Fault detection and diagnosis with parity trees for space compaction of test responses

This paper shows that accurate fault detection and diagnosis are possible when applying a parity tree for space compaction of test responses and continuously observing the parity-tree output. We exploit that each set of faults results in a unique parity-tree output sequence that can be used as a unique fault signature for truly accurate fault detection and diagnosis. Our theoretical analysis shows that probabilities for fault cancellation and fault aliasing are extremely low and decrease with circuit size. Experimental results show that for typical scan chain architecture in large industrial circuits, fault coverage is not affected by parity-tree space compaction for up to 256 scan chains, while diagnostic resolution is reduced by only 1% to 2%