Dynamic Test Set Selection Using Implication-Based On-Chip Diagnosis
−N detj τ , where Wi,j(t) represents the weight to be used for fault j when the t th pattern of the test subset is to be chosen and N detj represents the number of times that fault j has been detected by the t − 1 patterns that have already been added to the test subset. Once the new weights are obtained, we recompute the score for each pattern remaining in the superset. The unselected pattern with the highest score is then added to this implication’s test pattern assignment list. We repeat this process until the desired number of patterns are included in the list. The same procedure is followed for all of the implications/groups. Our results show that high diagnostic resolution may be obtained with the proposed method. When implications are not grouped, we can achieve highly targeted test sets that detect the suspect faults multiple times. Even when we group the implications to constrain the hardware overhead, the number of detections per suspect fault is still high, and thus these test sets are appropriate for fortuitous detection.