Detectability of CMOS stuck-open faults using random and pseudorandom test sequences

An analysis is presented of CMOS stuck-open faults tested with a pseudorandom test sequence, i.e. a test sequence consisting of some or all of the 2^N test patterns generated by a modified N-bit linear-feedback shift register (LFSR). Such a scheme is viewed as testing without replacement. When all 2^N test patterns are applied, then such a test sequence is called a pseudorandom exhaustive test sequence (PRETS). The alternative scheme is called random testing, which corresponds to sampling the population of test vectors with replacement. It is shown that some stuck-open faults require a single test vector of their detection, while most require an ordered pair of test vectors. A PRETS will detect all stuck-open faults that require a single test vector, but may not necessarily detect all faults that require an ordered pair of test vectors. The results obtained under pseudorandom testing are compared with results obtained under random testing