Optimization of multidimensional equalizers based on MMSE criteria for multiuser detection

Abstract

This thesis is about designing a multidimensional equalizer for uplink interleaved division multiple access (IDMA) transmission. Multidimensional equalizer can be classified into centralized and decentralized multidimensional equalizer. Centralized multidimensional equalizer (MDE) have been used to remove both inter-symbol interference (ISI) and multiaccess interference (MAI) effects from the received signal. In order to suppress MAI effects, code division multiple access (CDMA) has been used with MDE to minimize the correlation between users' signals. The MDE structure can be designed using linear equalizer (MLE) or decision feedback equalizer (MDFE). Previous studies on MDE employed adaptive algorithms to estimate filter co-effi cients during the training mode, i.e. the symbol equalization was not optimal, for two users. In our work, we applied MDE on IDMA receiver for multipath selective fading channels and also derived new equations to obtain the optimal filter taps for both types of MDE equalizers, i.e. MDFE and MLE, based on the minimum mean square error (MMSE) criterion. The optimal filter taps are calculated for more than two users. Moreover, we investigated the performance of the optimal MDFE using both IDMA (MDFE-IDMA) and CDMA (MDFE-CDMA) detectors. Generally, the MDE equalizer suffers from residual MAI interference effects at low signal-to-noise-ratios (SNR) due to the delay inherent in the convergence of the crossover filter taps. Therefore, a new decentralized multidimensional equalizer has been proposed to IDMA detector. Within design of decentralized equalizer, the convergence problem has been resolved by replacing the crossover filters with parallel interference canceler (PIC) for removing MAI dispersion. The proposed decentralized multidimensional equalizer shows a higher efficiency in removing MAI interference when compared with existing receivers in the literature. However, this is achieved at the expense of higher computational complexity compared to centralized multidimensional equalization.