Efficient Double Fault Diagnosis for CMOS Logic Circuits With a Specific Application to Generic Bridging Faults

Fault diagnosis that predicts the most likely fault sites in a faulty chip is an important step for manufacturing yield improvement or design debugging. In this paper, we address the problem of double fault diagnosis for full-scan designs. Our algorithm aims to identify both faults accurately. The features of our algorithm include the following. (1) The proposed algorithm is not limited to any particular fault type. (2) An effective selection heuristic is incorporated to significantly reduce the diagnosis time, while retaining a high success rate of catching faults. (3) The inject-and-evaluate paradigm proposed in [6] is incorporated to accurately screen out unlikely fault candidates. Experimental results on ISCAS85 benchmark circuits injected with a generic bridging fault, two stuck-at faults, or two gate-type faults show that both faults can be caught simultaneously within several minutes with a high success rate.