A novel digital IIR filter design strategy - Structure-based discrete coefficient filters

Most of the existing design methods for digital filters that are to be implemented with finite precision device are given by two independent stages. The first stage is to determine a transfer function (in infinite precision) that meets the specifications given, while the second one is to choose a proper filter structure to reduce finite wordlength (FWL) errors as much as possible. In this paper, we propose a new digital filter design strategy that merges the two stages such that both the performance and the structure issue are considered simultaneously with the FWL effects taken into account. The idea is developed with a recently proposed lattice-based filter structure [8]. A design example is given, which shows that the obtained structure-based filter not only outperforms existing lattice-based filters in terms of reducing the FWL effects but also has a much lower implementation complexity than the latter and the canonical filter realizations.