Wavelet-based multi-carrier code division multiple access systems

Abstract

This thesis proposes a wavelet-based multicarrier CDMA system that uses wavelet basis instead of DFT basis to deal with the problem of inter-channel interference (ICI) and spectral bandwidth efficiency. The theoretical analysis shows that the proposed system has lower side lobe components and higher bandwidth efficiency. The simulation further shows that the wavelet-based multi-carrier CDMA system gives lower ICI and better BER performance than that using DFT basis under multipath fading channel. Moreover the proposed system is more robust against narrow band interference signal. This thesis exploits the wavelet correlation properties to perform timing synchronisation in multicarrier CDMA systems. This synchronisation is required to maintain the signal orthogonality. This timing synchronisation consists of maximum likelihood acquisition for coarse time synchronisation and early late sample tracking for fine time synchronisation. The performance of M L acquisition and ELS tracking under AWGN channel have been obtained analytically and through simulation and compared with the Cramer-Rao Bound. The designed acquisition circuit performs close to the Cramer-Rao Bound. The characteristic of the tracking circuit has also been discussed and shown the unambiguous locking capability. The thesis also proposes a blind frequency error estimation based on the variation of the signal phase. The performance of the frequency error estimator has been obtained and shown that the error variance is inversely proportional to Eb/No value. This thesis considers a new approach where the detection to get the original data symbol from the multicarrier CDMA system signal is performed using minimum mean-square error (MMSE) algorithm in both levels of orthogonality (subband and user axis). The joined MMSE combining (MMSEC) and MMSE multiuser detection (MMSE-MUD) shows better BER performance under Rayleigh fading channel compared with decorrelator, MMSEC alone and MMSE-MUD alone. The proposed MUD gets BER improvement at the cost of higher computation complexity but still retains the near-far resistance. The adaptive MMSE multiuser detection has shown further improvement of BER performance and lower computation complexity but requires a training sequence.