Research On Frequency Offset Estimation And Channel Estimation Techniques In Wideband OFDM Systems

Orthogonal frequency division multiplexing (OFDM) is now considered as an effective technique for high rate data communications, since it enjoys high bandwidth efficiency and is more robust against frequency selective fading compared to other multiplexing technique. However, carrier frequency offset (CFO) is one of the major disadvantages in OFDM systems. CFO destroys the orthogonality among subcarriers, which in turn causes inter-carrier interference (ICI) and degrades the system performance. Therefore, CFO estimation is very crucial in OFDM systems. On the other hand, in OFDM systems, coherent decoding requires channel state information (CSI), which is usually difficult to obtain, especially for MIMO-OFDM systems since the received signal is the superposition of the signals transmitted from different transmit antennas simultaneously. Therefore, the technologies of CFO estimation and MIMO channel estimation for OFDM systems are investigated in this dissertation.Firstly, CFO estimate methods based on an OFDM training symbol with L identical parts are studied. We propose an efficient CFO estimation scheme based on an OFDM training symbol with L identical parts. In this scheme, the estimation is processed in the two steps of fine and coarse estimations. The fine estimation is obtained by the correlation of two identical halves in the training symbol. The received OFDM training symbol is then reshaped into several sub-symbols. Furthermore, a modified fast Fourier transform (FFT) is presented incorporating the compensation of the fine estimate of CFO. After applying the modified FFT to the sub-symbols, the coarse estimation is accomplished by finding the maximum magnitude of the combined frequency domain signals. It turns out that more sub-symbols involved in the combination process, more improved estimate performance can be expected at the cost of increased computational complexity. To reduce the computational load, a method of adaptive adjusting the scale of used sub-symbols based on estimate reliability evaluation is presented. The proposed CFO estimator has satisfactory estimate performance, while maintaining very low computer complexity.The estimate range of the traditional CFO estimate method based on L identical parts is±L/2. Since L should be a relatively small value to satify the requirement for the practical systems, only a relatively small CFO can be estimated based on the conventional preamble with a reasonable L. For the practical systems, a narrow estimate range means that oscillators with high precision must be used, thus increasing the system cost. The conventional large CFO estimate method processed in the frequency-domain necessitate much computational load due to the correlations for all possible CFOs or the discrete Fourier transform (DFT). To estimate large CFO while maintaining low computational complexity, the dissertation proposes a special preamble composed of two training symbol blocks with several wide-sense identical parts. Based on the designed preamble, two large CFO estimate methods are proposed. By elaborately selecting the numbers of identical parts for the two training blocks, the proposed methods have larger estimation range while achieving approximately the same estimate performance compared to the estimation method based on the conventional preamble with several identical parts. Moreover, they have the advantages of fast estimation speed and low computational complexity over the frequency-domain estimators.Channel state information of STBC-OFDM system is required for maximum likelihood decoding. A subspace-based semi-blind method is proposed for estimating the channels of STBC-OFDM systems. The channels are first estimated blindly up to an ambiguity parameter utilizing the nature structure of STBC, irrespective of the underlying signal constellations. Furthermore, a method is proposed to resolve the ambiguity by using a few pilot symbols. Our proposed semi-blind estimator does not need precoding and has not the constraint of constant-modulus (CM) modulating. Moreover, the estimator can converge with only a few tens of received data blocks. Simulation results show the proposed semi-blind estimator can achieve higher spectral efficiency and provide improved estimation performance compared to the non-blind estimator.Channel estimation in the time domain turns out to be more efficient since the number of the unknown parameters is greatly decreased compared to that in the frequency domain. In this dissertation, the time-domain channel estimation for MIMO-OFDM systems in sparse multipath channels is studied. A new time domain channel estimation approach is proposed. In this approach, the DFT based least square (LS) channel estimator is used for initial channel estimation. The parametric channel model is considered and the GAIC is introduced to estimate the significant taps and the channel length. This effectively reduces mean squared error of the DFT based LS channel estimation. Furthermore, a SLS-like algorithm is proposed to improve the channel estimation performance by utilizing the intersymbol interference (ISI)-free samples in the CP. The proposed channel estimator can considerably improve the estimate performance of the conventional LS estimate method.The channel estimation for MIMO-OFDM systems in time-varying fading channels using the expectation-maximization (EM)-type algorithm is studied. The conventional method uses the decision-directed method to track the channel varying. In that method, the channel estimate at the previous data block is used as the initial value for the iterative channel estimation at the current data block. However, the estimate performance will deteriorate greatly when used for fast time-varying channel. To compensate for the fast Rayleigh fading, a simple interpolation method is presented. The study shows that the interpolation can improve the estimate performance of the EM-type channel estimation. Moreover, it can fast the convergence rate of the EM algorithm. Considering the multipath propagation in practical wireless channels is aptly modeled by a few dominant specular paths, we propose a scheme to improve the channel estimation performance of EM-type algorithm by using a low rank approximation. By then using subspace tracking, it is observed that our proposed channel estimation method for MIMO-OFDM systems is feasible. The subspace tracking aided low rank approximation method can effectively reduce channel estimation error and thus improve systems performance compared to the method based on the traditional EM-type algorithm. Finally, we give a brief summary of the originalities in this dissertation, and the further research direction of carrier frequency offset estimation and channel estimation techniques in OFDM systems as well.