A New Class of Automatic Equalizers

A new class of automatic equalizers with very fast convergence is discussed in this paper. This fast convergence is accomplished by means of an iterative procedure which successively makes higher-order approximations on the desired equalizer function. The iterative procedure can be conveniently realized in the form of transversal filter stages in cascade. For a given distortion D < 1, the residual distortion at the equalizer output after n iterations can be reduced to no more than the equation in the noise-free cases. When this approach is applied to the feed-forward part of a recursive structure, the front distortion (due to precursors) can be reduced in a similar fashion without unduly increasing the overall distortion. The resulting distortion, mainly in the rear end, can then be cancelled out via feedback paths. Other topics treated include certain generalizations, the truncation error due to limited length in cascaded equalizer sections, and the effect of noise. Several numerical examples are also presented to illustrate the effectiveness of the approach.