Black-Box Modelling of Nonlinear Audio Circuits

Abstract

This Thesis details various black-box modelling techniques for nonlinear audio circuits. The aim of this work was to compare the different approaches and determine the current most accurate model. The models analysed were: Chebyshev polynomials, Hammerstein, polynomials of best fit and a parametric model. These implementations were compared based on their spectral accuracy and time-domain accuracy.

From the analysis of these models, a new method is proposed. This proposed model restructures the determined most accurate black-box model by placing a nonlinearity between a high-pass and a low-pass filter. The filters emulate the linear response of the nonlinear circuit. These filters also have a reduced filter order in comparison to the linear response. The proposed method results in a model that performs close to the current best but does result in a massive reduction in computational time, up to 95% less time required and 99% fewer operations.