Test limitations of parametric faults in analog circuits

This paper investigates the detectability of parameter faults in linear, time-invariant, analog circuits and sheds new light on a number of very important test attributes. We show that there are inherent limitations with regard to analog faults detectability. It is shown that many parameter faults are undetectable irrespective of which test methodology is being used to catch them. It is also shown that, in many cases, the detectable minimum-size parameter fault is considerably larger than the normal parameter drift. Sometimes the minimum-size detectable fault is two to five times the parameter drift. We show that one of the fault-masking conditions in analog circuits, commonly believed to be true, is, in fact, untrue. We illustrate this with a simple counter example. We also show that, in analog circuits, it is possible for a fault-free parameter to mask an otherwise detectable parametric fault. We define the small-size parameter fault coverage, and describe ways to calculate or estimate it. This figure of merit is especially suitable in characterizing the test efficiency in the presence of small-size parameter faults. We further show that circuit specification requirements may be translated into parameter tolerance requirements. By doing so, a test for parametric faults can, indirectly, address circuit specification compliance. The test limitations of parametric faults in analog circuits are illustrated using numerous examples.