Multichannel equalization using linear matrix inequality optimization
Abstract
In digital communication receiver systems equalization is either a linear or nonlinear process. The primary task of the equalizer is to remove or at least reduce intersymbol interference (ISI) between transmitted symbols. ISI arises from time dispersion whereby a transmitted symbol is smeared across several symbol time intervals, hence interfering with other transmitted symbols. Time dispersion stems from the characteristics of the channel medium. Some examples of time dispersion would be the band-limited nature of the channel medium and multi-path interference. Not compensating for ISI typically results in an unacceptable probability of error at the receiver. This research focuses on designing multi-channel equalizing filters for digital communication systems. The design methodologies developed in this research relies on numerical convex optimization based on Linear Matrix Inequalities (LMIs). Digital communication systems are modeled as having an additive noise source to account noise distortions. ISI and the additive noise source are primary performance impairments in most digital communication systems. Thus it is prudent to account for both of these factors in designing an equalizing filter. Under certain conditions it is well known that ISI reduction and output noise power are competing entities. The design techniques developed in this research exploit the trade-off between ISI and equalizer output noise power, by utilizing an equalizer design methodology based on LMI optimization. Additionally, concepts such as spatial diversity, i.e., multiple channels, and sample-spaced equalization (to be defined later), are used to solve problems associated with multi-path channels which exhibit severe fading. Performance gains are shown in terms the post-equalizer constellation mean-square error (MSE) and the bit-error-rate (BER) of the receiver.
Degree
Ph.D.
Advisors
Balakrishnan, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.