Digital circuit testing and design for testability

The design methods presented in earlier chapters will allow an initial specification to be translated into a practical circuit whose logical behaviour is correct and whose operating speed and other physical factors are predictable and well defined. However, it is also necessary that such circuits operate reliably, and that any faults which affect their operation may be detected easily (and hence cheaply). In most cases, our systems are constructed from components (from simple gates to VLSI devices) which are assembled on to printed circuit boards. In such cases, a wide range of different faults can exist, from faulty components to incorrect PCB manufacture and assembly, and we require methods to detect and locate such faults. Obviously, if all components were perfect and no errors ever occurred during system manufacture, testing would be unnecessary, but, if the probability of correct component operation is C, then the probability, P n , of correct operation of a system containing n such components can be no better than C n , assuming that the failure of any component signifies failure of the system (since P n = C 1 × C 2 × ... × C n ). This means for example, that if C = 99% and the system contains 100 components, then P n = 37%, so that almost two-thirds of the systems produced may contain faults.