Data-Driven Modeling of Weakly Nonlinear Circuits via Generalized Transfer Function Approximation

This paper presents an extension of the Vector Fitting algorithm with the purpose of constructing compact behavioral models of weakly nonlinear circuits starting from frequency-domain input-output data. Using the concept of generalized transfer function provided by Volterra series theory for nonlinear systems, the algorithm approximates a given dataset of generalized transfer function samples with a black-box multivariate rational model. The fitted model can then be recast into a bilinear state-space form for time-domain analysis. Practical extraction of the required data samples can be carried out by measurement or harmonic balance analysis available in commercial solvers. Examples demonstrating the accuracy and efficiency of the behavioral models include a Low-Dropout Regulator and a Low Noise Amplifier.