When Optimized N-Detect Test Sets are Biased: An Investigation of Cell-Aware-Type Faults and N-Detect Stuck-At ATPG

Fanchen Zhang, Micah Thornton, Jennifer Dworak

Abstract: Cell-aware faults have previously been proposed to more effectively detect defects within gates. At the same time, n-detect test sets that provide multiple detections of each stuck-at fault are often used to maximize the detection of unmodeled defects. However, n-detect test sets are often not particularly effective at fortuitously detecting all untargeted cell-aware faults. In this paper, we investigate the effectiveness of different types of n-detect ATPG test sets for efficiently detecting difficult cell-aware-type faults and explain why optimizing test sets for n- detect using stuck-at faults while still keeping pattern counts low can actually bias those test sets against the detection of some cell-aware type faults. We then investigate the addition of cell-aware top-off patterns for cell-aware-type faults that are shown to be functionally relevant through good state simulation, allowing such faults to be prioritized when testing resources are limited.