Poage has constructed a complex fault detection algorithm which generates a complete and minimal test set of all multiple stuck-at faults of a given combinational network. Several authors have derived from his method fast and simple multiple fault detection algorithms, which are claimed to generate complete test sets with a "near-minimal" or "near-optimal" number of tests. We show that the algorithms by Bossen and Hong and the algorithm by Yang and Yau may generate test sets with an exponential number of tests (relative to the number of inputs) where a linear number of tests is sufficient for a complete multiple fault detection test set.

A design procedure is shown which allows the construction of one-dimensional iterative systems of combinational cells with good single stuck-at fault test complexity. Let S be a system where one of the M input words and two of the N states are defined for test purposes and the cells are realized as AND–EXOR normal form circuits, then S is testable with at most [log2(M x N) + 6] tests for all single faults and for any number of cells. A weaker restriction is shown, where no normal form realization is necessary, and which gives good testability for single or multiple stuck-at faults of the input/output and interconnection lines of cells.