
8The Process
- Scan-in of test vectors in the chain composed of
D type flip-flops - Circuit Under Test (CUT) takes the scanned-in
test vector as input - Response from the CUT is given out to the chain,
which is later scanned-out for signature
comparison
9Conventional Scan Cell Testing (CSCT)
- Conventional scan architecture (Q-D)
-
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1011
trans
0
-
-
-
-
10Conventional Scan Cell Testing (CSCT)
- Conventional scan architecture (Q-D)
-
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1011
trans
0
-
-
-
-
1
0
-
-
-
11Conventional Scan Cell Testing (CSCT)
- Conventional scan architecture (Q-D)
-
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1011
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
1
12Conventional Scan Cell Testing (CSCT)
- Conventional scan architecture (Q-D)
-
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1011
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
1
0
1
-
1
1
13Conventional Scan Cell Testing (CSCT)
- Conventional scan architecture (Q-D)
-
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1011
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
1
0
1
-
1
1
1
2
1
1
0
Total 3
14Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
15Previous Works
- ATPG Stage
- Using don't-care bits to improve the correlation
between two consecutive test vectors - Redundancy based approach like fault dropping
(filters out redundant test vectors which target
same faults).
16Previous Works
- Post ATPG Stage
- Scan cell reordering based techniques
- Drawback Long and congested wire routing
- Timing problems and decoder buffer problems
- Reordering of test vectors
- A combination of above two approaches
- Adding extra DFT logic to reduce switching
activity - Drawback Area overheads are high
17Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
18Optimized Scan Cell Testing (OSCT) Our Approach
Scan Architecture Modification
Test Vector Adaptation
Test Vector Reordering
19Scan Architecture Modification (SAM)
- Does not involve scan cell reordering.
- We suggest including \Q-D type of connections at
well chosen position. (instead of only Q-D type
connections) - Type of connection is dependent upon the set of
test vectors and responses provided by ATPG.
D
Q
D
Q
\Q
\Q
20SAM How do we do it?
- Formula deciding flip-flop connection at
- ith index of the scan chain
- Cost incurred (in terms of number of transitions)
is compared at every index of the scan chain to
choose between Q-D and \Q-D
21SAM A simple example
- We calculate the VBitTotal (VBT) and RBitTotal
(RBT) for the input and response vectors - Using these values, we can determine transition
Cost at every scan chain index
Vector
Response
VBT011-001 VBT111-000 VBT211-003
RBT011-000 RBT111-002
RBT211-003 VBT010-013 VBT110-014
VBT210-011 RBT010-014
RBT110-012 RBT210-011
22SAM A simple example
Cost011-00 15 Cost111-00 12
Cost211-00 4 Cost010-01 1
Cost110-01 4 Cost210-01 12
\Q D \Q D
Q D
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Thus depending upon the Cost values we choose
among Q-D and \Q-D type of connections
23Test Vector Adaptation (TVA)
- \Q-D connection flip the bits of test pattern
passing through them. - Consequently, the adaptation of the test vectors
is required to have the original vector after the
scan-in. - Rule for adaptation
- Flip a bit if it passes odd number of \Q-D
connections - Keep it the same otherwise
-
24TVA A simple example
- Take 1011, with target architecture
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
1
0
1
Original vector
1
Even \Q-D
Odd \Q-D
Even \Q-D
Even \Q-D
1
1
1
Adapted vector
1
25Optimized Scan Cell Testing (OSCT)
- Modified Scan design (with \Q-D)
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1111 (adapted)
trans
0
-
-
-
-
26Optimized Scan Cell Testing (OSCT)
- Modified Scan design (with \Q-D)
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1111 (adapted)
trans
0
-
-
-
-
1
0
-
-
-
27Optimized Scan Cell Testing (OSCT)
- Modified Scan design (with \Q-D)
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1111 (adapted)
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
0
28Optimized Scan Cell Testing (OSCT)
- Modified Scan design (with \Q-D)
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1111 (adapted)
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
0
1
0
-
1
0
29Optimized Scan Cell Testing (OSCT)
- Modified Scan design (with \Q-D)
Q
D
Q
Q
Q
D
D
D
SIN
SOUT
\Q
\Q
\Q
\Q
Test vector 1111 (adapted)
trans
0
-
-
-
-
1
0
-
-
-
1
0
-
-
0
1
0
-
1
0
1
0
1
1
0
Total 0
30Test Vector Reordering (TVR)
- Concept of Clash Difference in MSB of last
test response and LSB of new test vector
0 1 1 1
Response of last test vector
New vector for shift
No Clash
0 1 1 0
Response of last test vector
New vector for shift
Clash
31Test Vector Reordering (TVR)
- Each Clash propagates through the complete FF
chain. - (trans) (clashes) X ( FF in scan chain)
- We will reorder test vectors so as to minimize
the number of clashes, thereby reducing number of
transitions further.
32TVR How do we do it?
- Classify the test patterns into 4 groups
represented as 00, 01, 10, 11 of the format
LSB of Vector, MSB of Response - List all vectors of group 00
- List one vector from group 01
- List all vectors of group 11
- List all vectors from group 10 and 01
alternately - If either of the list exhausts, ignore the
alternate picking policy and list the remaining
vectors of the other group.
33TVR A simple example
(0101, 1001 ) (1100, 1000 ) (0110, 0011
) (1101, 0000 ) (0000, 1001 ) (1101, 0101
) (0100, 0001 )
(0110, 0011 ) (0100, 0001 )
(0110, 0011 ) (0100, 0001 )
All from 00
(1100, 1000 ) (0000, 1001 )
(1100, 1000 )
One from 01
(0101, 1001 )
All from 11
(1101, 0000 ) (1101, 0101 )
(1101, 0000 )
Alternately from 10 and 01
3 Clashes before ordering
(0000, 1001 )
(0101, 1001 )
(1101, 0101 )
0 Clashes after ordering
Division into groups
34Advantages
- Significant test power reduction
- Circumvents scan cell reordering based approaches
in terms of feasibility - No interference with circuit parameters like
delay or area - No compromise on the fault coverage
- No extra DFT logic
35Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
36Simulation Resultsusing ISCAS89 benchmarks
37Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
38Conclusions and future work
- We were effectively able to reduce number of
transitions thereby reducing the power consumed. - 24.4 for s510
- Our method is independent of other parameters
like delay and fault coverage. - Future work would involve molding ATPG patterns
generated to meet our ends.
39Presentation Outline
- Motivation
- Scan based DFT
- Previous Works
- Proposed approach
- Simulation results
- Conclusions and future work
- References
40References
- 1 A. Crouch, Design-for-Test for Digital ICs
and Embedded Core Systems, Number IBSN
0-13-08427-1 Prentice Hall, 1999. - 2 V. Dabholkar and S. Charkravarty, Techniques
for minimizing power in scan and combinational
circuits during test application, IEEE Trans, on
Computer Aided Design, 17(12)1325-1333, 1998. - 3 S. Devadas and S. Malik, A survey of
optimization techniques targeting low power VLSI
circuits, In Proc. Of Design Automation
Conferences, pages 242-247, 2002. - 4 D.Bryan F.Brglez and K.Kozminski,
Combinational profiles of sequential Benchmark
circuits, IEEE ISCAS, 31929-1934, May 1989. - 5 S. Gerstendorfer and H.J.Wunderlich,
Minimized power consumption for scan-based Bist,
In Proc. IEEE International Test Conference Pages
77-84, 1999. - 6 H.K.Lee and D.S. Ha, On the generation of
test patterns for combinational circuits,
Technical Reports 12-93, Dept. of Elec. Eng.
Virginia Polytechnic Institute and State
University. - 7 I. Bayraktarouglu O. Sinanoglu and A.
Orailoglu. Scan power reduction through test data
transition frequency analysis. In Proc. Of
International Test Conference, pages 844-850,
2002. - 8 C. Laundrault P. Girard, L. Guiller and S.
Pravossoudovitch, A test vector ordering
technique for switching activity reduction during
test operation. In IEEE Great Lakes Symposium on
VLSI, pages 24-27, 1999. - 9 C. Laundrault Y. Bonhomme, P. Girard and S.
Pravossoudovitch, Power driven chaining of
flip-flops in scan architectures. In Proc. IEEE
International Test Conference, pages 786-803,
2002. - 10 Y. Zorian. A distributed BIST control scheme
for complex VLSI devices. In Proc.11th IEEE VLSI
Test Symposium, pages 49, 1993.
41Thank You