Parametric Fault Detection in Analog Circuits: A Statistical Approach

A statistical approach is proposed to detect parametric fault in linear and weakly non-linear analog circuits by mapping faults to a statistical metric, Bhattacharyya distance which is measured from the Probability Density Function (PDF) of the outputs. The non-Gaussian Auto-regressive (AR) model is used to estimate the PDF. To validate the proposed statistical approach, we have simulated two benchmark circuits, Sallenkey band pass filter and cascade amplifier with CADENCE Virtuoso using umc-180nm technology. Defect screening is also measured with linear regression analysis. Detectability of the proposed method for parametric fault is reasonably large compared to the functional test method.