Study On Channel Estimation And Equalization Techniques For OFDM Systems

Orthogonal frequency division multiplexing (OFDM) is a multi-carrier modulationtechnology. It has many well known advantages such as robustness to multi-path fadingchannels, high bandwidth efficiency, and efficient implementation. OFDM has beenchosen as a solution for physical layer transmission in many wireless communicationstandards.In order to transmit data rapidly with high spectral efficiency, the coherentdetection is usually employed in OFDM system, where the real-time and accuratechannel state information (CSI) should be known by the receiver. The CSI also shouldbe known in other situations, which combine OFDM and adaptive modulation, beamforming or multiple-input and multiple-output (MIMO) transmission. Therefore,channel estimation is the essential part of OFDM systems. In this dissertation, newsolutions and algorithms related on OFDM channel estimation and interferencecancellation techniques are proposed for time-varying channels.In the first part of the dissertation, the properties of the fading channel and itsmathematical model is discussed. After the briefly introduction of the conventional LSestimator, the LMMSE estimator, the DFT-based transform domain estimator, a newchannel estimation algorithm based on statistical analysis is proposed.The traditional DFT-based channel estimator only removes the noise in the channelimpulse response (CIR) beyond the length of the cyclic prefix (CP)，while the noisewithin the length of the CP is not suppressed at all．In most of the improved DFT-basedchannel estimation algorithm, one decision threshold is obtained by some methods.When this threshold was applied to the CIR within the length of the CP to remove noise,the accuracy of channel estimation is improved. To obtain the desired results, a crucialproblem is how to find an optimal threshold value. In addition, these kinds of methodsare invalid, when the value of CIR estimation is close to the threshold. From themultivariate analysis method, the noise suppression problem is a typical discriminantanalysis problem, since the CIR is composed of the ideal channel taps and the addednoise taps. To discriminate the uncertain taps, the Mahalanobis distance is used todiscriminate the CIR. By introducing the distance measure, the defects of thethreshold-discrimination method are overcome effectively. The channel estimationmethod based on DFT for OFDM is suitable for sample spaced channel. The energy of channel impulse response will leak to all taps when the non-sample spaced channel istransformed by the DFT. By changing the discriminated range, the improved channelestimation method is presented, which is suitable for the non-sample spaced channelbased on DFT for OFDM systems.In the interval of one OFDM symbol, the wireless channel is usually assumed to besmooth. The high speed motion of subscribers gives rise to Doppler effects that destroyorthogonality between sub-carriers, leading to ICI and performance degradation for theOFDM system. An algorithm for fast-varying channel estimation was proposed by Choi.This method need to know the channel statistical knowledge, which cannot be obtainedaccurately in practice. Furthermore, its computational complexity is very high.The time-varying channel can be approximated by a linear model during anOFDM block, when the normalized Doppler frequency shift is less than0.2. In this case,the channel is modeled as a linear function, and the CIR can be obtained by onlyestimating the slope and the reference signal of linear model. When the linear model isused to approximate the time variations of channel transmission function, there will becertainly some errors. Based on total least squares (TLS) criterion，a novel method forestimating channel parameters variation within an OFDM symbol is proposed. Thesimulation results show that the new method is more robust when Doppler frequency ishigh.Using the CP as the guard interval is a simple way to combat the inter-symbolinterference (ISI). However, it also reduces the transmission efficiency of the system.For some channels with a very long delay spread, it is possible that the length of the CPis beyond the channel length. The interference due to the insufficient CP was analyzedand the SIR was derived for OFDM systems. In order to cancel interference, a newmethod for interference cancellation, based on total least squares (TLS) criterion, isproposed.Due system terminal mobility and multi-path effects channel stability is difficult tobe maintained in wireless communication systems. Time variations of doubly-selectivechannels lead to the loss of orthogonality between sub-carriers, resulting in the increaseof the ICI energy and error of the channel estimation at pilot sub-carriers. Furthermore,the one tap frequency equalizer is not applicable any more in this case. The new methodwas proposed for the signal restoration via the regularized constrained total leastsquares. Errors caused by channel inaccuracies can be mitigated after equalization.Theoretical analysis and simulation results show that the method can not only eliminate noise and interference, but also recover the transmitted data．...