Joint Iterative Channel and Data Estimation in High Mobility MIMO-OFDM Systems

Kurzfassung

In recent years, wireless mobile communication systems have experienced an unprecedented growth in voice and data services. Future communication systems try to satisfy the increasing demand for these services by guaranteeing throughput rates of more than 100 Mbps outdoors and 1 Gbps indoors. Due to the large expenses on licensed spectrum, future communication systems require high spectral efficiency for broadband transmission. Orthogonal frequency division multiplexing (OFDM) in combination with bit-interleaved coded modulation (BICM) has turned out to be a robust yet implementation efficient technique for reliable broadband communication over fading channels. To achieve high spectral efficiency, BICM-OFDM systems employ higher order bit to symbol mappings and multiple-input multiple-output (MIMO) techniques to transmit several bits per time and frequency bin. In this thesis, joint iterative channel and data estimation schemes for high mobility MIMO OFDM systems are developed and analyzed. Besides OFDM, we particularly consider OFDM code division multiplexing and its multiple access variant multi-carrier code division multiple access. The convergence behavior of iterative receivers and the resulting system performance is analyzed with extrinsic information transfer (EXIT) charts and bit-error rate or frame-error rate curves. Furthermore, we have developed adaptive bit loading (ABL) schemes that exploit channel state information at the transmitter (CSIT) in combination with iterative receivers and algorithms to estimate CSIT not only for time division duplex but also for frequency division duplex systems. In the first part, we research iterative receiver algorithms for OFDM and OFDM code division multiplexing with BICM and perfect channel state information. We explain why it is necessary to iterate extrinsic instead of a-posteriori information. Then, we present a novel, analytic derivation to compute EXIT charts efficiently. We compare the EXIT charts to bit-error rate transfer charts with the result that EXIT charts predict the performance of iterative receivers more accurately. Further, we show that EXIT charts can predict the performance of MIMO systems and channels only accurately if the channels are weakly correlated and the MIMO algorithms directly influence the demapper. In the second part, we study iterative receivers with different channel estimation algorithms. Both joint a-posteriori probability iterative channel estimation and demapping and the optimum linear minimum mean square error channel estimation have high computational complexity. Hence, we propose suboptimum iterative channel estimation (ICE) algorithms with less complexity. When data symbols that have small amplitude values are used in ICE, noise enhancement can degrade the channel estimation. To avoid noise enhancement, we propose the modified least-squares error method or the modified minimum mean square error methods. To extend the ICE algorithms to MIMO systems, we either compute for each transmit antenna an interference-reduced receive signal or we exploit the orthogonality of the space-time-frequency codes. For a realistic MIMO channel model, simulated throughput results demonstrate that iterative receivers with ICE can double the throughput at medium signal-to-noise ratios. In the third part, we investigate ABL algorithms. The ABL approach differs from error rate optimizing ABL schemes in the literature as we directly aim at minimizing the overall bit-error probability at the decoder output under the assumption of perfect CSIT. First, we develop a simple ABL scheme, whose complexity to search for the optimum bit loading scales linearly with the number of data symbols. Second, we extend the algorithm to higher order alphabets and an arbitrary bit-rate constraint. We combine the ABL scheme with transmit diversity techniques and iterative receivers so that ABL can yield significant performance gains versus uniform bit loading. In addition to perfect CSIT, we also develop a simple scheme to estimate the CSIT based on received pilot symbols that also takes into account the transmit diversity techniques. Thus, we consider the major factors contributing to unreliable CSIT for ABL, which are channel estimation errors, limited capacity in the feedback, and feedback delays.