Neural network based technique for detecting catastrophic and parametric faults in analog circuits

The approach to transient functional test of analog circuits is considered. The artificial neural network is proposed for realization of the circuit under test (CUT) response analysis. The coefficients of wavelet decomposition of CUT transient output responses reflecting the dynamical behavior of analog circuit are used for neural network training sensitivity analysis is applied for selecting the test frequencies and test nodes. The experimental results for analog benchmark circuits are provided.