Browse Prior Art Database

# Generation of Correlated Bit Patterns for Weighted Random Pattern Testing

IP.com Disclosure Number: IPCOM000104287D
Original Publication Date: 1993-Mar-01
Included in the Prior Art Database: 2005-Mar-19
Document File: 1 page(s) / 33K

IBM

## Related People

Linsker, R: AUTHOR

## Abstract

In conventional weighted random pattern testing of logic circuits, each input bit is generated with a specified probability, independently of the other input bits. Disclosed is a method for generating patterns in which pairs (or larger numbers) of input bits occur with specified joint probabilities. This enables improved fault coverage for both DC and AC test applications.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 89% of the total text.

Generation of Correlated Bit Patterns for Weighted Random Pattern Testing

In conventional weighted random pattern testing of logic
circuits, each input bit is generated with a specified probability,
independently of the other input bits.  Disclosed is a method for
generating patterns in which pairs (or larger numbers) of input bits
occur with specified joint probabilities.  This enables improved
fault coverage for both DC and AC test applications.

The method generates a set of bit patterns that satisfy a
desired joint probability distribution over a subset B1...Bn of the
bits, and comprises the steps of:

1.  predetermining, for said desired joint probability distribution,
a set of Boolean functions F1...Fn, where each Fi is a function
of k input bits;

2.  generating, at each test cycle, a set of pseudorandom bit values
A1...Ak;

3.  computing the values of B1...Bn according to B1=F1(A1...Ak),
B2=F2(A1...Ak), ..., Bn=Fn(A1...Ak); and

4.  repeating steps 2 and 3 a specified number of times.

A suitable set of functions F1...Fn may be constructed as
follows, for the case of a preferred embodiment in which the input
bits A1...Ak are unweighted and uncorrelated.  Choose k such that the
desired joint probabilities can be denoted approximately by
N(B1...Bn)/2 sup k where each N(B1...Bn) is an integer.  Generate a
truth table having inputs A1...Ak and outputs B1...Bn such that each
output string B1...Bn occurs N(B1.....