Memory Testing

 Introduction
 Memory Fault Model
 Memory Test Algorithms
Classical algorithms
 March algorithms
 Memory Fault Simulation
 Memory Test Generation
 Memory BIST

1 VLSI Test 16.2 © National Taiwan University

 Memory Test Algorithms
 A Test algorithm is a finite sequence of test elements
 A test element contains a number of
1. Memory operations
 Read or Write
2. Data pattern (aka. data background)
 Zero or One
3. Address sequence
 Ascending or Descending
 Test (time) Complexity of test algorithm is expressed in terms of N
 N = memory size = number of memory cells
 Higher complexity means longer test time

Mem. Test Alg. is Different from ATPG Alg.

2 VLSI Test 16.2 © National Taiwan University
 Memory Test (Time) Complexity
Size N 10N N lg N N1.5 N2
1M 0.01s 0.1s 0.2s 11s 3h
16M 0.16s 1.6s 3.9s 11m 33d
64M 0.66s 6.6s 17s 1.5h 1.43y
256M 2.62s 26s 1.23m 12h 23y
1G 10.5s 1.8m 5.3m 4d 366y
4G 42s 7m 22.4m 32d 59c
16G 2.8m 28m 1.6h 261d 936c
N = number of memory cells; 100MHz test speed

Linear Complexity Feasible for Production Tests

3 VLSI Test 16.2 © National Taiwan University

 QUIZ
Q: Which one is correct about memory testing?

A) Higher complexity means longer CPU time

B) Test algorithm means a finite sequence of test elements,

which contain: memory operation, data pattern, and address
sequence

C) (N log N) Complexity is acceptable for large memory

4 VLSI Test 16.2 © National Taiwan University

 Memory Testing
 Introduction
 Memory Fault Model
 Memory Test Algorithms
Classical algorithms
 MSCAN
 Checkerboard
 GALPAT
 Butterfly
 March algorithms
 Memory Fault Simulation
 Memory Test Generation
 Memory BIST

5 VLSI Test 16.2 © National Taiwan University

 MSCAN
 aka. zero-one algorithm MSCAN
 Detects all SAF 1.Write zero to every cell
 Detects </0> TF, not </1>
2.Read zero from every cell
3.Write one to every cell
 Does NOT detect all AF, CF 4.Read one from every cell
 Complexity is 4N
 4 operations each cell

addr content addr content

write 0 read 0 write 1 read 1
to all A3 000 from all to all A3 111 from all

A2 000 A2 111
A1 000 A1 111

6 VLSI Test 16.2 © National Taiwan University

 Checkerboard
 Detects all SAF and half TF CHECKERBOARD
 Does NOT detect all AF, CF 1. Write chbd pattern to all cells
 Complexity is 4N 2. Read ckbd pattern from all cells
3. Write ckbd’ pattern to all cells
4. Read ckbd’ pattern from all cells
Same as MSCAN but
Detects Bridging Fault ckbd means 01 alternating
ckbd’ is complement of ckbd

addr content addr content

write read write read
ckbd A3 101 ckbd ckbd’ A3 010 ckbd’

A2 010 A2 101
A1 101 A1 010

1 0 1 0 1 0
0 1 0 1 0 1
cell
1 0 1 0 1 0
7 VLSI Test 16.2 © National Taiwan University
 QUIZ
Q1: Can checkerboard detect OR-type AF between A1 and A2?
ANS:

Q2: Can checkerboard detect OR-type AF between A1 and A3?

ANS:

addr content addr content

write read write read
ckbd A3 101 ckbd ckbd’ A3 010 ckbd’

A2 010 A2 101
A1 101 A1 010

8 VLSI Test 16.2 © National Taiwan University

 Galloping Test (GALPAT)
 Aka. Ping-pong test GALPAT
 Detects all SAF, TF 1. Write background 0 to all cells;
 Detects CF, and AF 2. For BaseCell = 0 to N-1
Complement BC;
 Complexity is 4N2 For OtherCell = 0 to N-1, OC != BC;
 Line 2: 2N2 Read BC; Read OC;
 Line 4: 2N2 Complement BC;
3. Write background 1 to all cells;
OC 4. Repeat Step 2;

BC

memory cell array

Too Long! Only for Characterization

9 VLSI Test 16.2 © National Taiwan University
 QUIZ
Suppose BC is aggressor and OC is victim.
Q1: Can we detect CFst <1; 0/1>?
Q2: Can we detect CFid <↑; 0/1> ?
Q3: Can we detect CFin <↑ ; /↨ >?

0 0 0
OC

0 BC 0
010

0 0 0

0 0 0

10 VLSI Test 16.2 © National Taiwan University

 Butterfly Algorithm
 Detects SAF and TF BUTTERFLY // given MAXDIST < 0.5 col/row size
 Does not detect all CF, AF 1. Write background 0;
2. For BaseCell = 0 to N-1
 Complexity is 10 N log N Complement BC; dist = 1;
 5 reads for each dist While dist  MAXDIST
 dist doubles each loop Read cell @ dist north from BC;
Read cell @ dist east from BC;
 Repeated in line 4 Read cell @ dist south from BC;
6 Read cell @ dist west from BC;
Read BC; dist = dist * 2;
1 Complement BC;
BC 3. Write background 1;
9 4 2 7
5, 10 4. Repeat Step 2;
3

8
cell

memory cell array

11 VLSI Test 16.2 © National Taiwan University
 Summary
 Test algorithm is a finite sequence of test elements

which are: Memory operation, Data pattern, Address sequence
 Test complexity means test time (not CPU time)
 Four classical test algorithms
 MSCAN and Checkerboard’s fault coverage NOT good
 GALPAT and Butterfly are too slow
 Need linear-time test algorithms with good fault coverage
 March test algorithms (see 16.3)

SAF AF TF CF Complexity
MSCAN D - - - 4N D: all detected;
- : not all detected
Checkerboard D - - - 4N
GALPAT D D D D 4N2
Butterfly D - D - 10 N log N

12 VLSI Test 16.2 © National Taiwan University

 FFT
 Q: In GALPAT, choosing all cells as base cell is slow. Can we
choose only cells on the diagonal line as BC? What is complexity?
 Q: Why do we choose cells in the same column and row of BC in
Butterfly test?

GALPAT-DIAGONAL
1. Write background 0 to all cells;
2. For BaseCell = 0 to N-1
BC // BC must be on diagonal line
Complement BC;
For OtherCell = 0 to N-1, OC != BC;
Read BC; Read OC;
Complement BC;
3. Write background 1 to all cells;
4. Repeat Step 2;

13 VLSI Test 16.2 © National Taiwan University

 MSCAN [Breuer & Friedman 1976]
 aka. zero-one algorithm MSCAN
 Detects all SAF 1.Write zero to every cell
 Detects </0> TF, not </1>
2.Read zero from every cell
3.Write one to every cell
 Does NOT detect all AF, CF 4.Read one from every cell
 Complexity is 4N

addr content addr content

write 0 read 0 write 1 read 1
to all A3 000 from all to all A3 111 from all

A2 000 A2 111
A1 000 A1 111

SAF AF TF CF Complexity
D: all detected
MSCAN D - 1/2 - 4N - : not all detected

14 VLSI Test 16.2 © National Taiwan University