Efficient test set evaluation

Abstract

The fault coverage obtained by a set of test patterns is usually determined by expensive fault simulation. Even when using fault dropping techniques, fault simulation provides more information than actually needed. For each fault, the pattern is determined which detects this fault first. This is mainly redundant information if diagnosis is not required. One can dispense with this high resolution and restrict interest to the set of faults which is detected by a set of patterns. It is shown theoretically and practically that this information is obtainable in an highly efficient way.<>