A New Approach to Effective Digital Filter Design

Abstract

In this paper the effect of modifying digital filter weights derived on a least squares error basis through multiplication by certain weighting functions is investigated. It is shown that this approach amounts to adopting frequency resolution as a filter performance criterion, in a manner analogous to resolution in spectral estimates along the lines of Blackman and Tukey. Weight formulation using the method is very simple, and working transfer functions are easily estimated without computation. Simple, effective designs are presented for low, high and band-pass (or band-stop) filters of the "cosine" or in-phase type, for cosine "low low-pass" and "sampling" filters, and for "sine" or quadrature "sampling" filters, as well as differentiators.