Coefficient optimization of recursive digital filters

Abstract

To realize the digital filter in digital signal processing systems, all the coefficients should be expressed with limited word length. In this thesis, a general computer program to optimize the coefficients of a recursive digital filter with a given finite word length, to meet magnitude specifications in the frequency domain, is developed. Starting with the rounded coefficients of the continuous optimum filter which was designed in the continuous parameter space, optimum coefficients with given finite word length are searched in the discrete parameter space by using the "branch and bound" algorithm. A general description of the branch and bound algorithm and some design examples are given.