Soft fault detection in analog circuits from probability density function

Abstract

A new parametric fault detection scheme for linear and weakly non-linear analog circuits is proposed from probability density function (PDF) of the output. Non-parametric kernel density estimation (KDE) technique is used to estimate the PDF from the random output of the circuit excited with random input stimuli. Two benchmark circuits viz. Continuous-time low pass State Variable Filter circuit and Cascade Amplifier are tested to validate the proposed framework. All the circuits are simulated with CADENCE Virtuoso using UMC-180nm technology. Detectability of the proposed method of soft fault detection is appreciably higher than that of functional test method.