Computationally Efficient Equalisation of Broadband Multiple-Input Multiple-Output Systems

Multiple-input multiple-output (MIMO) systems are encountered for example in communications if several transmit and receive antennas are empoyed, such that a separate transmit channel exists between every possible pairing of transmitter and receiver antennas. As a results if this spatial diversity, the channel capacity is dramatically increased over the single-inout single-output (SISO) case. While this increase is desired, the use of high data rates requires sophistiocated equalisation and/or detection schemes in the receiver to compensate for spatial and temporal dispersion in broadband MIMO channels, since a time-dispersive, in addition ot spatially-dispersice channel, must be assumed. The estimation of the broadband MIMO channel or its inverse is in general difficult and calls for training sequences that reduce the slot time for the transmission of actual data, which may counteract the promised gain in channel capacity. Another problem can be the computational cost since in many systems this will be at least proportionaly to the added spatial diversity. This thesis is concerned with the application of techniques that find the best broadband MIMO equaliser in terms of MSFR or BER performance while keeping the computational cost as realistically low as possible. It examines established adaptive and analytic methods of doing this and then moves on to the application of suibband adaptive filtering techniques to perform MIMO channel equalisation and detection, since this tehcnique has been found to give considerable advantages wit respect to computational complexity and converence rate fir related SISO applications. For many slow-converging low-cost adaptive algoirhtms applied to the inversion of channels, the convergence rate can be increased by use of subband processing, where, in independent frequency bands, separate smaller-scale adaptive algorithms are operated at a reduced update rate. We will apply such methods to the identification and inversion of MIMO channels. Fractionally spaced systems also are known to outperform theuir symbol-spaced counterparts hence these are factored into the subband MIMO system developed. Many simulation results demonstrating the benefits of MIMO systems with respect to the channel capacity, the performance of various adaptive and analytic MIMO inversion techniques and the potential complexity and convergence rate improvements of the subband approach in the MIMO context are presented. Adaptation of MIMO systems generally takes much longer than for SISO systems. For adaptive identification the time increases by an amount approximately equal to the dimensions of the MIMO system. A frequency-domain inversion method shows the best performance compromise between MSE and BER performance and the requirement to minimise computational complexity, though it suffers from in-accuracies at high SNR values. The subband approach shows benefits for highly time-dispersive channels, and its potential improvements in tracking ability for dynamic channels means is can be beneficial in time-varying fading environments. Finally , fractionally-spaced equalisation often shows considerable benefits and is even capable of equalising channels that are not possible to equalise using standard symbol-spaced methods.