Training Sequence Design And Demodulation For TDS-OFDM Transmit Diversity System

The transmit diversity technique can significantly improve the transmission performance of wireless communication systems under multipath fading channels. And the application of transmit diversity in the Time Domain Synchronous Orthogonal Frequency Division Multiplexing (TDS-OFDM) system is the requirement and trend of Chinese Digital Television Terrestrial Broadcasting (DTTB). The essential issue in the design of TDS-OFDM system with transmit diversity is to design and choose proper training sequences as the guard interval paddings in TDS-OFDM signal frames, which also facilitate fast frame synchronization, accurate channel estimation and correct demodulation at receiver. Therefore, this dissertation focuses on the training sequences design for the TDS-OFDM system with transmit diversity, the demodulation algorithms at the receiver, as well as the Inter-Block Interference (IBI) cancellation in OFDM system.Firstly, several types of sequences with zero cross-correlation property and good auto-correlation characteristic are designed, including Binary Partial Block Pass (BPBP) sequence, Pseudo Noise (PN) based Partial Block Pass (PN-PBP) sequence, PN based Partial Comb Pass (PN-PCP) sequence, and PN Extended-and-Rotated (PN-ER) sequence. A near-ergodic algorithm is proposed to help search the BPBP sequences with low Peak-to-Average Power Ratio (PAPR). The quasi-constant-amplitude property for the non-zero elements in frequence domain and the near-optimal auto-correlation characteristic belonging to the PN-PBP, PN-PCP and PN-ER sequences are proved.Secondly, the proposed and carefully designed training sequences are applied to the TDS-OFDM transmit diversity system and several demodulation algorithms for the receiver are studied, including channel estimation, frame synchronization and Carrier Frequency Offset (CFO) estimation. Two channel estimation methods at receiver are investigated, one is with multiple frame information and the other is with single frame information. Which one can bring better performance depends on the properties of the adopted training sequences, the padding modes, and the strength of time and frequency selectivity of the channel. However, when the proposed training sequences are used, both methods can reach the lower bound of channel estimation error under the Least Squared (LS) criterion, and make the system Bit Error Rate (BER) be approximate to the case of ideal channel estimation. At the receiver, fast frame synchronization is possible via cross-correlation between received signal and local reference sequences followed by peak detection. The cross-correlation based frame synchronization also has the ability to perform well under channels with large CFO and large delay spread. For the TDS-OFDM transmit diversity system with PN-ER sequences, a new CFO estimation algorithm is developed, where the CFO is estimated by computing a correlation vector and constructing a Hermitian matrix which contains the information of CFO.Finally, a differential IBI cancellation method for OFDM system is proposed. Through differential operation upon the adjacent signals in frequency domain, the IBI can be effectively removed and the differential outputs construct a convolutional-like structure, which can be demodulated using the maximum likelihood sequence detection algorithm. Analysis and simulations show that the proposed algorithm can eliminate the impact of IBI very well, even under the channels with severe multi-path.