A Simple Adaptive Equalizer for Efficient Data Transmission

A small and relatively inexpensive automatic equalizer has been developed for use with a vestigial sideband AM data set presently utilized to provide a high-speed Bell System voiceband data service. The equalizer uses a transversal filter structure comprising a tapped analog delay line with a variable-gain circuit associated with each tap. Its operation is adaptive in the sense that the tap-gain-adjustment information is derived from the received data, so that the tap gains are continuously optimized. A number of adaptive tap-gain-adjustment algorithms and transversal filter realizations were considered. The algorithms range in order of implementation complexity from "zero forcing," involving binary correlation, to "mean square," which requires linear correlation. The transversal filter realizations vary from an early scheme, which accomplished the variable-gain function with up-down counters controlling a switchable resistive ladder network, to an all-digital approach, where the analog delay line is replaced by shift registers and the variable-gain function is achieved with binary multipliers. Although the all-digital approach, using the mean-square algorithm and implemented with large-scale integration, will undoubtedly be the best choice in the near future, the equalizer to be described was selected on the basis of its being the cheapest to implement at the present time. The equalizer uses a modified zero-forcing tap-gain-adjustment algorithm, which is no more costly to implement than the zeroforcing algorithm, but which has superior convergence properties in the presence of severe distortion. The variable-gain function is achieved by means of a field-effect transistor (FET) used as a variable resistance. A single operational amplifier is used both to control the resistance of the FET and to provide feedback necessary to insure precise circuit performance without requiring close tolerances on the FET characteristics.