Worst-case analysis for pseudorandom testing

The testing problem of combinational circuits with pseudorandom patterns is investigated. Work by previous workers for single faults is extended to multiple faults situations; in addition, masking effects between disjoint/conjoint faults are considered. An analytical model based on Markov chains with any number of states is proposed for random/pseudorandom testing and relationships between test length and test confidence are developed. Evaluations of the model for double and triple faults are presented using well-known examples. The results presented in this paper are useful for BIST systems that use random/pseudorandom input patterns.<>