Fault Modeling of Analog Circuits Using System Identification Automated Model Generation Approaches from SPICE level Descriptions

Fault modeling and simulation (FMAS) of analog circuits is considered to be time consuming and expensive as compared to digital circuits. FMAS of analog circuits is heavily dependent on transistor level (TL) circuits and slow speed of TL fault simulation (TLFS) increase overall testing cost. Therefore, Automated Model Generation (AMG) techniques are employed to model nonlinear faults in analog circuits and achieve speed up in simulation. In this paper, we model faults of an operational amplifier (opamp) circuit using System Identification (SI) based AMG techniques: nonlinear autoregressive with exogenous input (NLARX) and hammerstein-wiener (H-W) techniques from SPICE transistor level descriptions. To investigate performance of AMGs for nonlinear behavior of faults, several nonlinear functions are employed such as sigmoid network, wavelet network, and tree partition etc. A comparison of simulation speeds of TLFS and AMGs is also provided. Simulation results show that more accurate and efficient AMGs should be considered for the modeling of nonlinear behavior of analog faulty circuits and achieve speedup in simulations.