Recursive Coverage Projection of Test Sets

In the generation of test sets for the detection of stuck-type faults in combinational switching networks, it is an expedient and reasonably common assumption to consider explicitly faults only of specified sizes (for example, all single faults), and then to assume (or hope) that most or all faults of larger sizes will be covered (that is, detected) as well. This paper systematically addresses this aspect of multiple fault coverage in a quantitative manner for combinational networks, wherein only primary input fanout is allowed. A procedure is given to estimate (or project) the multiple fault coverage capability of a test set based on the known coverage capability of that test set for subsets of the multiple faults. This is accomplished by means of a recursive use of a detailed formula which exploits two fundamental interrelationships between test sets and faults. Based upon these results, it can be shown that the above-mentioned assumption must be made, in general, with discretion as its validity is highly network structure/test set dependent.