Fault diagnosis of analog integrated circuits using response surface methods

Abstract

This paper demonstrates how response surface methods can be applied in the fault diagnosis of analog integrated circuits. The method involves experimental design and principal components analysis to identify which parameters appear to be important and which are irrelevant, so that they can be fixed at some reasonable value and omitted from further considerations. When the principal parameters have been chosen, regression techniques are used to obtain models that approximate the performance of the analog integrated circuit. The constructed models have enough information to find the circuit fault when it is tested.