Abstract
The author points out that to make pseudorandom testing usable for circuits containing random pattern resistant faults, input probabilities are weighted. New results on numerical weight optimization are presented. First, two cost functions which are aimed at minimizing the expected test length and at maximizing the expected fault coverage, respectively, are derived. Optimizations based on precomputed tests are then considered. Applying numerical methods to such optimizations yields a method which improves existing strategies.

This publication has 11 references indexed in Scilit: