A unimodal adaptation criterion for a non-linear IIR self-learning equalizer

In this work, we propose a nonlinear structure for self-learning equalization, which can be easily updated using the direct-decision error criterion. Such a structure consists in three different systems: an IIR predictor that provides the magnitude equalization, an automatic gain control and a nonlinear phase equalizer. The main contribution of the paper concerns the demonstration of the unimodality of the adaptation criterion. An original theoretical analysis is carried out for the nonlinear phase equalizer. Finally an example with a severe radio-communication channel is presented.