The application of nonlinear architectures to adaptive channel equalisation

The authors investigate the potential value of applying nonlinear architectures to the adaptive equalization of a bipolar signal passed through a dispersive channel in the presence of additive noise. The problem is described as a geometrical classification problem and the factors affecting the complexity of the problem are considered. In particular, it is shown that some degree of nonlinearity is desirable in the equalizer structure when signal-to-noise ratios are poor. The authors describe two nonlinear architectures which might be considered as adaptive equalizers, and evaluate their performance in this role by means of computer simulation. The advantage of the structures over their linear counterparts is readily discerned. It is concluded that by incorporating a degree of nonlinearity in the design of an equalizer, it is possible to produce a structure which can achieve a BER (bit-error rate) superior to that offered by linear structures when equalizing bipolar signals passed through linear channels, especially when signal-to-noise conditions are poor