Turbo receivers for equalizing frequency-selective MIMO channels: algorithms and implementations

Received signals of modern wireless communications systems are impaired by temporal and spatial interferences between the symbols. Consequently, a sophisticated equalizer is required as an essential component of the receiver. When a priori knowledge based on decoder results is taken into account in the equalization procedure, interference can be reduced more efficiently than in conventional low-complex receivers. These schemes are referred to as turbo equalization. This work focuses on turbo receiver schemes to equalize frequency-selective MIMO channels, which introduces intersymbol interference (ISI) and inter-antenna interference (IAI). The thesis is divided into two parts: The first part (Chapter 4) investigates different MMSE-based algorithms in terms of communications performance and computational complexity. In this context, a novel multi-iterative turbo receiver is proposed to enable interference-aware receiving with respect to the given channel scenario. It is furthermore shown, that normalizing the filter coefficients reduces the equalization bias and improves significantly the reliability of the transmission. The second part (Chapter 5) demonstrates an implementation of the developed algorithm based on a programmable solution. The equalizer application has been profiled whereas matrix inversion was identified as the most computational intensive operation. Starting from a RISC processor instruction set architecture, new equalizer-specific functional units (FUs) have been developed and integrated in order to accelerate both non-recursive and recursive operations. The design was optimized towards high area-efficiency, thus increasing the performance throughput with reasonable low additional hardware effort. In contrast to ASIC implementations, the resulting ASIP core enables very high flexibility to support different equalizer modes along with moderate to high performance.