| In this thesis we present a parallel N x N MIMO detection algorithm that is suitable for implementation on multi-core low-power processors. This algorithm has multiple versions which, depending on the SNR and target FER, allows for the complexity, performance and power consumption to be traded off on a packet-by-packet basis.;Despite the use of the capacity-lossy Alamouti code we show that the outage capacity and FER performance of the proposed algorithm at practical SNR levels below 20dB are comparable to, or sometimes outperform, those offered by spatial multiplexing algorithms such as MMSE-VBLAST. Furthermore, we show that our algorithm is more robust than MMSE-VBLAST in the presence of channel imperfections such as spatial correlation and channel estimation errors, two commonly encountered imperfections in practice. We also show that the complexity of our algorithm is equivalent to that of a lower complexity version of MMSE-VBLAST. Through FER comparisons with other well-known hybrid algorithms we also show that our algorithm achieves similar or better performance at a much lower complexity.;The features offered by the proposed algorithm that is: parallel architecture, good FER/capacity performance, robustness against channel imperfections make it an ideal candidate for use in 4G and beyond-4G wireless standards.;The proposed hybrid algorithm essentially consists of coding the transmit symbols in pairs using the Alamouti space-time block code. At the receiver, one or multiple QR decompositions are performed in parallel after which alternate successive interference cancellation, Alamouti decoding and coherent combining are performed. Iterative detection can also be performed should better error performance be required. |
