A Nyquist-rate adaptation algorithm for fractionally spaced equalizers.

Synchronous transversal equalizers have long played a significant role in many advanced data communication systems. In recent years, however, fractionally spaced equalizers, with coefficient weighting spaced less than a baud interval, have gained in popularity because of their notable advantages, chiefly an ability to form an optimal linear receiver and provide performance that can be quite insensitive to receiver timing phase. In this work we describe a new time-domain Nyquist-rate algorithm that broadens the utility of fractional equalizers by permitting more rapid convergence relative to synchronous coefficient updating, stochastic gradient or "true" zero-forcing adaptation, user-defined specification of the end-to-end Nyquist channel, and the potential for minimization or complete elimination of the coefficient drift phenomenon. The algorithm is described in the context of decision-aided equalization, though its use with training sequences is evident. The investigation presented herein includes an analytical description of the algorithm, a functional circuit architecture, and brief reference to a computer simulation illustrating stable equalizer operation in the presence of dispersion on a digital subscriber loop.