On the relationship between stuck-at fault coverage and transition fault coverage

The single stuck-at fault coverage is often seen as a figure-of-merit also for scan testing according to other fault models like transition faults, bridging faults, crosstalk faults, etc. This paper analyzes how far this assumption is justified. Since the scan test infrastructure allows reaching states not reachable in the application mode, and since faults only detectable in such unreachable states are not relevant in the application mode, we distinguish those irrelevant faults from relevant faults, i.e. faults detectable in the application mode. We prove that every combinatorial circuit with exactly 100% stuck-at fault coverage has 100% transition fault test coverage for those faults which are relevant in the application. This does not necessarily imply that combinatorial circuits with almost 100% single-stuck at coverage automatically have high transition fault coverage. This is shown in an extreme example of a circuit with nearly 100% stuck-at coverage, but 0% transition fault coverage.