Blind equalization of linear and nonlinear channels

Inter-symbol interference (ISI) in a digital communications system can be a severe impediment to recovery of transmitted symbols at the receiving end of the channel. Reduction or elimination of ISI—equalization of the channel—is accomplished using a variety of techniques, most of which assume linear ISI. The equalization is blind in the absence of known training sequences sent by the transmitter to facilitate the setting of equalizer parameters at the receiver. Nonlinear channels compound the problem by disrupting the operation of blind equalizers designed for linear channels. The advantages and drawbacks of existing blind equalization algorithms applied to nonlinear channels are discussed in this dissertation. This leads to a novel structure which replaces a simple linear filter with a relatively simple serial combination of a linear filter and a nonlinear filter. Asymmetry is introduced as a design feature of symbol constellations. Asymmetric constellations can be used for blind estimation of nonlinear phase and amplitude distortions in nonlinear channels. For a linear channel, third-order statistics of asymmetric constellations allow blind identification of the channel taps, which in turn enables use of maximum-likelihood sequence estimation or zero-forcing, linear MMSE, or decision-feedback equalizers.