A Test Pattern Quality Metric for Diagnosis of Multiple Stuck-at and Transition faults

The test patterns computed for detecting the manufacturing defects in the electronic circuits are generally insufficient for diagnosis. The test set compaction and failure log truncation lead to loss of critical failure observations that diagnosis might depend on. In this context, it is beneficial to know the diagnostic usefulness of failures so that we can log the more useful failures instead of logging the initial failures. In this paper, we evaluate three metrics to gauge such diagnostic usefulness in real-time by observing the circuit responses on the tester. We implement a pattern selection framework for failure logging and compare the results with those achieved by logging the initial failures. Using one of our proposed metrics, we were able to improve the diagnosis quality for a significant number of faulty instances of ISCAS'89 and IWLS'05 benchmarks having 1-7 inserted stuck-at and transition faults.