titel und thema des vortrages - uni-potsdam.de

86
ASIC Test

Upload: others

Post on 11-May-2022

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Titel und Thema des Vortrages - uni-potsdam.de

ASIC Test

Page 2: Titel und Thema des Vortrages - uni-potsdam.de

Outline

1 Introduction

2 Fault models

3 Test Generation Flow

4 Design-for-Testability

5 Problems of Testing

6 Conclusions

Page 3: Titel und Thema des Vortrages - uni-potsdam.de

Introduction

1

Introduction

Goals and Problems of Testing

Types of Tests

Some Definitions of Testing

Page 4: Titel und Thema des Vortrages - uni-potsdam.de

Verification vs. Testing

Model

Netlist

Chip Implementation/

Synthesis Production

Simulation/Verification Testing

Bug Defect

Page 5: Titel und Thema des Vortrages - uni-potsdam.de

Verification vs. Testing

Model

Netlist

Chip Implementation/

Synthesis Production

Simulation/Verification Testing

Bug Defect

Page 6: Titel und Thema des Vortrages - uni-potsdam.de

Goal & Problems of Testing

• Goal

• The goal of testing is to apply a minimum set of

input vectors to each device to determine if it contains

a defect.

• Problem

• ⇒ Apply a set of test vectors to a device and compare

its outputs to the known good response.

• ⇒ The optimum test set will detect the greatest

number of defects that can be present in a device

with the least number of test vectors

Page 7: Titel und Thema des Vortrages - uni-potsdam.de

Types of Tests

• Exhaustive – apply every possible input vector

• Functional – test every function of the device

• Fault Model Derived – find a test for every modelled fault

Page 8: Titel und Thema des Vortrages - uni-potsdam.de

Which test type is closest to optimum?

ALU

Consider a 16-bit ALU

• 8 operations

4 arithmetical operations (add, sub, mul, div)

4 logical operations (and, or, xor, inv)

• 2 16-bit operands

a b

c

16 16

16

3 op

Page 9: Titel und Thema des Vortrages - uni-potsdam.de

Which test type is closest to optimum?

Exhaustive testing

• Will detect 100 % of detectable faults

• Can done automatically

• Requires 235 = 17.179.869.184 test vectors

• Test at 10 MHz would take 57 min 16 s

Page 10: Titel und Thema des Vortrages - uni-potsdam.de

Which test type is closest to optimum?

Functional testing

• Will detect 100 % of detectable faults

• No algorithmic way to verify that all functional modes have

been tested (designer expertise required)

• Total functional test will take 4(log_ops) × 4(vecs) + 4(arith_ops)× ≈ 20(vecs) ≥ 96 vectors

• Test at 10 MHz would take 9.6 µs

Page 11: Titel und Thema des Vortrages - uni-potsdam.de

Which test type is closest to optimum?

Modelled Fault Testing

• Will detect 100 % of detectable faults

• Vectors can be generated and analyzed in terms of fault

coverage

• Number of detected defects depends on the quality of the fault

model

• Requires ≈ 40 vectors (for a single-stuck-at-faults)

• Test at 10 MHz would take ≈ 4 µs

Page 12: Titel und Thema des Vortrages - uni-potsdam.de

Which test type is closest to optimum?

Exhaustive testing

• Requires 235 = 17.179.869.184 test vectors

• Test at 10 MHz would take 57 min 16 s

Functional testing

• Total functional test will take 96 vectors

• Test at 10 MHz would take 9.6 µs

Modelled Fault Testing

• Requires ≈ 40 vectors (for a single-stuck-at-faults)

• Test at 10 MHz would take ≈ 4 µs

Page 13: Titel und Thema des Vortrages - uni-potsdam.de

Definitions

Fault coverage

Percentage of total faults referring to a fault model for which test

patterns have been generated

Fault coverage = 100 × Number of faults detected by patterns

Total number of faults of the given FM

Page 14: Titel und Thema des Vortrages - uni-potsdam.de

Definitions cont.

Testability

. . . is determined by controllability and observability.

Page 15: Titel und Thema des Vortrages - uni-potsdam.de

Definitions cont.

Testability

. . . is determined by controllability and observability.

Controllability

. . . is the capability of controlling the state of a unit.

⇒ Application of patterns

Page 16: Titel und Thema des Vortrages - uni-potsdam.de

Definitions cont.

Testability

. . . is determined by controllability and observability.

Controllability

. . . is the capability of controlling the state of a unit.

⇒ Application of patterns

Observability

. . . is the capability of observing the state of a unit.

⇒ Reception of test responses

Page 17: Titel und Thema des Vortrages - uni-potsdam.de

Testability

Tes

t inputs

Tes

t res

ponse

UUT-Environment (Rest of Device)

Unit-Under-

UUT inputs UUT outputs

Test

Tester

Testability

Controllability Observability

Page 18: Titel und Thema des Vortrages - uni-potsdam.de

Definitions cont.

Sensitization

Process of driving the circuit to a state where the fault causes an

actual erroneous value at the point of the fault

Propagation

Process of driving the circuit to a state where the error becomes

observable at the primary outputs

Justification

Process of determining the input combination necessary to drive

an internal circuit node to a specific value

Page 19: Titel und Thema des Vortrages - uni-potsdam.de

Fault models

2 Fault models

Types of Faults

Stuck-at Fault Model

Stuck-open Fault Model

Bridging Faults

Delay Fault Model

IDDQ Model

Page 20: Titel und Thema des Vortrages - uni-potsdam.de

Types of Faults

Functional faults (bugs)

. . . are defined in conjunction with a functional model.

⇒ Introduced in design process

⇒ Shall be detected during simulation/verification

⇒ Require a feasible test bench

⇒ No automated test generation

Page 21: Titel und Thema des Vortrages - uni-potsdam.de

Types of Faults

Structural faults (defects)

. . . are defined in conjunction with a structural model of the system

and a logical fault model.

⇒ Introduced during fabrication

⇒ Not detectable during simulation, detected via testing

⇒ Automated test generation is possible, since the function of the

system in case of a fault of the given fault model can be

determined.

Page 22: Titel und Thema des Vortrages - uni-potsdam.de

Goals of Fault Modelling

• Fault model

• Abstract model of a defect, i.e. something that can go

wrong during the fabrication of a system.

• Goal of fault models

• Model defects at the highest level of abstraction

possible

• Model as high a percentage as possible of the actual

physical defects that can occur

Page 23: Titel und Thema des Vortrages - uni-potsdam.de

Stuck-at Fault Model

Page 24: Titel und Thema des Vortrages - uni-potsdam.de

Stuck-at Fault Model

• Stuck-at-0: A net is permanently connected to GND

s-a-0

GND

• Stuck-at-1: A net is permanently connected to VDD

VDD

s-a-1

Page 25: Titel und Thema des Vortrages - uni-potsdam.de

Stuck-at Fault Model Cont.

x s-a-0 (1)

x2

• Output stuck-at fault model:

x = s − a − 0 /1 ⇒ x = x1 = x2 = s − a − 0 /1

s-a-0 (1) x1 x

x2

• Input stuck-at fault model:

x1 = s − a − 0 /1 ≠> x2 = s − a − 0 /1

x1

Page 26: Titel und Thema des Vortrages - uni-potsdam.de

Single vs. Multiple Stuck-at Fault Model

Single stuck fault model

• Assumes that only one wire is permanently stuck

Reasonable number of faults 2n (n number of circuit nodes)

Efficient algorithms for ATPG and fault simulation are well

developed

Covers about 90% of possible manufacturing defects in CMOS

circuits

Other fault models can be mapped into stuck-at faults

Does not cover all defects

Does not consider that faults can mask each other

Page 27: Titel und Thema des Vortrages - uni-potsdam.de

Single vs. Multiple Stuck-at Fault Model

Multiple stuck fault model

• Assumes that 2 or more wires are permanently stuck

Fault coverage higher than for single stuck-at fault model

Does not cover a significantly larger number of defects than

single stuck-at

Large number of faults 3n − 1 (n number of circuit nodes)

Algorithms for ATPG and fault simulation are much more

complex

Page 28: Titel und Thema des Vortrages - uni-potsdam.de

Stuck-open Fault Model

Page 29: Titel und Thema des Vortrages - uni-potsdam.de

Stuck-open Fault Model

• A physical line is broken (high

impedance)

• Such breaks result in memory effects for

some undetermined discharge time

Covers physical defects not covered by

stuck-at fault models

Large number of tests

Algorithms for ATPG and fault simulation

are much more complex

Requires a low level circuit description for

faults within logic elements

A

B

VDD

GND

F Line

break

Page 30: Titel und Thema des Vortrages - uni-potsdam.de

Bridging Faults

Page 31: Titel und Thema des Vortrages - uni-potsdam.de

Bridging Faults

• Two nodes of a circuit are shortened together

• Usually assumed to be a low resistance path

• Three classes

Bridging within a logic element

Bridging between logic nodes without feedback

Bridging between logic nodes with feedback VDD

GND

A

B

Bridging

F

Page 32: Titel und Thema des Vortrages - uni-potsdam.de

Bridging Faults

• Two nodes of a circuit are shortened together

• Usually assumed to be a low resistance path

• Three classes

Bridging within a logic element

Bridging between logic nodes without feedback

Bridging between logic nodes with feedback

Bridging

Page 33: Titel und Thema des Vortrages - uni-potsdam.de

Bridging Faults

• Two nodes of a circuit are shortened together

• Usually assumed to be a low resistance path

• Three classes

Bridging within a logic element

Bridging between logic nodes without feedback

Bridging between logic nodes with feedback

Bridging

Page 34: Titel und Thema des Vortrages - uni-potsdam.de

Bridging Faults

• Two nodes of a circuit are shortened together

• Usually assumed to be a low resistance path

• Three classes

Bridging within a logic element Bridging between logic nodes without feedback

Bridging between logic nodes with feedback

Covers a large number of physical defects – 30 % of all defects

are bridging faults

ATPG algorithms are very complex

Requires a low level circuit description for bridging faults

within logic elements

Page 35: Titel und Thema des Vortrages - uni-potsdam.de

Delay Fault Model

Page 36: Titel und Thema des Vortrages - uni-potsdam.de

Delay fault model

• The logic function of the UUT is error free

• Some physical defect (e.g. process variation) makes some

delays greater than some defined bounds

• Two delay fault models:

Gate delay, or transitional fault model

Path delay fault model

Page 37: Titel und Thema des Vortrages - uni-potsdam.de

Transitional delay fault model

• A logical model for a defect that delays either a rising or

falling transition

Slow-to-rise

Slow-to-fall

011

000

001

011

Slow-to-rise

Fault free

• Tests are similar to stuck-at fault tests

Initialize a line to 0 and test for a s-a-0 fault to detect

slow-to-rise transition fault

• Two patterns are required: initialization/launch and transition

detection/capture

Reasonable number of faults 2n (n number of nodes)

Faulty delays of gates can be compensated by other gates

Page 38: Titel und Thema des Vortrages - uni-potsdam.de

Path delay fault model

Delay = 2

Delay = 2

Delay = 6

Delay = 2

B

A

C

E

Z

Delay = 2 D

Delay = 2

• The total delays of a path from its inputs to its outputs exceed

some maximum value

Detects more faults than transition delay fault model

Can be used with more aggressive statistical design

philosophy

Large number of possible paths in the circuit

Algorithms for ATPG are complex

Page 39: Titel und Thema des Vortrages - uni-potsdam.de

IDDQ Model

Page 40: Titel und Thema des Vortrages - uni-potsdam.de

IDDQ Model

defect

no defect

I

O I

VDD

GND

O

defect

• Many defects in CMOS circuits can be detected by current

measuring techniques

• A fully static CMOS gate consumes significant current only

when switching

• Most physical defects will raise the quiescent current for MOS

devices (IDDQ) by several orders of magnitude

Page 41: Titel und Thema des Vortrages - uni-potsdam.de

IDDQ Model and Testing

Test generation is easier since faults must be activated, but

not propagated to a primary output

IDDQ testing detects defects not modelled by stuck-at fault

models

Normal IDDQ is very low ⇒ precise measurement required

Measurement takes a significant amount of time

Problematical in combination with small cell geometries

Page 42: Titel und Thema des Vortrages - uni-potsdam.de
Page 43: Titel und Thema des Vortrages - uni-potsdam.de

Test Generation Flow

a-s-a-0

a-s-a-1

b-s-a-0

b-s-a-1

c-s-a-0

c-s-a-1

d-s-a-0

d-s-a-1

...

Generate

list of

undetected

faults

Select

undetected

fault for test

generation

Generate a test

vector for that

fault

Generate list of

other faults

detected by

that vector

a-s-a-0

c-s-a-1

d-s-a-1

h-s-a-0

l-s-a-1

m-s-a-1

s-s-a-0

...

Select a fault

for test

generation

101110...

X a-s-a-0

a-s-a-1

b-s-a-0

b-s-a-1

c-s-a-0

X c-s-a-1

d-s-a-0

X d-s-a-1

...

Circuit-Under-

Test

a-s-0

Yes Exit All faults detected or

proven untestable?

Page 44: Titel und Thema des Vortrages - uni-potsdam.de

Test Generation Flow

a-s-a-0

a-s-a-1

b-s-a-0

b-s-a-1

c-s-a-0

c-s-a-1

d-s-a-0

d-s-a-1

...

Generate

list of

undetected

faults

Select

undetected

fault for test

generation

Generate a test

vector for that

fault

Generate list of

other faults

detected by

that vector

a-s-a-0

c-s-a-1

d-s-a-1

h-s-a-0

l-s-a-1

m-s-a-1

s-s-a-0

...

Select a fault

for test

generation

101110...

X a-s-a-0

a-s-a-1

b-s-a-0

b-s-a-1

c-s-a-0

X c-s-a-1

d-s-a-0

X d-s-a-1

...

Circuit-Under-

Test

a-s-0

Ye

s All faults detected or

proven untestable?

Exit

Adequate

fault coverage?

Ye

s

Redesign/DfT-

insertion

Page 45: Titel und Thema des Vortrages - uni-potsdam.de

Test Generation Flow

• Test vectors developed by the design team often must be

significantly modified by the test team since the vectors are

incompatible with Automatic Test Equipment (ATE)

Overflow of scan-data, format-data, or timing-data

memory

Vectors set is not compact enough to fit in pattern

memory

Vectors include comparison with tristate values

Page 46: Titel und Thema des Vortrages - uni-potsdam.de

Design-for-Testability

4 Design-for-Testability

Definition of Design-for-Testability

Structured DfT – Scan Design

Built-In Self-Test (BIST)

On-line Testing

Page 47: Titel und Thema des Vortrages - uni-potsdam.de

Design-for-Testability

The goal of Design-for-Testability

. . . is to increase the ease with which a device can be tested.

⇒ Increase the controllability and observability of internal points.

Categories of Techniques

• Ad-Hoc

Partitioning

Test points

• Structured Techniques

Scan Design

Boundary Scan

Page 48: Titel und Thema des Vortrages - uni-potsdam.de

Ad-Hoc DfT

Device-Under-Test

Page 49: Titel und Thema des Vortrages - uni-potsdam.de

Ad-Hoc DfT

Module

1

Module

3

Module

2

• Decompose the system into subcomponents

Page 50: Titel und Thema des Vortrages - uni-potsdam.de

Ad-Hoc DfT

Module

1

Module 2

TDI1

TEN

TDI2

Module

3

TDO1

TDO2

• Decompose the system into subcomponents

• Add control and observation points

Page 51: Titel und Thema des Vortrages - uni-potsdam.de

Structured DfT – Scan Design

Page 52: Titel und Thema des Vortrages - uni-potsdam.de

Scan test

• Make a FSM testable by making the internal state

variables controllable and

observable

• Replacing latches and flip-flops by their scannable

counterparts

• Connect latches and

flip-flops to shift-registers

• Provide a tester interface

(SEN, SDI, SDO)

• Test: Shift vectors in, apply 1 functional clock cycle,

shift response out

SI

SO

MU

X

SE SET

RS

D

SI

Q

Q=SO

CK

Q D

Q RST

SET

Page 53: Titel und Thema des Vortrages - uni-potsdam.de

Scan test

• Make a FSM testable by making the internal state

variables controllable and

observable

• Replacing latches and flip-flops by their scannable

counterparts

• Connect latches and

flip-flops to shift-registers

• Provide a tester interface

(SEN, SDI, SDO)

• Test: Shift vectors in, apply 1 functional clock cycle,

shift response out

SI1

SO1

SI2

SO2

SI3

SO3

Page 54: Titel und Thema des Vortrages - uni-potsdam.de

Level Sensitive Scan Design (LSSD)

• Used to realize scannable latches

• Introduces a second latch used to avoid transparency

• Requires 2 additional non-overlapping shift clocks

L2 L1 TCK1

SI

D

C

Q

TCK2

SO

Page 55: Titel und Thema des Vortrages - uni-potsdam.de

Level Sensitive Scan Design (LSSD)

• Used to realize scannable latches

• Introduces a second latch used to avoid transparency

• Requires 2 additional non-overlapping shift clocks M

UX

SI

D CK

TCK2

Q

SO

SE

RS

SET

D Q

G RST

SET

SET

D Q

G RST

MU

X

TCK1

L2

L2

L2

Q1

Q0

Q2

SO

SI

C

D1

D2

D0

TCK1

TCK2

S I

TC K 1

L1 D

C

S I

TC K 1

L1

D

C

S I

TC K 1

L1

D

C

TC K 2

TC K 2

TC K 2

Page 56: Titel und Thema des Vortrages - uni-potsdam.de

LSSD – Design Rules

1. All internal storage is implemented in hazard-free polarity-hold

latches

2. The latches are controlled by two or more non-overlapping

clocks such that latches that feed one another can not have

the same clock

3. It must be possible to identify a set of clock primary inputs

from which the clock inputs to to shift-register latches (SRLs)

are controlled either through simple powering trees or through

logic that is gated by SRLs and/or non-clock primary inputs

4. Clock primary inputs my not feed the data inputs to latches

either directly or through combinational logic, but may only

feed the clock input to the latches or the primary outputs

Page 57: Titel und Thema des Vortrages - uni-potsdam.de

Boundary Scan

Source: Application-Specific Integrated Circuits - Michael J. S. Smith

Page 58: Titel und Thema des Vortrages - uni-potsdam.de

Boundary Scan Cell

Source: Application-Specific Integrated Circuits - Michael J. S. Smith

Page 59: Titel und Thema des Vortrages - uni-potsdam.de

TAP Controller

Source: Application-Specific Integrated Circuits - Michael J. S. Smith

Page 60: Titel und Thema des Vortrages - uni-potsdam.de

Boundary Scan – Modes

• Normal mode Mode=0: Data

passes from IN to OUT

• Scan mode Shi f t=1 ,

Clock=scan clock: Serial

data is shifted in from SIN

and to SOUT

• Capture mode Shi f t=0 , Clock=std clock: Data on

the IN line is clocked into QA

• Update mode with QA

loaded, Mode=1: Data in QA

is applied to OUT

TDI

MU

X

Application Logic

Optional:

BIST registers

Scan registers

T

A

P Bypass Regs.

Instr. Regs.

Op. Regs.

SI

SO

TMS

TCK

TDO

Boundary-scan cell I/O Pad

Boundary-scan path

MU

X

IN

SIN

D Q

QA

D Q

QB

MU

X

OUT

SO

Shift Clock Update

Page 61: Titel und Thema des Vortrages - uni-potsdam.de

Advantages of scan technique

Transforms the testing problem from one of sequential circuit

testing to one of combinational testing

Can be used to test for a variety of fault models (stuck-at,

bridging, delay faults, IDDQ tests)

Eases functional testing due to additional controllability and

observability

Good diagnosis capabilities

Area and speed overhead due to scan cells

Limited at-speed testing

Clock generation and distribution for LSSD

Page 62: Titel und Thema des Vortrages - uni-potsdam.de

Built-In Self-Test (BIST)

Page 63: Titel und Thema des Vortrages - uni-potsdam.de

Built-In Self-Test

• Perform at-speed tests of internal IPs

• Collecting each output response and off-loading it for

comparison is too inefficient to be practical

⇒ Compress the entire output stream into a single

signature value

⇒ Evaluate and analyze signature at the end of the test

Page 64: Titel und Thema des Vortrages - uni-potsdam.de

Built-In Self-Test

output input Tout Ten

TPG

TRA

UUT

MU

X

• Test-Pattern-Generator (TPG) generates patterns for the

unit-under-test (UUT)

• Test-Response-Analyser (TRA) receives the output of the UUT

and generates a signature

Page 65: Titel und Thema des Vortrages - uni-potsdam.de

Built-In Self-Test – TPG

D Q

R

D Q

S

D Q

S

rst

clk

• Implemented using a simple Linear-Feedback-Shift-Register

(LFSR)

• Generates all possible values except →0

• Often made scannable for setting an initial seed

Page 66: Titel und Thema des Vortrages - uni-potsdam.de

BIST – Canonial Form of LFSRs and MISRs

+

c1

+

c2 cn-1

+

D Q

Q1 Q2 Q3 Qn

D Q D Q D Q

cn

(a) LFSR Type 1

cn-1 cn-2 cn c1

+ D Q +

Q1 Q3 Qn

D Q +

Q2

D Q D Q

(b) LFSR Type 2

P (x) = 1 + c1x + c2x2 + . . . + cnxn

Page 67: Titel und Thema des Vortrages - uni-potsdam.de

Built-In Self-Test – TRA

D Q

R

D Q

S

D Q

S

Tout

Input data

rst

clk

ienable

• Implemented using a Multi Input Signature Register (MISR)

• Input enable signal for excluding undesirable response

vectors (e.g. that includes X values)

• Made scannable for evaluation of the stored pattern

Page 68: Titel und Thema des Vortrages - uni-potsdam.de

Signature Analysis

• Based on the concept of cyclic redundancy checking (CRC)

• Choose ci, such that the possibility of masking faults is

minimized

+

cn-1

+

cn-2 cn c1

+ D Q

Q1 Q2 Q3 Qn

D Q D Q D Q +

D1 D2 D3 Dn

(c) MISR

Page 69: Titel und Thema des Vortrages - uni-potsdam.de

Signature Analysis

• Based on the concept of cyclic redundancy checking (CRC)

• Choose ci, such that the possibility of masking faults is

minimized

D Q

Q

D Q

Q

D Q

Q

D Q

Q

MU

X

Q1 Q2 Q3

Q4=SO

D1 D2 D3 D4 B1

B2

SI

(d) BILBO

Page 70: Titel und Thema des Vortrages - uni-potsdam.de

BIST with LUTs

• Use look-up tables in case

predefined patterns are required

• Can be combined with LFSRs to

keep the overhead low, e.g.

Generate frame via LUT

Generate payload via LFSR

Pattern

Response

MISR Signature

Pattern-LUT

Test enable

Counter

Pattern address

MUX

MISR

UUT

Datain

Dataout

Page 71: Titel und Thema des Vortrages - uni-potsdam.de

BIST with LUTs

• Use look-up tables in case

predefined patterns are required

• Can be combined with LFSRs to

keep the overhead low, e.g.

Generate frame via LUT

Generate payload via LFSR

Pattern

Response

Test enable

MUX

UUT

Datain

Dataout

Address

Pattern-LUT

Counter

Response

LUT

Comparator

Fault

Page 72: Titel und Thema des Vortrages - uni-potsdam.de

Hierarchical BIST

TEn

Din Dout

TEn TEn TEn

IP

IP

Tester Interface

BIST Controller

IP IP

MU

X

MU

X

MU

X

TPG

TPG

TPG

TPG

TPG

TPG

TPG TPG

IP

• Divide and conquer

• Insert global test as well as local test capabilities

Page 73: Titel und Thema des Vortrages - uni-potsdam.de

Advantages of BIST

Can be used in case of synchronization problems with tester,

e.g. for heterogeneous systems (GALS, asynch. circuits)

Can be used to perform at-speed tests

Detects transient and intermittent faults

Limited diagnosis capabilities

Faults within check circuitry may reduce yield

Page 74: Titel und Thema des Vortrages - uni-potsdam.de

On-line Testing

Page 75: Titel und Thema des Vortrages - uni-potsdam.de

On-line Testing

• Introduce redundant logic to perform on-chip tests during

normal operation

• Predictor-Circuits: Usage of error-detecting (EDC) (or even

-correcting (ECC)) codes

• Watchdog circuits: Often used in case the UUT has an

invariant property

Unit-Under-Test

Generator

Predictor Com

para

tor

input output

fault

(e) Predictor-Circuits

Unit-

Under-

Test

fault

WatchDog

(f) Watchdog

Page 76: Titel und Thema des Vortrages - uni-potsdam.de

On-line Testing – An example

• Arithmetic operations (addition, subtraction, multiplication)

can be checked through residue class arithmetic modulo p by

using the following equations:

Page 77: Titel und Thema des Vortrages - uni-potsdam.de

On-line Testing – An example

Page 78: Titel und Thema des Vortrages - uni-potsdam.de

On-line Testing – An example

c = a b

n n

k

r2 = < e >p

m

f = (r1 == r2)?

2

f

a b

c

< a >p < b >p

m m

Predictor

r1 = << a >p < b >p>p

m

Generator

Comparator

Page 79: Titel und Thema des Vortrages - uni-potsdam.de

Comparators

• Use self-checking comparators for test response generation

a

f

a’

b

b’

f'

Page 80: Titel und Thema des Vortrages - uni-potsdam.de

Advantages of On-line testing

Allows at-speed tests

Detection of intermittent and transient faults

Allows graceful degradation techniques

Additional hardware overhead for check components

Faults within check circuitry may reduce yield

Page 81: Titel und Thema des Vortrages - uni-potsdam.de

Redundant logic and undetectable faults

• Redundant logic introduces undetectable faults

• Additional DfT (control and observation points) may help

FF

FF

FF

Voter

x

Clk

s-a-x

Page 82: Titel und Thema des Vortrages - uni-potsdam.de

Conclusions

Page 83: Titel und Thema des Vortrages - uni-potsdam.de

Summary of DfT-Techniques

Scan Test BIST On-line Test Observability & controllability

Fault coverage

Influences in area

Influences in

Performance

At-speed testing

Diagnose

Capabilities

Test time

Page 84: Titel und Thema des Vortrages - uni-potsdam.de

Conclusions

• Integrate DfT-Techniques into design specification

• Divide and Conquer

• Make use of hierarchical DfT techniques

• Functional tests are not sufficient!

• Insert scan as a prerequisite for performing structural tests!

• Be aware of test problems with redundant logic! (Requires

additional DfT)

Page 85: Titel und Thema des Vortrages - uni-potsdam.de

Thank you for your attention!

Page 86: Titel und Thema des Vortrages - uni-potsdam.de

References

• RASSP Educational & Facilitation Program, Test Technology

Overview, h t t p : / /www.cedcc .psu.edu/ee497i / rassp_43 /s ld001.htm ,

1998

• M. Abramovici, M.A. Breuer and A.D. Friedman: Digital

Systems Testing and Testable Design, 1990

• Michael Gössel and Steffen Graf: Error Detection Circuits,

1993