Browse Prior Art Database

Random Adaptive Test Technique

IP.com Disclosure Number: IPCOM000080085D
Original Publication Date: 1973-Oct-01
Included in the Prior Art Database: 2005-Feb-26
Document File: 2 page(s) / 14K

Publishing Venue

IBM

Related People

Storey, TM: AUTHOR

Abstract

This is a technique for analyzing large-scale integrated (LSI) circuitry and identifying reset lines independent of circuit density or logic breakdown, by using a random adaptive test technique.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Random Adaptive Test Technique

This is a technique for analyzing large-scale integrated (LSI) circuitry and identifying reset lines independent of circuit density or logic breakdown, by using a random adaptive test technique.

The fundamentals of random testing were first promulgated in US Patent 3,614,608 issued on October 19, 1971. This patent sets forth the requirements for and advantages of testing with statistically random-test patterns. It was subsequently found that the weighting of random-test patterns significantly improved testability. This latter concept of weighted test patterns is set forth in US Patent 3,719,885 issued March 6, 1973. Since the concept of weighted random-test patterns was invented, various improvements to this basic concept related to the method of the "weighting" have been devised.

One such method of determining the weighting related to reset identification is based upon the behavior of "reset-like" primary input lines showing initial high jumps in activity, and then tapering off as the number of patterns simulated increases. On the other hand, the "clock-like" lines show a steady slope upward of activity. With the advent of LSI and its wide variety circuit density, the approach is incapable of recognizing these two types without prior human intervention. Namely, the program is incapable of the following calculations: 1. Determination of what activity increases justify examination as a "reset" candidate. 2. Determination of what interval should constitute a significant number of patterns in which to monitor activity increments, and how many. 3. How many patterns in total reflect activity response. 4. Determination of optimum reset weight.

Improper determination of the above values could result in over-emphasis, underemphasis, or quite possibly, no observations concerning the true reset lines. On the other hand, many more lines than necessary are examined, wasting much computer processing time. A truly automatic algorithm must be able to properly analyze the internal circuitry of the circuit under test to determine these critical interval values.

By the present technique, the above-named problems are solved by using the following properties of a "reset-line" line: 1) Only those lines which can produce a high level of activity can seriously effect the degree of percent tested acquired. 2) Although a line activity may remain "dormant" for the majority of patterns applied, a sudden jump in...