VLSI Design Verification and Testing Overview: Partial-Scan & Scan Variations

Design for Testability (DFT) - 2
Definition

Partial-scan architecture

Scan flip-flop selection methods

Cyclic and acyclic structures

Partial-scan by cycle-breaking

Scan variations
Mohammad Tehranipoor

Scan-hold flip-flop (SHFF)
Electrical and Computer Engineering

Summary
University of Connecticut

8 April 2008 1 8 April 2008 2

Partial-Scan Definition Partial-Scan Architecture

PI PO

A subset of flip-flops is scanned.
Combinational

Objectives: circuit
 Minimize area overhead and scan sequence length, yet
achieve required fault coverage
CK1
 Exclude selected flip-flops from scan:

Improve performance FF

Allow limited scan design rule violations FF
 Allow automation: CK2 SCANOUT

In scan flip-flop selection SFF

In test generation
TC
 Shorter scan sequences – reduce application time
SFF

SCANIN

8 April 2008 3 8 April 2008 4

Scan Flip-Flop Selection Methods Cycle Breaking

Testability measure based:

Difficulties in ATPG
 Use of SCOAP: limited success.

S-graph and MFVS problem

Structure based:
 Cycle breaking

Test generation and test statistics
 Balanced structure
Partial vs. full scan

Sometimes requires high scan percentage
Partial-scan flip-flop

ATPG based:
 Use of combinational and sequential TG

8 April 2008 5 8 April 2008 6

1
 Difficulties in Seq. ATPG Benchmark Circuits

Poor initializability. Circuit s1196 s1238 s1488 s1494

PI 14 14 8 8

Poor controllability/observability of state variables. PO 14 14 19 19
FF 18 18 6 6

Gate count, number of flip-flops, and sequential Gates 529 508 653 647
depth do not explain the problem. Structure Cycle-free Cycle-free Cyclic Cyclic
Sequential depth 4 4 -- --

Cycles are mainly responsible for complexity. Total faults 1242 1355 1486 1506
Detected faults 1239 1283 1384 1379

A Sequential ATPG experiment: Potentially detected faults 0 0 2 2
Untestable faults 3 72 26 30
Circuit Number of Number of Sequential ATPG Fault Abandoned faults 0 0 76 97
gates flip-flops depth CPU s coverage Fault coverage (%) 99.8 94.7 93.1 91.6
Fault efficiency (%) 100.0 100.0 94.8 93.4
Cyclic TLC 355 21 14* 1,247 89.01% Max. sequence length 3 3 24 28
Total test vectors 313 308 525 559
Acyclic Chip A 1,112 39 14 269 98.80% Gentest CPU s (Sparc 2) 10 15 19941 19183

* Maximum number of flip-flops on a PI to PO path

8 April 2008 7 8 April 2008 8

Cycle-Free Example Relevant Results

Circuit

Theorem 8.1: A cycle-free circuit is always
initializable. It is also initializable in the presence of
F2
any non-flip-flop fault.
2

Theorem 8.2: Any non-flip-flop fault in a cycle-free
F3 circuit can be detected by at most dseq + 1 vectors.
F1
3
ATPG complexity: To determine that a fault is
Level = 1 F2 untestable in a cyclic circuit, an ATPG program using
2 nine-valued logic may have to analyze 9Nff time-
s - graph frames, where Nff is the number of flip-flops in the
F1 F3 dseq = 3 circuit.
Level = 1 3

All faults are testable. See Example 8.6.

8 April 2008 9 8 April 2008 10

A Partial-Scan Method The MFVS Problem

For a directed graph find a set of vertices with smallest
cardinality such that the deletion of this vertex-set makes the
graph acyclic.

Select a minimal set of flip-flops for scan to

The minimum feedback vertex set (MFVS) problem is NP-
eliminate all cycles. complete; practical solutions use heuristics.

Alternatively, to keep the overhead low only long
A secondary objective of minimizing the depth of acyclic
cycles may be eliminated. graph is useful. This further reduces time to generate
patterns.

In some circuits with a large number of self-loops,
all cycles other than self-loops may be eliminated. 3 3

L=3
1 2 4 5 6 1 2 4 5 6
L=2
L=1

A 6-flip-flop circuit s-graph

cardinality of a set is a measure of the "number of elements of the set".

8 April 2008 11 8 April 2008 12

2
 Test Generation Partial Scan Example

Scan and non-scan flip-flops are controlled from
Circuit: TLC
separate clock PIs:
355 gates

Normal mode – Both clocks active
21 flip-flops

Scan mode – Only scan clock active
Scan Max. cycle Depth* ATPG Fault sim. Fault ATPG Test seq.

Seq. ATPG model: flip-flops length (L) CPU s CPU s cov. vectors length

Scan flip-flops replaced by PI and PO

Seq. ATPG program used for test generation 0 4 14 1,247 61 89.01% 805 805

Scan register test sequence, 001100…, of length nsff + 4 4 2 10 157 11 95.90% 247 1,249
applied in the scan mode

Each ATPG vector is preceded by a scan-in sequence to set 9 1 5 32 4 99.20% 136 1,382
scan flip-flop states
10 1 3 13 4 100.00% 112 1,256

A scan-out sequence is added at the end of each vector
sequence 21 0 0 2 2 100.00% 52 1,190

Test length = (nATPG + 2) nsff + nATPG + 4 clocks * Cyclic paths ignored

Same as full scan except that nsff is significantly smaller

8 April 2008 13 8 April 2008 14

Partial vs. Full Scan: S5378 Flip-flop for Partial Scan

Seq ATPG
Normal scan flip-flop (SFF) with multiplexer of the LSSD flip-flop is
Original Partial-scan Full-scan used.

Scan flip-flops require a separate clock control:
Number of combinational gates 2,781 2,781 2,781
Number of non-scan flip-flops 179 149 0
Either use a separate clock pin
(10 gates each)
Or use an alternative design for a single clock pin
Number of scan flip-flops 0 30 179
(14 gates each) D
Gate overhead 0.0% 2.63% 15.66% Master Slave
MUX Q
Number of faults 4,603 4,603 4,603 latch latch
PI/PO for ATPG 35/49 65/79 214/228 SD
Fault coverage 70.0% 93.7% 99.1% TC
Fault efficiency 70.9% 99.5% 100.0% SFF
CK
CPU time on SUN Ultra II 5,533 s 727 s 5s (Scan flip-flop)
200MHz processor
Number of ATPG vectors 414 1,117 585 TC
Scan sequence length 414 34,691 105,662
CK
Normal mode Scan mode

8 April 2008 15 8 April 2008 16

Scan Variations Scan Set

Integrated and Isolated scan methods

PI Logic PO
 Scan path: NEC 1968 And
Flip-flops
 Serial scan: 1973
 LSSD: IBM 1977 CK

 Scan set: Univac 1977

TC
 RAS: Fujitsu/Amdahl 1980 SCANIN SCANOUT

8 April 2008 17 8 April 2008 18

3
 Scan Set Applications Random-Access Scan (RAS)

The scan function is implemented like a random-

Advantages access memory (RAM)
 Potentially useable in delay testing.
All flip-flops form a RAM in scan mode
 Concurrent testing: can sample the system state
A subset of flip-flops can be included in the RAM if
while the system is running partial scan is desired

Used in microrollback
In scan mode, any flip-flop can be read or written

Disadvantages
 Higher overhead due to routing difficulties

8 April 2008 19 8 April 2008 20

Random-Access Scan (RAS) RAS Flip-Flop (RAM Cell)

PI PO
Combinational D Q
logic From comb. logic To comb.
RAM SD logic
SCANIN Scan flip-flop
(SFF)
nff
CK
bits CK
TC
SCANIN SCANOUT TC
SEL
Address decoder SCANOUT
SEL
ADDRESS Address scan
register
ACK
log2 nff bits

8 April 2008 21 8 April 2008 22

RAS Applications Scan-Hold Flip-Flop (SHFF)

To SD of

Logic test: reduced test length. We don’t have shift D
next SHFF

an entire pattern again as in full scan method. Only Q

SD
the differences between existing data in flip-flops
SFF

Delay test: Easy to generate single-input-change TC
Q
(SIC) delay tests. CK

Advantage:
HOLD

RAS may be suitable for certain architecture, e.g., where
memory is implemented as a RAM block.
The control input HOLD keeps the output steady

Disadvantages: at previous state of flip-flop.
Not suitable for random logic architecture



Applications:

High overhead – gates added to SFF, address decoder,
address register, extra pins and routing
Reduce power dissipation during scan

Isolate asynchronous parts during scan test

Delay testing

8 April 2008 23 8 April 2008 24

4
 Summary

Partial-scan is a generalized scan method; scan

can vary from 0 to 100%.

Elimination of long cycles can improve testability
via sequential ATPG.

Elimination of all cycles and self-loops allows
combinational ATPG.

Partial-scan has lower overheads (area and
delay) and reduced test length.

Partial-scan allows limited violations of scan
design rules, e.g., a flip-flop on a critical path
may not be scanned.

8 April 2008 25