A Fault-Independent Transitive Closure Algorithm for Redundancy Identification

We present a fault-independent redundancy identification algorithm. The controllabilities and observabilities are defined as Boolean variables and represented on an implication graph. A major enhancement over previously published results is that we include all direct and partial implications, as well as node fixation. The transitive closure, whose computation now requires a new algorithm, provides many redundant faults in a single-pass analysis. Because of these improvements, we obtain better performance than all previous faultindependent methods at execution speeds that are much faster than any exhaustive ATPG. For example, in the s9234 circuit more than half of the redundant faults are found in just 14 seconds on a Sparc 5. All 34 redundant faults of c6288 are found in one pass. Besides, our single pass procedure can classify faults according to the causes of their redundancy. The weakness of our method, as we illustrate by examples, lies in the lack of a formulation for the observabilities of fanout stems.