| In this thesis we develop new practical transmission and detection techniques for wireless multiple-input multiple-output (MIMO) communication systems operating over frequency-flat block Rayleigh fading channels. Both the coherent scenario where the fading coefficients are known at the receiver but not at the transmitter and the non-coherent scenario where these coefficients are not known at either the transmitter or the receiver are considered.; For coherent systems, we develop a design method for a class of codes known as Linear Dispersion (LD) codes. This class of codes subsumes several standard designs and can be used with systems that have an arbitrary number of transmit and receive antennas. We show that for systems that employ a large number of transmit antennas, LD codes constructed from random unitary coding matrices become asymptotically optimum from different design perspectives, viz., Minimum Mean Square Error (MMSE), mutual information and average Pairwise Error Probability (PEP). Those measures have a direct impact on the detection complexity, data rate and error performance that a space-time code can achieve. Using the insight generated by the asymptotic result, we provide a structured design technique for these codes that suits a broad class of finite configurations. Although our main result is based on large systems, we demonstrate via simulations that even for small systems, our codes can support higher data rates and provide performance advantages over known designs.; Based on the asymptotic results, we also propose a row interleaving scheme. This interleaving scheme is shown to result in significant performance enhancement without incurring additional detection complexity.; For the non-coherent communication scenario we focus on wireless systems operating at moderate-to-high SNRs. At high SNRs, the capacity achieving distribution for this channel corresponds to the isotropic distribution on a Grassmann manifold. Using a subspace perturbation analysis, an appropriate metric for the distance between Grassmannian constellation points is determined. Based on this metric, a greedy technique for designing constellations that mimic the isotropic distribution is then proposed. Our numerical examples show that our choice of the metric yields constellations that outperform other designs as the data rate approaches the high SNR non-coherent capacity limit. In addition to the proposed constellation design technique, we use the subspace perturbation analysis to develop an efficient suboptimum detector. The performance of this detector is comparable to that of the maximum likelihood detector, but it requires considerably less computational effort. Finally, seeing as the pairwise error probability (PEP) is indicative of the system performance at high SNR, we derive an exact expression for this quantity. As compared to other PEP expressions, our expression is easier to compute and can be used to extract more insight into the key parameters that affect the high SNR performance of the non-coherent system. |
