Design of lattice-based digital IIR filters with discrete coefficients

In this paper, the problem of designing discrete coefficient infinite impulse response (IIR) digital filters is discussed. The digital filters are characterized with a lattice-based filter structure that has three sets of parameters. The filters, obtained in our proposed design procedure, are determined with all the sets of parameters optimized over a discrete power-of-two coefficient space such that the signal power ratio, an FWL-related measure, is minimized subject to the frequency response specification satisfied. A genetic algorithm (GA) is derived to solve the corresponding optimization. Two design examples are given, which show that the filters designed by the proposed method not only have a low implementation complexity but also possess much improved finite word length properties such as very low parameter sensitivity and very uniform signal powers across signal nodes.