Classes of difficult-to-diagnose transition fault clusters
Irith Pomeranz · 2013
An earlier paper showed that the accuracy of a fault diagnosis procedure based on single faults is reduced when it attempts to diagnose transition faults that are clustered in an area. This is due to the increased possibility that the effects of clustered faults will interact. This paper defines three different types of clusters. In path-based clusters, transition faults are clustered along a subpath. In adjacency-based clusters, transition faults are clustered on adjacent lines. In cone-based clusters, transition faults are clustered in a logic cone. Adjacency-based clusters were considered earlier. This paper shows that cone-based clusters are more difficult to diagnose. They thus provide a more challenging target for diagnosis procedures based on single faults, and they can explain reports of difficult-to-diagnose defects.