As was shown in the preceding sections, a fault hypothesis is not needed to perform a random test. However, the estimation of a random test length requires some assumptions.

This chapter describes the calculation of the required test length for combinational circuits. This is a very important case since, in addition to purely combinational circuits it also covers parts of sequential circuits designed using the scan-path method. Recall that scan design (Section 1.1.2) transforms a sequential circuit into a combinational circuit plus scan-path registers.

Most of this chapter is devoted to combinational faults, i.e., assuming that both the fault-free and the faulty circuits are combinational. If the faulty circuit is sequential, then the techniques developed for sequential circuits (Chapter 7) must be used. However, it will appear (particularly in Chapters 7 and 9) that approximate methods for sequential faults use the results presented in Section 6.1 for combinational circuits.

It is assumed that the random input vectors correspond to a constant distribution, i.e., drawing at time t is independent from the input vectors previously drawn.

Calculation of random test length is presented in Section 6.1. Methods for calculating the detection probabilites are explained in Section 6.2. Numerical results for usual circuits are given in Section 6.3.