Study And Realization Of Channel Estimation Technology For OFDM System

OFDM technology having the advantages of high spectrum efficiency, strong anti-interference capability and using IDFT/DFT realization modulation and demodulation, has been widely adopted by wireless LAN standards. Because of OFDM system is very sensitive to frequency offset, and the propagation environment of wireless channel is poor, causing amplitude and phase distortion of the transmission signals, and limiting speed of them. Therefore reliable channel estimation is the key issue to achieve high-speed data communication for OFDM system.This paper analyzes the transmission characteristics of wireless channel together whit principle and anti-interference mechanism of OFDM technology, studies various channel estimation algorithms and their performance in OFDM system, and realizes LS channel estimation and equalization algorithm based on IEEE802.11a's long training sequence on FPGA. Channel estimation methods can be broadly divided as the data symbol auxiliary and blind channel estimation. LS, MMSE, SVD, LMS algorithm based on training and the Wiener filter algorithm, linear interpolation, Gaussian interpolation, Cubic interpolation, insert in transform domain algorithm based on pilot are researched in the thesis. After studying in-depth of channel estimation method based on decision direct of training sequence and interpolation of pilot, a training guided decision feedback equalization (DFE) method is proposed for OFDM system. By comparing the symbol error rate (SER) curve of receiving signal after equalization, MATLAB simulation is employed to analyze and compare performance of several common and improved channel estimation algorithms. Simulation result indicates that LS algorithm based on training is only suitable for slow time-varying channel; MMSE algorithm performance is the best, the performance of MMSE algorithm improved significantly compared to LS; compared to MMSE, SVD have a low complexity while a poor performance. Based on Pilot subchannels interpolation algorithm is a rough method of curve fitting, the performance of equalization dependences on accuracy of interpolation method and the number of insertion pilots, relatively poor. The proposed algorithm can effectively combat the fading in received signals caused by the multi-path propagation and Doppler shift, and can achieve obviously improved performance to the single long training sequence and interpolation algorithm based channel estimation. The implementation in high speed mobile circumstance is also applicable.Finally, the LS channel estimation algorithm based on IEEE802.11a's long training sequence is realized on FPGA. The successfully debugged channel estimation module is added the overall IEEE802.11a baseband processing module to be tested. The results indicate that the receiving data of various modulation after equalized are ideal, to meet the actual system design requirements.