On Precoding Techniques In OFDM And MIMO Systems

In recent years, with the rapid increase of data services, e.g., real-time multimedia communications and high-speed Internet access, enhancing the transmission rate, bandwidth efficiency and quality of service (Qos) of communication systems has become a problem demanding prompt solution. Orthogonal frequency division multiplexing (OFDM) is one key technology of broadband wireless communication, because it can resist inter-symbol interference (ISI) caused by multi-path fading. While Multiple-input multiple-output (MIMO) can make use of space resource to increase the channel capacity, spectral efficiency and communication reliability significantly without increasing the transmit power and bandwidth. The combination of MIMO and OFDM is widely regarded as the core scheme of future wireless communication systems. This dissertation deals with the precoding techniques of OFDM systems and MIMO systems respecitively. The main achievements and results of this dissertation are listed as follows:1. In OFDM systems, the very different output SNR values of the subchannels will lead to poor bit error performance when equal bit allocation (EBA) is adopted. So, we proposed three novel joint precoder and detector (JPD) schemes that can transform all subchannels of an OFDM system into subchannels with identical channel gain. Two schemes are designed based on ZF criterion and the other scheme based on minimum mean square error (MMSE) criterion. Numerical analysis helps us to obtain the theoretical approximate BER values of the JPD schemes. Simulation results verify the numerical analysis and confirm that the performance of our proposed JPD schemes greatly outperform linear equalizer with EBA at high SNR values. JPD schemes outperforms ZF equalizer with EBA about 8 dB for QPSK modulation and about 6dB for 16QAM modulation at BER=10-3.2. A lattice-reduction aided linear MMSE (LMMSE) precoding with limited feedforward for spatial correlated MIMO channels is proposed to achieve lower BER by use of complex Lenstra-Lenstra-Lovasz (CLLL) algorithm. Then, we prove that the proposed algorithm obtains the full transmit diversity in correlated MIMO channels through diversity order derivation. The simulation results show that in TDD MIMO systems assuming correlated block flat fading channel, with QPSK modulation and MMSE uplink channel estimator at the transmitter, the uncoded BER of LR-aided LMMSE precoder is lower than that of traditional LMMSE precoder when Eb/NO is greater than 10dB, while with (2,1,3) convolutional channel coding and Viterbi decoding the coded BER of LR-aided LMMSE precoder is lower than that of traditional LMMSE precoder when Eb/NO is greater than 12dB at all correlation coefficients. Furthermore, the LR-aided LMMSE precoding also obtains the full transmit diversity M (the number of transmit antenna).3. In point to point MIMO systems, uniform channel decomposition (UCD) has been proven to be optimal in bit error rate (BER) performance and strictly capacity lossless when perfect channel state information (CSI) are assumed to be available at both the transmitter and the receiver side. However, in practice, CSI can be obtained at the transmitter if there is reciprocity between the forward and reverse channels in time division duplex (TDD) systems or can be conveyed from the receiver to the transmitter via a feedback channel. In any case, channel error is inevitable. In the first section, we consider the case of imperfect CSI at the receiver (CSIR) and imperfect CSI at the transmitter (CSIT) are the same. A novel robust UCD scheme and corresponding optimal robust power allocation are proposed, which are capable of improving the BER performance compared to the conventional UCD scheme and the robust linear precoding scheme. Simulation results show that the MIMO channel capacity of the proposed robust UCD scheme is higher than that of the conventional UCD scheme. By deriving and analyzing the MIMO channel capacity lower bound of the robust UCD scheme, we prove that our proposed robust UCD scheme is capacity lossless in a channel estimation error existing MIMO system. In the second section, we consider the case of imperfect CSIR and imperfect CSIT are different. A simple practical scheme for mismatch between CSIR and CSIT in limited feedback MIMO joint transceiver design is proposed. The proposed scheme designs the nulling vector or matrix at the receiver side with the quantized estimated CSI, which seems losing some information about channel, but eliminates the mismatch between CSIR and CSIT that may cause greater deterioration than the lost channel information. Moreover, we applied our scheme to three popular joint transceiver designs based on ZF and MMSE criterion respectively (SVD, GMD and UCD) and analyzed why the matching architecture of CSI will achieve better performance. Subsequently, using Shannon rate-distortion theory and generalized Lloyd vector quantization algorithm (GLA), we obtain the approximate variance of channel quantization error which can be substituted into the expression of nulling matrix of robust UCD scheme. The approximation enhances the practicability of our proposed robust UCD scheme in the more pratical scenario that imperfect CSIR and imperfect CSIT are different. The first set of simulation results show that the proposed matching architecture outperforms the conventional architecture at different channel estimation errors and vector quantization errors. The second set of simulation results in BER and ergodic capacity show the validation of the approximation of the variance of channel quantization error.4. Three criteria based multi-user multi-stream VP (MUMS-VP) algorithms are proposed:zero forcing expanded channel inversion (ZF ECI)-VP algorithm, minimum total mean square error criterion (MTMSE) based two MUMS-VP algorithms and maximum signal to leakage and noise ratio (MSLNR) criterion based MUMS-VP algorithm. The general expression of achievable rates of MUMS-VP algorithms is derived. Analysis and simulation results show that the proposed ZF MUMS-VP is equivalent to BD-VP, while MTMSE MUMS-VPâ…¡has the maximum achievable sum rates among these algorithms. All proposed algorithms have much lower complexity than BD-VP. Furthermore, MTMSE MUMS-VPâ… ,â…¡and MSLNR MUMS-VP greatly outperform BD-VP in BER performance for both fixed modulation scenario and adaptive modulation scenario.