Self-Testing with Correlated Faults

Accurate prediction of the number of self-test random patterns that will provide low defect levels is crucial. This prediction is complicated by fault-to-fault correlating effects. This paper presents a technique to assess the number of random patterns required to test to a given defect level in the presence of fault correlation. This paper will also develop approximation algorithms that enable test engineers to readily compute defect levels as a function of test coverage with correlated faults present. This paper thus enhances previous work which assumed fault-to-fault independence in determining random pattern sufficiency and defect levels.