Simplified detection and power-efficient signaling for a coded DSTTD-OFDM system

Abstract

Many people can currently enjoy high-speed multimedia stream services with very high quality-of-service (QoS) in wired communication systems. In wireless communications technologies with single-transmit and single-receive antenna (STSR), most of today``s systems are difficult to provide such high data rate and good QoS. However, information theories have recently shown that space-time architecture systems with multiple transmit/receive antenna techniques can either achieve enormous data rate or improve channel reliability over STSR systems in rich scattering environments. A hybrid space-time architecture, called double space-time transmit diversity (DSTTD) system, was proposed as a simple way to obtain diversity gain and high data rate simultaneously. It consists of two Alamouti``s space-time transmit diversity (STTD) units. Alamouti``s STTD itself gives transmit diversity gain and its dual structure can support double transmission data rate. It also demands smaller number of receive antennas than transmit antennas, so that it is favorable in practical applications. By employing orthogonal frequency division multiplexing (OFDM) modulation, so called the DSTTD-OFDM system can overcome intersymbol interference (ISI) due to frequency-selective fading channels. As combined with bit-interleaved coded modulation, the coded DSTTD-OFDM system can additionally order coding gain. The main bottleneck of implementing the coded DSTTD-OFDM system is the complexity of signal detection at the receiver. The maximum likelihood detector can produce optimal log-likelihood ratio (LLR), while its complexity grows exponentially with the number of transmit antennas. It is often prohibitive when high order modulation is used. Several suboptimal LLR computation methods were previously introduced, on the other hand, they still require very high computational loads and their performances are not comparable to that of the ML detector. In this thesis, we suggest a simplified ML (S-ML)...