Computational and performance aspects of minimax equalizers

The subject of this report is the computer-aided design of an optimum filter for time-domain equalization of a linear, time-invariant system. Given advance specification, in the time domain, of the unequalized system response and the desired equalized response, the equalizer impulse response is found by deconvolving subject to the minimax or Chebyshev error criterion. Deconvolution is accomplished numerically using the second algorithm of Remez. The unique advantages of this design procedure are uniform control and minimization of time-domain equalization errors. Computed results for about 50 FIR equalizer designs are summarized in graphical forms which relate accuracy, convergence, and computation time to filter length. Also illustrated are sensitivities of equalization error to solution-interval width, inaccurate input data, and limited computational precision. Some related IIR equalizer designs are considered.