Abstract

The data throughput or bandwidth efficiency of high-speed digital transmission over band-limited channels can be increased by employing a blind equalizer. The traditional blind equalization schemes, employing only adaptive finite impulse response (FIR) filter-based structures, perform poorly for maximum-phase communication channels having deep spectral nulls. This problem is due to the difficulty in modeling the inverse of the maximum-phase communication channel with a finite-length adaptive FIR filter. Attempts have been made to solve this problem by including both infinite impulse response (IIR) and FIR filters in the equalizer structure. However, these techniques are computationally demanding and are very complex for a hardware implementation. In this thesis a new sufficient-order blind equalization scheme is proposed that attempts to address some of the problems of the existing techniques. The proposed sufficient-order blind equalization scheme is based on a structure that makes use of an adaptive IIR predictor, an FIR filter, an adaptive FIR filter, and two block-based time reversers. New algorithms for the adaptation of IIR predictor and FIR filter are proposed. The adaptation algorithm for the IIR predictor is derived based on the well known least mean square error criterion using a gradient method, and for the adaptation of the FIR filter, a constant modulus adaptive algorithm is proposed. It is shown that the proposed blind equalization scheme requires a substantially less amount of computations and is easy to implement. The proposed sufficient-order blind equalization scheme is applied to various linear and deep-null minimum-phase, maximum-phase, and mixed-phase communication channels. From the experimental results, it is demonstrated that the proposed scheme estimates the unknown channels correctly and provides a better performance as compared to the existing schemes, in terms of the channel estimation, symbol error rate, and spectral response.