Combinational Test Pattern Generation
Combinational Test Generation
Test Generation (TG) Methods
(1) From truth table (2) Using Boolean equation (3) Using Boolean difference (4) From circuit structure
TG from Circuit Structure
Common Concepts Algorithms : D-Algorithm (Roth 1967 [2]), 9-V Algorithm (Cha 1978 [3]), PODEM (Goel 1981 [4]), FAN (Fujiwara 1983 [5])
Automatic Test Pattern Generation (ATPG)
2
A Test Pattern
A test pattern
0 0 1 1 stuck-at 1 0/1 0/1 1
A test pattern with don't cares
1 x x x stuck-at 0 1/0 x 1/0 (Good value and faulty value are different at a PO.) x
Test generation: generates a test for a target fault.
3
Test Generation Methods
(From Truth Table)
Ex: How to generate tests for the stuck-at 0 fault (fault )?
a b f c
abc 000 001 010 011 100 101 110 111
f 0 0 0 0 0 1 1 1
f 0 0 0 0 0 1 0 1
Impractical !!
Test Generation Methods
(Using Boolean Equation)
Since f = ab+ac, f = ac => T = the set of all tests for fault = ON_set(f) OFF_set(f) + OFF_set(f) ON_set(f) = {(a,b,c) | (ab+ac)(ac)' + (ab+ac)'(ac) = 1} = {(a,b,c) | abc'=1} = { (110) }.
High complexity !! Since it needs to compute the faulty function for each fault.
* ON_set(f): All input combinations that make f have value 1. OFF_set(f): All input combinations that make f have value 0.
5
Boolean Difference
Physical Meaning of Boolean Difference
For a logic function F(X)=F(x1, ..., xi, ..., xn), find all the input combinations that make the change of value in xi also cause the change of value in F.
Logic Operation of Boolean Difference
The Boolean difference of F(X) w.r.t. input xi is
dF(X) = F i(0)Fi(1) = Fi(0)Fi(1)+Fi(0)Fi(1), dxi where Fi(0) = F(x1, ..., 0, ..., xn) and Fi(1) = F(x1, ..., 1, ...,
Relationship between TG and Boolean Difference
x1 0 xi xn 1 circuit or 0 1 F 1 0 1 x1 xi xn 0 circuit o r 0 1 F 0 1
xn).
Applying Boolean Difference to Test Generation (1/2)
Case 1: Faults are present at PIs.
a b f c
f = ab+ac => df = f a(0)f a(1) = 1(b+c)+0 = b+c da The set of all tests for line a s-a-1 is {(a,b,c) | a' (b+c)=1} = {(01x), (0x1)}. The set of all tests for line a s-a-0 is {(a,b,c) | a (b+c)=1} = {(11x), (1x1)}.
Applying Boolean Difference to Test Generation (2/2)
Case 2: Faults are present at internal lines.
a b h f c
f = h+ac, h = ab =>df =f h (0)f h (1)=ac 1+ac 1=a+c dh The set of all tests for line h s-a-1 is { (a,b,c)|h' (a'+c')=1 } = { (a,b,c)|(a'+b') (a'+c')=1 } = { (0xx), (x00) }. The set of all tests for line h s-a-0 is {(a,b,c)|h (a'+c')=1} = {(110)}.
8
Test Generation Methods (From Circuit Structure)
1 a 1 b
1/0
fault activation
f
0 c 0 fault propagation
Two basic goals: Fault activation (FA) => Line justification (LJ) Fault propagation (FP)
where 1/0 means that the good value is 1 and the faulty value is 0 and is denoted as D. Similarly, 0/1 is denoted as D'. D and D' are called fault effects (FE).
9
Common Concepts for Structural TG
The FA problem => a LJ problem. The FP problem => (1) Select a FP path to a PO => decisions. (2) Once the path is selected => a set of LJ problems. The LJ problems => decisions or implications .
ex: a
b c
To justify c=1 => a=1 and b=1. (implication) To justify c=0 => a=0 or b=0. (need make decisions) Incorrect decision => Backtracking => Another decision. Once the fault effect is propagated to a PO and all line values to be justified are justified, the test is generated. Otherwise, the decision process must be continued repeatedly until all possible decisions have been tried. (Since the searching is exhaustive, it guarantees to find a test if one exists.)
10
Ex: Decisions When Fault Propagation
d a b c G2 G1 G6 G3 e G4 f2 G5 f1
{G5,G6} G5 F G6 S
The corresponding decision tree
FA => a=1, b=1, c=1 => G1= D', G3=0; FP => through G5 or G6. Decision: through G5 => G2=1 => d=0, a=0. => inconsistency => backtracking!! Decision: through G6 => G4=1 => e=0. => done!! The resulting test is 111x0. D-frontier: The set of all gates whose output value is currently x but have one or more fault signals on their inputs. Ex: Initially, the D-frontier of this example is {G5, G6}.
11
Ex: Decisions When Line Justification
a b c d k q l m n o e f h p r s
l=1 F m=1 S n=1 q=1 k=1 r=1 o=1
The corresponding decision tree
FA => h=D'; FP => e=1, f=1 (=> o=0); FP => q=1, r=1. To justify q=1 => l=1 or k=1. Decision: l=1 => c=1, d=1 => m=0, n=0 => r=0. => inconsistency => backtracking!! Decision: k=1 => a=1, b=1. To justify r=1 => m=1 or n=1 (=> c=0 or d=0). => done!! J-frontier: The set of all gates whose output value is known but is not implied by its input values. Ex: Initially, the J-frontier of the example is {q=1, r=1}.
12
Implications
Implication: computation of the values that can be uniquely determined.
- Local implication: propagation of values from one line to its immediate successors or predecessors. - Global implication: the propagation involving a larger area of the circuit and reconvergent fanout.
Maximum implication principle: perform as many implications as possible. Maximum implications help us to either reduce the number of problems that need decisions or to reach an inconsistency sooner.
13
Local Implications (Forward)
Before 0 1 1 1 x D' D x x x 0 J-frontier={ ...,a } a D' x D-frontier={ ...,a } D a 0 x 1 1 1 0
After 0 1 0 a
J-frontier={ ...
0 D-frontier={ ... a
14
Local Implications (Backward)
Before x x x 1 x x 1 0 0 a x 1 1 0 1 J-frontier={ ... } 1 x x x After 1 0 0 a 1
J-frontier={ ...,a 1
15
Global Implications
x D x Before d x x e x After x x g D x x x 1 D
(1) Future unique D-drive. * The D-frontier is {d,e}, so we do not have unique D-drive. We can observe, however, that no matter how we will decide to propagate the fault (through d or e or both), eventually we will reach a unique D-drive situation, because the fault must propagate through g. 1 D A D A 1 0 B F B F E E C C (2) F=1 implies B=1. (3) F=0 implies B=0 when (Static learning) A=1. (Dynamic learning) * (2), (3) are based on contraposition law: (A=>B) <=> (!B => !A). [6],[6']
16
D-Algorithm: Example (1/6)
Logic values = {0, 1, D, D', x}.
d' d h 1 i j n k f' f l m
e' e
0 a 1 b 1 c
g
D
17
D-Algorithm: Example (2/6)
1d
d' 0
e' e
0 a 1 b 1 c
h 1 i D' j n k
g D f' f
l m
18
D-Algorithm: Example (3/6)
1d
d' 0
1e 0 a 1 b 1 c
e' 0
h 1 i D' j
1 k 1=D' l 1 m 1
19
n D
g
D
f' f
conflict=>backtracking!
D-Algorithm: Example (4/6)
1d
d' 0
1e 0 a 1 b 1 c
e' 0
h 1 i D' j
1 k D' l
g
D
f' f
20
D-Algorithm: Example (5/6)
1d
d' 0
1e 0 a 1 b 1 c
e' 0
h 1 i D' j
1 k D' l 1 m 1=D'
21
n D
g
D 1f
f' 0
conflict=>backtracking!
D-Algorithm: Example (6/6)
1d
d' 0
1e 0 a 1 b 1 c
e' 0
g
D 1f
f' 0
h 1 i D' j 1 k D' l 1 m
D'
n D
22
D-Algorithm: Value Computation
Decision Implication Comments a=0 h=1 b=1 c=1 g=D d=1 i=D d=0 j=1 k=1 l=1 m=1 n=D e=0 e=1 k=D Propagate via n f=1 m=D f=0 l=1 n=D Contradiction
23
Active the fault Unique D-drive
e=1 k=D e=0 j=1 l=1 m=1 n=D f=0 f=1 m=D
Propagate via k
Propagate via n
Propagate via i
Contradiction Propagate via m
D-Algorithm: Decision Tree
{i,k,m} i {k,m,n} n F n F k {m,n} m S
Two times of backtracking!!
* decision node: the associated D-frontier. branch: the decision taken, i.e., the gate selected from the D-frontier. * The D-algorithm first tried to propagate the fault solely through i, then through both i and k, and eventually succeeded when all three paths were simultaneously sensitized.
24
