PPT

Report
Substantial Fault Pairs at-A-Time (SFPAT):
An Automatic Diagnostic Pattern
Generation Method
Jing Ye1,2, Xiaolin Zhang1,2, Yu Hu1, and Xiaowei Li1
1Key
Laboratory of Computer System and Architecture
Institute of Computing Technology
Chinese Academy of Sciences
2Graduate
University of Chinese Academy of Sciences
Motivation
Fault Diagnosis Quality
Efficiency of
Diagnosis Method
Distinguishability
of Used Patterns
Distinguish as Many
Fault pairs as possible
Few More Patterns
Than Test Patterns
2
Outline

Key Observation
• Distinguishability of 1-detect compressed Test Patterns
• Distinguishability of N-detect Test Patterns

Related Work

Proposed Diagnostic Pattern Generation Method
• Diagnostic Pattern Generation Method Overview
• Circuit Transformation and Fault List Creation
• Diagnostic Pattern Generation Flow

Experimental Result
• Experimental Setting
3
Key Observation
Distinguishability of 1-Detect Compressed Test Patterns

Experiment Setting
• ISCAS’89 benchmark circuits
• 1-detect compressed test patterns (TetraMax Ver.A-2007.12)

Fault Pairs Classification
FP1 type
FP2 type
two faults in
the fault pair
are in the
same FFR.
FP3 type
two faults in the fault pair
are in different FFRs but
with the same
observation points.
two faults in the fault pair
are in different FFRs but
with at least one different
observation points.
qp
a
b
c
d
e
f
h
g
qp
qp
m
q
i
j
Fanout Free Region (FFR)
n
p
p
4
Key Observation
Percentage of FP3 type fault pairs among all the fault pairs
Percentage
ofPercentage
indistinguishable
FP3
type
fault
pairs
among
allthe
the
indistinguishable
faultofpairs
Percentage
of FP3
type
fault
pairs
among
all
the
fault
pairs
indistinguishable
FP3FP3
type
fault
pairs
among
all the
indistinguishable
faultfault
pairs
Percentage
of
FP3
type
fault
pairs
among
all
fault
pairs Percentage
Percentage
of indistinguishable
FP3
type
fault
pairs
among
all
the
indistinguishable
fault
pairs
of FP3
type
fault
pairs
among
all the
fault
pairs
Percentage
of indistinguishable
type
fault
pairs
among
all the
indistinguishable
pairs
Distinguishability of 1-Detect Compressed Test Patterns
Percentage of FPi-type
fault pairs among all
the fault pairs
FP3 type
Percentage of FP2 type fault pairs among all the fault pairs
Percentage
ofPercentage
indistinguishable
FP2
type
fault
pairs
among
allthe
the
indistinguishable
faultofpairs
Percentage
of FP2
type
fault
pairs
among
all
the
fault
pairs
indistinguishable
FP2FP2
type
fault
pairs
among
all the
indistinguishable
faultfault
pairs
Percentage
of
FP2
type
fault
pairs
among
all
fault
pairs Percentage
Percentage
of indistinguishable
FP2
type
fault
pairs
among
all
the
indistinguishable
fault
pairs
of FP2
type
fault
pairs
among
all the
fault
pairs
Percentage
of indistinguishable
type
fault
pairs
among
all the
indistinguishable
pairs
Percentage of FP1 type fault pairs among all the fault pairs
Percentage
ofPercentage
indistinguishable
FP1
type
fault
pairs
among
allthe
the
indistinguishable
faultofpairs
Percentage
of FP1
type
fault
pairs
among
all
the
fault
pairs
indistinguishable
FP1FP1
type
fault
pairs
among
all the
indistinguishable
faultfault
pairs
Percentage
of
FP1
type
fault
pairs
among
all
fault
pairs Percentage
Percentage
of indistinguishable
FP1
type
fault
pairs
among
all
the
indistinguishable
fault
pairs
of FP1
type
fault
pairs
among
all the
fault
pairs
Percentage
of indistinguishable
type
fault
pairs
among
all the
indistinguishable
pairs
100%
100%
100%
100%
90%
90%
90%
90%
80%
80%
80%
80%
70%
70%
70%
70%
60%
60%
60%
60%
50%
50%
50%
50%
40%
40%
40%
40%
FP3 type
FP2 type
FP1 type
Percentage of indistinguishable
20%
20%
20%
20% FP -type fault pairs among all
i
FP1 type
10%
10%
10%
10%
the indistinguishable fault pairs
Percentage of FP3 type fault pairs among all the fault pairs
Percentage of indistinguishable FP3 type fault pairs among all the indistinguishable fault pairs
0%
0%
0%
0%
Percentage of FP2 type fault pairs among all the fault pairs
Percentage of indistinguishable FP2 type fault pairs among all the indistinguishable fault pairs
AVERAGE
Percentage of FP1 type fault pairs among all the fault pairs
Percentage of indistinguishable FP1 type fault pairs among all the indistinguishable fault pairs
FP2 type
30%
30%
30%
30%
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
08 98 44 49 82 86 20 44 10 26 41 13 20 32 38 53 96 38 23 88 94 78 34 07 50
s2 s2 s3 s3 s3 s3 s4 s4 s5 s5 s6 s7 s8 s8 s8 s9 s11 s12 s14 s14 s14 s53 s92 132 158
s
s
5
Key Observation
Distinguishability of 1-Detect Compressed Test Patterns


FP1 type fault pairs
• Two faults in the fault pair are in the same FFR
FP2 type fault pairs
• Two faults in the fault pair are in different FFRs
but with the same observation points

50%
Percentage of FP1 type fault pairs among all the fault pairs
Percentage
60%
40%
FP1 > FP2 > FP3
30%
‘>’ : harder to be distinguished
f
Percentage
80%
but with at least one different observation point
20%
10%
h
g
Percentage of FP2 type fault pairs among all the fault pairs
90%
70%
e
Percentage
100%
FP3 type fault pairs
• Two faults in the fault pair are in different FFRs
a
b
c
d
Percentage of FP3 type fault pairs among all the fault pairs
q
m
i
j
n
0%
AVERAGE
p
6
Key Observation
Distinguishability of 1-Detect Compressed Test Patterns


FP1 type fault pairs
• Two faults in the fault pair are in the same FFR
FP2 type fault pairs
• Two faults in the fault pair are in different FFRs
but with the same observation points

50%
Percentage of FP1 type fault pairs among all the fault pairs
Percentage
60%
40%
FP1 > FP2 > FP3
30%
‘>’ : harder to be distinguished
f
Percentage
80%
but with at least one different observation point
20%
10%
h
g
Percentage of FP2 type fault pairs among all the fault pairs
90%
70%
e
Percentage
100%
FP3 type fault pairs
• Two faults in the fault pair are in different FFRs
a
b
c
d
Percentage of FP3 type fault pairs among all the fault pairs
q
m
i
j
n
0%
AVERAGE
p
7
Key Observation
Distinguishability of N-Detect Test Patterns

N-detect test pattern
• A fault may be detected for multiple times in different ways.
pat
FP1 type fault pairs
Flt.1
Flt.2
55
Num. of indistinguished fault pairs
50
s208
s1196
s1423
s1488
s1494
s5378
45
40
35
30
25
20
15
10
5
0
1-detect
compressed
1-detect
2-detect
3-detect
4-detect
8
Related Work

Test elimination process of modifying test patterns
• [I. Pomeranz, S. M. Reddy TCAD2000]
• [I. Pomeranz, S. M. Reddy ETS2007]

Exclusive test pattern generation
• [V. D. Agrawal, D. H. Baik, et al. ICVD2003]

Pattern generation for fault-tuple modeled faults
• [N. K. Bhatti, R. D. Blanton ITC2006]

Integer linear program formulation
• [M. A. Shukoor, V. D. Agrawal ETS2009]

Pattern distinguishability and N-detect patterns
• [Z. Wang, M. Marek-Sadowska, et al. ICCD2003]

Pattern reordering algorithm for truncated fail data
• [C. Gang, S. M. Reddy, et al. DAC2006]
9
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Method Overview
Consider Substantial Fault Pairs at-A-Time
+ =>
Original circuit netlist
pat.A
pat.B
pat.C
Target fault list
Transferred circuit netlist and fault list
=>
Flt.1 flt.2 flt.3
flt.4 flt.5
ATPG tool
No!
=>
END
More indistinguished fault pairs?
Yes!
10
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Method Overview
Consider Substantial Fault Pairs at-A-Time
+ =>
Original circuit netlist
pat.A
pat.B
pat.C
Target fault list
Transferred circuit netlist and fault list
=>
Flt.1 flt.2 flt.3
flt.4 flt.5
ATPG tool
No!
=>
END
More indistinguished fault pairs?
Yes!
11
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Method Overview
Consider Substantial Fault Pairs at-A-Time
+ =>
Original circuit netlist
pat.A
pat.B
pat.C
Target fault list
Transferred circuit netlist and fault list
=>
Flt.1 flt.2 flt.3
flt.4 flt.5
ATPG tool
No!
=>
END
More indistinguished fault pairs?
Yes!
12
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Method Overview
Consider Substantial Fault Pairs at-A-Time
+ =>
Original circuit netlist
Target fault list
Transferred circuit netlist and fault list
=>
END
No!
Cont.
=>
Reduce noise
Lower power ATPG tool
Compress patterns
More indistinguished fault pairs?
Yes!
13
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation


Miter circuit
• Miter circuit is a circuit consisting of two modified duplication
D1 and D2 of the original circuit.
• Different connection of D1 and D2 is proposed in previous works.
S-fault
• The pattern which can detect a S-fault in the transformed circuit
can distinguish its related fault pair in the original circuit.

Example
• Stuck-at v fault at l: l/v.
• We will work on other fault models in the future.
• Distinguish the fault pair (a/1,c/1) and the fault pair (b/1,d/1).
a
b
c
d
e
f
h
g i
j
m
n
14
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)
SA1-module
• ‘out’ = ‘sel’ | ‘in’
SA1
•
in
S0
1S
1
MUX

out
a
b
c
d
a
b
c
d
e
f
h
g i
j
m
h
g i
j
m
n
D1
e
f
n
D2
sel
15
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)
SA1-module
• ‘out’ = ‘sel’ | ‘in’
SA1
•
in
S0
1S
1
MUX

out
A
B
C
D
a
b
c
d
a
b
c
d
e
f
h
g i
j
m
h
g i
j
m
n
D1
e
f
n
D2
sel
16
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)
SA1-module
• ‘out’ = ‘sel’ | ‘in’
SA1
•
in
S0
1S
1
MUX

out
A
B
C
D
a
b
c
d
a
b
c
d
SA1
e
SA1
f
h
g i
j
m
h
g i
j
m
n
D1
e
f
SA1
SA1
M
N
n
D2
sel
17
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)

SA1-module
A
B
C
D
a
b
c
d
a
b
c
d
• ‘out’ = ‘sel’ | ‘in’
•
in
S0
1S
1
MUX
SA1
out
sel

S-fault
sel1
0
SA1
e
SA1
f
h
g i
j
m
h
g i
j
m
n
D1
e
f
SA1
M
N
n
D2
SA1
sel2
1
• sel1/1 – (a/1,c/1)
• sel2/1 – (b/1,d/1)
18
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)

SA1-module
A
B
C
D
a
b
c
d
a
b
c
d
• ‘out’ = ‘sel’ | ‘in’
•
in
S0
1S
1
MUX
SA1
out
sel

S-fault
sel1
0
SA1
SA1
h
m
i
FAULT-FREE
g
f
j
n
D1
e
h
m
i
FAULT-FREE
g
f
j
n
D2
M
N
e
SA1
SA1
0
0
sel2
1
• sel1/1 – (a/1,c/1)
• sel2/1 – (b/1,d/1)
19
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)

SA1-module
A
B
C
D
a
b
c
d
a
b
c
d
• ‘out’ = ‘sel’ | ‘in’
•
in
S0
1S
1
MUX
SA1
out
sel

S-fault
sel1
0
SA1
h
i
INJECTg a/1
f
j
D1
m
h
i
INJECTg c/1
f
j
D2
m
e
SA1
e
SA1
SA1
n
1
M
N
n
sel2
1
• sel1/1 – (a/1,c/1)
• sel2/1 – (b/1,d/1)
20
Proposed Diagnostic Pattern Generation Method
Circuit Transformation and Fault List Creation

Target fault pair
• (a/1,c/1)
• (b/1,d/1)

SA1-module
A
B
C
D
a
b
c
d
a
b
c
d
• ‘out’ = ‘sel’ | ‘in’
•
in
S0
1S
1
MUX
SA1
out
sel1
h
i
INJECTg h/1
f
j
D1
SA1
e
SA1
m
n
h
m
i
FAULT-FREE
g
f
j
n
D2
M
N
e
SA1
SA1
0
sel2
sel

S-fault
• sel1/1 – (a/1,c/1)
• sel2/1 – (b/1,d/1)

0
Fault in original circuit
• Constrain the value of sel to 0
21
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Flow
+ =>
Original circuit netlist
FP1 type fault pairs
Transferred circuit netlist and fault list
=>
FP1 > FP2 > FP3
ATPG tool
=>
+ =>
Original circuit netlist
Indistinguished
Target
fault list fault pairs
Some of faults are
certain distinguished
if they are detected.
(1) All the faults in
original circuits.
(2) S-faults of FP1 type
fault pairs which may
not be distinguished
when they are detected.
Transferred circuit
circuit netlist
netlist and
and S-fault
fault list
Transferred
list
=>
ATPG tool
No!
=>
END
More indistinguished fault pairs?
Yes!
22
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Flow
+ =>
Original circuit netlist
FP1 type fault pairs
Transferred circuit netlist and fault list
=>
=>
hATPG tool
m
i
g
f Indistinguished
j
fault
n pairs
e
+ =>
a
b
c
Original circuit netlist
d
Some of faults are
certain distinguished
if they are detected.
(1) All the faults in
original circuits.
(2) S-faults of FP1 type
fault pairs which may
not be distinguished
when they are detected.
=>
Transferred circuit netlist and S-fault list
Is a fault pair distinguished if both faults are detected?
=>
(a/1,b/1) (a/1,e/0) (a/1,c/1) (a/1,d/1)
ATPG tool
Yes
Yes
No
No
(b/1,e/0) (b/1,c/1) (b/1,d/1) (b/1,f/0)
No!
END
More indistinguished
fault pairs?Yes
Yes
No
No
Yes! (e/0,g/1) (e/0,g/0)
(e/0,d/1)
(e/0,f/0)
Yes
Yes
Yes
Yes
(c/1,g/0)
(d/1,f/0)
(d/1,g/1) (d/1,g/0)
Yes
Yes
Yes
Yes
(a/1,f/0)
Yes
(b/1,g/1)
Yes
(c/1,d/1)
Yes
(f/0, g/1)
Yes
(a/1,g/1)
Yes
(b/1,g/0)
Yes
(c/1,f/0)
Yes
(f/0,g/0)
Yes
(a/1 g/0)
Yes
(e/0,c/1)
Yes
(c/1,g/1)
Yes
(g/1,g/0)
Yes
23
Proposed Diagnostic Pattern Generation Method
Diagnostic Pattern Generation Flow
+ =>
Original circuit netlist
FP1 type fault pairs
Transferred circuit netlist and fault list
=>
ATPG tool
=>
Indistinguished fault pairs
+ =>
Original circuit netlist
Some of faults are
certain distinguished
if they are detected.
(1) All the faults in
original circuits.
(2) S-faults of FP1 type
fault pairs which may
not be distinguished
when they are detected.
Transferred circuit netlist and S-fault list
=>
ATPG tool
No!
=>
END
More indistinguished fault pairs?
Yes!
BREAK
(1) No patterns can distinguish
a target fault pair.
(2) ATPG tool cannot achieve 100%
S-fault coverage.
24
Experimental Result
Experimental Setting

Benchmark circuit
• ISCAS’89
• ITC’99

Test Pattern
• TetraMax Ver.A-2007.12
• 1-detect compressed test patterns
25
Experimental Result
Experimental Data
Circuit
s5378 s9234 s13207 s15850 s35932 s38417 s38584
Stuck-at faults
4563
6473
9664 11336 35110 31015 34797
Test patterns
123
156
264
128
Indistinguished fault pairs
593
1621
2276
2971 14438 3850
FP1 type
S-faults
24
111
144
3242
18006 24370 42099 26584 18924 34057 67871
The number of S-faults is mainly
146
28
118 structure
155
determined
by the circuit
Diagnostic patterns
124
164
273
S-faults
578
1552
2062
Diagnostic patterns
20
71
5
26
Total diagnostic patterns
144
235
278
172
137
168
Fault pairs which cannot be
distinguished by any patterns
523
1229
2043
2801 12893 3372
2696
Remaining
indistinguished
2909 13123 3673
2808
The number of S-faults
3
19
13
becomes
much smaller
31
26
Experimental Result
Comparison with Previous Work

Comparison with [12]
Number of diagnostic patterns in [12]
Percentage of distinguished fault pairs
• ISCAS’89: almost
the same
for
the
small circuits
under diagnostic patterns among
Number of diagnostic
patterns
in this
work
indistinguishable fault pairs under test patterns
Number of test patterns in [12]
• ITC’99: different
version
of
benchmark
circuits
[12]
this work
Number of test patterns in this work
Percentage
Number
350
100%
300
90%
250
80%
200
70%
150
60%
100
50%
50
40%
0
30%
s5378
s9234
s13207
s15850
s35932
[12] I. Pomeranz and S. M. Reddy, "Diagnostic Test Generation Based on Subsets of Faults,"
Proc. of European Test Symposium (ETS), pp. 151-158, 2007.
27
Experimental Result
Comparison with Previous Work

Comparison with [12]
Percentage of distinguished fault pairs
Number of diagnostic patterns in [12]
under diagnostic patterns among
Number of diagnostic patterns in this work
indistinguishable fault pairs under test patterns
Number of test patterns in [12]
[12]
this work
Number of test patterns in this work
About 90% of distinguished
100% in this work
Number
fault pairs under diagnostic
patterns
among indistinguished
350 fault pairs
under test patterns in [12]
Percentage
100%
300
90%
Number of diagnostic
patterns in this work
250
Number of diagnostic
200
patterns in [12]
80%
70%
150
Number of test
patterns in this work
100
60%
50%
Number of test
50 [12]
patterns in
40%
0
30%
s5378
s9234
s13207
s15850
s35932
[12] I. Pomeranz and S. M. Reddy, "Diagnostic Test Generation Based on Subsets of Faults,"
Proc. of European Test Symposium (ETS), pp. 151-158, 2007.
28
Conclusion
Distinguishability of patterns are important !

Distinguishability of 1-detect compressed test patterns
• FP1 > FP2 > FP3

Miter-circuit and S-fault
• The pattern which can detect a S-fault in the miter-circuit can
distinguish its related fault pair in the original circuit.
• There is no need to modify the ATPG tool, and the functions of
ATPG tool can also be applied.
29

Thank You for Your Attention !

Any Questions?

similar documents