Coding and advanced signal processing for MIMO systems

The use of multiple transmit and multiple receive antennas in mobile communications offers a high potential to improve the bit rate and the link quality. This can be achieved by using a higher multiplexing rate and by exploiting the diversity contained in the channel, under the constraint of acceptable complexity. The channel knowledge availability has an important impact on the system design. In fact, the Channel State Information (CSI) at the transmitter (Tx) has an impact on the coding, whereas the quality of the channel knowledge at the receiver (Rx) side has an impact mainly on the detection and the channel estimation. The first part of this thesis considers the case of absence of CSI at the Tx and perfect at the Rx. We propose the Space-Time Spreading (STS), which is a space-time coding scheme based on linear precoding that use a MIMO convolutive prefilter. STS achieves full multiplexing rate and is optimized to exploit maximum diversity and coding gains and to save the ergodic capacity. STS allows to use various receiver structures of low complexity. The Stripping MIMO Decision Feedback Equalizer (DFE), is a non-iterative receiver that detects the streams successively. The performances of the Stripping are evaluated in term of diversity versus multiplexing tradeoff. Another non-iterative receiver is the Conventional DFE applied to the MIMO case. It detects jointly symbols for different streams but proceed successively in time. The third proposed receiver is an iterative one. It takes advantage of the presence of the binary channel code, and iterates between the linear equalizer and the binary channel decoder. Simulations are provided to evaluate its performance. In the second part we consider channels with partial CSI at the Tx and perfect CSI at RX. The partial knowledge in these cases can come from the decomposition of the channel in slow varying and fast varying parameters. It can also be the result of the reciprocity of the downlink and uplink physical channels. For those cases we provide suitable channel models and study the ergodic capacity. In the last part, the case of absence of CSI at both Tx and Rx is considered. The capacities of two channel models, block fading and time selective, are studied. Due to the absence of CSI at Rx in this case the channel needs to be estimated in practical systems. We propose semi-blind estimators that combine training and blind information. Identifiability conditions are derived and simulations are presented to evaluate performances.