Research On Theory And SofT-Demodulation TeChniques Of Bit-Interleaved Coded Modulation Systems

According to Shannon information theory, Massey proved that the system performance can be improved greatly by optimizing channel coding and modulation togather in1974. Then Ungerboeck proposed the famous Trellis coded modulation (TCM) technique in1982. TCM is considering as a milestone in the field of communications after Shannon information theory. However, the performance of TCM is not robust in fading channel. Bit-interleaved coded modulation (BICM) technique inserts the bit interleaver between coder and modulator to improve the degree of diversity in fading channel which can improve the performance of BICM system in fading channel and enhance the flexibility of system. In the recent years, bit-interleaved coded modulation (BICM) has become the de facto scheme of modern communication systems due to its flexibility, high spectrum efficiency and robust performance over fading channels, such as DVB-T, DAB, DRM, HSDPA,802.11,802.16and LTE etc..In order to reduce the performance loss for cascading encoder and modulator, the soft bit metrics are computed by the demodulator and provided to the the soft-decision decoder. The optimal bit metrics for BICM systems are maximum likelihood (ML) bit metrics, but the calculation for ML bit metrics are too complicated to implication in high-order modulation. So Max-Log is the most common bit metric generation method in practice. To meet the need of adaptive modulation in moderm communication systems, the dissertation researchs the methods of bit metric calculation for some modulation types and proposed some general bit metric generation approaches with low complexity and similar even better performance than Max-Log.Firstly, the conventional paralell sub-channel model and the new dismatched decoding model are described for BICM systems. Then the generalized mutual information (GMI) is introduced. To make the GMI as a performance measurement for different bit metrics, this dissertation derived some properties of GMI. The convex property simplifies the complexity to computer GMI. The relationship of the bit metric and GMI is further elaborated.This dissertation proposes a low complexity bit metirc calculation method for PSK signaling with Gray labeling. This approach converts all bit metrics to the bit metric of most significant bit (MSB) by recursion. The complexity of the method is very low compared to Max-Log bit metric. Because of the recursive formula is unrelated to modulation order, the proposed method can be used for any modulated signals of PSK and is very suitable for adaptive modulation systems. The GMI and bit error rate (BER) are used to measure the performance of this simplifed bit metrics. The simulation results demonstrate this method has similar performance compared to ML bit metirc over AWGN channel and fading channels.This dissertation extends the above method to ASK and QAM signals which is more common in practice for high-order modulation. This method is based on the bit metric calculation function of MSB by recursion. For the importance of the metric function of the MSB, this dissertation derived non-linear and peicewise linear approximate functions and analyse their performances in BICM systems. The the mixed apprximate function is derived which has better performance the Max-Log. When the number of bits per symbol is odd, the Cross-QAM and pseudo-Gray is the best constellation and labeling type. In BICM systems,32-QAM and128-QAM are the most common modulation types. In this dissertation, a simplified bit generation method is proposed base on region dividing. The complex plane is divided to some regions by Max-Log bit metrics firstly. Then we merged these regions by its bit metric generation function which can make regular region boundary.