Adaptive Multipath Cancellation for Digital Data Communications

In this paper, an adaptive spectral equalization technique is developed for multipath cancellation, and more generally for any minimum-phase distortion, which is implemented by a constrained form of adaptive linear prediction. The technique does not use decisiondirected feedback, has guaranteed convergence and stability, and is simple to realize using conventional circuit technology. Binary amplitude-shift-keying (ASK) is considered as an example of the technique, and the theoretical performance characteristics are derived for a bandlimited pulse shape. Experimental results are also presented for an operational real-time processing system which has been developed for this application and are shown to conform favorably to theory. The analytic and experimental results presented in this paper clearly indicate the usefulness of this concept for a single multipath channel, yielding a performance advantage of between 5 and 20 dB. Particularly significant is the potential of this adaptive technique to acquire and track the multipath without any a priori knowledge of the specific parameters.