Browse Prior Art Database

Index of Slack as Delay Fault Measurement Tool

IP.com Disclosure Number: IPCOM000083387D
Original Publication Date: 1975-May-01
Included in the Prior Art Database: 2005-Mar-01
Document File: 4 page(s) / 62K

Publishing Venue

IBM

Related People

Snethen, TJ: AUTHOR [+2]

Abstract

SUMMARY: With the advent of LSI (Large-Scale Integration), increasing importance is being placed on an ability to provide a screening of logic chips by delay testing. Assuming delay tests can be generated, it is necessary to be capable of providing data on the quality (or effectiveness) of those tests applied. Described is a measurement which must be made in order to accurately compute delay test coverage.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 41% of the total text.

Page 1 of 4

Index of Slack as Delay Fault Measurement Tool

SUMMARY: With the advent of LSI (Large-Scale Integration), increasing importance is being placed on an ability to provide a screening of logic chips by delay testing. Assuming delay tests can be generated, it is necessary to be capable of providing data on the quality (or effectiveness) of those tests applied. Described is a measurement which must be made in order to accurately compute delay test coverage.

THE PROBLEM: Many oE the problems in delay fault coverage computation result from inadequate tests, but also on the behavior of the delay fault, the tester environment, the logic configuration, and others. An accurate computation of coverage must include such factors. Previous discussion on probability of detection has been concerned with only certain forms of delay tests (have excluded latch-to-latch tests, for example). However, delay faults may be detected on all forms of delay tests. It must be possible, to provide data which not only covers all tests but measures the difference between these tests. Even given a single delay test, it is possible that certain delay faults may be tested more 'stringently' than others.

Clearly there is a need for a measurement which provides detection data for all forms of delay tests, and for all degrees of detection within these tests.

THE SOLUTION: The solution is to measure the difference in delay between the path of which a given delay fault lies, and the 'measurement path'. This difference is call 'slack'.

In a simple PI (input) to PO (output) delay test, a test can appear as shown in Fig. 1.

A transition is placed on the P1 and propagates to the PO. Then waiting 'delta' after putting the transition on the PI, the PO is sampled. It is clear that there is only one path of propagation on the chip, and that all delay faults along this path are detected to the same degree. If 'delta' is adjusted so that both paths have the same delay, all detected faults are detected with slack equal to zero.

In Fig. 2, the value of delta would normally be set based on the delay in the upper path of transition, since this path causes the PO to change. Note the lower path transition must occur also for the AND to switch, but has 2 additional block delays of 'leeway' before causing the delay test to fail. [A failure hero would be manifested as an absence of a transition by the time the sample was taken].

Note the delay along the lower path is 2 block delays less than the measurement path. Thus, those faults detected along that path are detected with a slack equal to 2 block delays, while the upper path faults are detected with zero slack.

Logic as shown in Fig. 2, should be rare. However, a relatively common type of test is shown in Fig. 3. This type of test is called a secondary test and may represent a significant percentage of the forms of detection in latch-trigger designs. Here again there is an upper and lower path of transition. If the two

1

Page 2 of 4

path...