Iterative Receivers with Channel Estimation for Multi-Carrier Systems

Kurzfassung

The dilemma of mobile communications is multipath fading. There are several techniques developed in the recent past to combat its effect such as orthogonal frequency division multiplexing (OFDM). This scheme gives increased resilience against frequency and time selectivity caused by the multipath environment at the cost of a small reduction in the throughput. To fully exploit the potential of OFDM, coherent reception is used which requires channel knowledge in order to operate. However, channel estimation in case of multicarrier transmission is a demanding task. Usually pilots are inserted along time and frequency in the transmitted data, which cause a further reduction in throughput. Iterative receivers with channel estimation provide the possibility of coherent transmission with virtually no pilot insertions, while giving better estimates of the channel compared to conventional one-shot receivers. In this thesis, a coherent, bit interleaved coded modulation (BICM) transmission using OFDM over time variant multipath channels is investigated. Extrinsic information transfer (EXIT) charts are used to investigate the convergence behavior of iterative receiver along with bit error rate (BER) and mean square error (MSE) plots. In the literature, analysis is commonly done for an inner component, the channel estimator (CE) and demodulator, and an outer component, the decoder. Here, we additionally try to separate the analysis of the CE from the demodulator. The results show up to 5dB of gain in SNR at BER of 10−4 with iterative receivers, compared to one shot receivers in case of convolutional codes. With iterative CE, after few iterations the system performs as good as it does with perfect CE. In case of parallel concatenated codes (PCC) we also get 5dB gain compared to the one shot receiver with the same PCC. However, an iterative receiver using PCC gives a reduction up to 103 in the BER floor compared to the iterative receiver with a convolutional code. Excluding CE and demodulation from the iterative loop of the receiver, each causes more than 2dB of loss. A-posteriori information feedback to the channel estimator gives the gain in SNR of 2dB compared to extrinsic feedback in the turbo cliff region. We further observed that using hard decision feedback improves MSE up to about one decade in turbo cliff region compared to soft decision feedback. Finally a conversion mapping from mutual information to BER for EXIT chart analysis is given. A CE EXIT chart model is proposed, results indicate good match in medium to high a-priori mutual information. Hence a 3-D EXIT chart investigation is done with the output mutual information of the CE, demodulation and decoding each along one axis.