Nonlinear self-training adaptive equalization for multilevel partial-response class-IV systems

Abstract

Summary form only given, as follows. Self-training adaptive equalization for multilevel partial-response class-IV systems is addressed. The author considers an adaptive equalizer realized with distributed-arithmetic architecture, where the process of multiplying the tap signals with tap gains and summing the resulting product is replaced by a procedure involving only table lookup values and shift-and-add operations. Self-training adaptation schemes devised for linear adaptive equalizers do not converge if applied to a distributed-arithmetic equalizer because of the inherent nonlinearity of the system during the adaptation process. A new algorithm allowing the convergence of the lookup values is analyzed. Under the usual assumption of independent signal vectors in the equalizer delay line at different updating times, a sufficient condition allowing the temporal evolution of the lookup values to be modeled as an ergodic Markov process is given. Numerical results are presented with reference to a multilevel PRIV system for high-rate data transmission over twisted-pair cables.