Synthesis of efficient low-order FIR filters from primitive sections

Abstract

A synthesis procedure for low-order FIR filters based on the cascading of computationally simple (and in some cases very computationally efficient) primitive sections is described. An integer programming approach effectively yields an optimal factorisation of a low-order transfer function H(z) that satisfies the filter specification and, hence, yields also an optimal cascade of available primitives. If the primitive cascade approach is used, synthesised filters are considerably more computationallly efficient than optimal direct form realisations. The design process is noniterative, completely automatic and particularly suitable for high sample rate applications.