Automatic equalization using a successive overrelaxation iterative technique

Abstract
The successive overrelaxation iterative technique is used as the algorithm for the iterative adjustment of transversal equalizer tap gains during a training period using isolated pulses for the minimization of the mean-square error. For certain channel models, substantial improvement is demonstrated in the rate of convergence over the commonly used gradient techniques. Numerical simulations indicate that the relaxation algorithm consistently converged much faster than the gradient techniques. The behavior of the algorithm in noise is also investigated and is shown to be no worse than any other equalizer algorithm.