Digital filter design optimization using partial cost functions

Abstract

A novel optimization procedure for the design of IIR digital filters using partial cost functions is introduced. The method is specially useful in the design of transfer functions with different numbers of zeros and poles, which find accurate hardware realizations of digital and analog sampled-data IIR filters. In addition, as the number of poles is reduced so does phase distortion, which suits modern communication systems. The use of partial cost functions avoids the occurrence of local minima, increasing the robustness of the optimization algorithm. Analytical step size adjustment, based on Taylor Series, is developed for accurate update of the transfer function parameters at each iteration, increasing the convergence rate. Simulation results are presented to verify the efficiency of the proposed approach.