A New, Fast-Converging Mean-Square Algorithm for Adaptive Equalizers With Partial-Response Signaling

A New, Fast-Converging Mean-Square Algorithm for Adaptive Equalizers With Partial-Response Signaling By K. H. MUELLER (Manuscript received December 5, 1972) A new, generalized mean-square algorithm is presented to adjust the taps of a transversal adaptive equalizer. Any knowledge of the channel or the signaling format can be taken into account and will speed up the convergence process. The main applications are seen in partial-response signaling, where the new algorithm eliminates the interaction between the individual tap increments. This is achieved by decorrelating the components of the gradient in a fixed weighting matrix prior to the adjustments. Convergence is then extremely fast. If the channel has only phase distortion, a single iteration is sufficient to obtain the optimum tap vector (for any timing and carrier phase). This is verified by computer simulations. Finally, the new equalizer is compared with another structure recently proposed, and some advantages of the new system with regard to implementation and flexibility are pointed out. I. INTRODUCTION Together with sophisticated modulation techniques and bandwidthconserving signal designs, automatic transversal equalizers are the prime characteristics of the new generation of high-speed data modems that have appeared on the market in the past few years. These modems allow data rates in the range of 3600 to 9600 b/s to be transmitted over voice-band channels. In addition to point-to-point transmission between computers and data processing centers, a large demand for high-speed data sets exists in multiparty polling systems.