A PSAM Channel Estimation Based On DFT And DCT For OFDM System

OFDM technology having advantages of strong anti-interference capability, high spectrum efficiency and using IDFT/DFT to realize modulation and demodulation, is an attractive method for high data transmission. Owing to its distinctive advantages, it has been widely adopted by wireless local area network (WLAN) standards, including outdoor WLAN, European digital audio broadcasting and digital video broadcasting.One of the defects of OFDM system is its sensitivity to the frequency offset. In the poor propagation environment of wireless mobile communications, transmitted symbols will have distortions in amplitudes and phases, which will also limit the speed of transmission. Therefore reliable channel estimation methods are the keys to achieve high-speed data communications for OFDM system, which are the foundation to complete dynamic bit allocation and signal coherent detection. The accuratechannel state information (CSI) which is obtained by channel estimation technique can improve estimation quality and reduce BER for OFDM system. Channel estimation algorithm for OFDM systems is divided into three types. The first is a channel estimation method based on pilot symbol assisted modulation (PSAM), which uses pilots inserted in data streams to estimate channel frequency responses (CFR). Although this algorithm is very easy to realize, it occupies a large number of spectrum resource of the system and brings inevitable time delay. The second channel estimation is based on decision direct (DD). It takes advantage of the channel response estimation value of the previous frame or symbol to estimate the channel response in current time. In order to improve performance of channel estimation, this algorithm also utilizes some pilot symbols. The third is a kind of blind channel estimation (BCE), which estimates channel state information through statistical characteristic of transmitted symbols. On account of no pilot insertion in the data sub-carriers, this approach saves a lot of bandwidth, but its convergence rate is slow and computational complicacy is high.This paper focuses on the research fields on the PSAM channel estimation algorithm. At the same time, we make use of matlab software to simulate the OFDM system and analyze the channel estimation performance.Aiming at the various patterns of pilot insertion, to begin with, the paper studies the inserting methodology including block-type pilot, comb-type pilot and rectangular pilot; then, there is analysis of channel estimation methods in details, such as least square (LS), linear minimum mean square error (LMMSE), singular-value decomposition (SVD), discrete fourier transform (DFT) and discrete cosine transform (DCT); thirdly, it introduces several interpolation algorithms, for instance, constant interpolation, linear interpolation, Gauss interpolation, Cubic interpolation, DFT interpolation and so forth; finally, it compares and analyzes the channel frequency responses derived from applications under different channel estimations, interpolation algorithms and pilot patterns. The simulation results menifest that the performance of LMMSE channel estimation algorithm is the best, but its calculation is the largest; SVD simplifies the computational complexity because of matrix dimension reduction, whose performance is nearly consistent with LMMSE; poor as performance of LS is, it is simple to compute and prone to inplementation; the estimation effects become bad with the"error floor"at high signal-noise-ratio (SNR) for the interpolation algorithms based on DFT and DCT.There is some improvement about the"error floor"exsiting in the channel estimation algorithms based on DFT and DCT interpolation with the application of the comb-type pilot in this paper. For channel estimation algorithm based on the IDFT/DFT, at the beginning, channel frequency response is obtained under the LMMSE criteria at the pilot locations; then, regard the CFR of pilot sub-carriers as the input of IDFT; afterwards, transform channel impulse response linearly from the last step in the time domain; finally, overall CFR is acquired through DFT interpolation algorithm. There still exsits the"error floor"in the bit error ratio (BER) curve for this method, that maybe because the interpolation algorithm applied in this approach is not precise, which leads to the inaccuracy for channel estimation. So another modified step is proposed, that is getting CFR through linear interpolation algorithm which is transforming CIR of pilot positions into CFR by DFT and then computing complete CFR in the frequency domain. The BER curve from the modification eliminates the"error floor"thoroughly and improves the system performance. For channel estimation algorithm based on the DFT/IDFT, after transform CFR through DFT under LMMSE criteria, execute low-pass filter and add zeros for the data sequences in the transform domain, then transform the sequences into frequency domain by IDFT. Due to adopting LMMSE criteria, computation is larger but there is no obvious improvement about the performance. Maybe it is because the cutoff frequency in the transform domain and the adjustment coefficient in the frequency domain are selected incorrectly. For channel estimation algorithm based on the IDCT/DCT, transform CFR derived from LMMSE criteria into time domain by IDCT operation after multiplying the gain factor, and then conduct DCT interpolation after padding zeros at the tail of CIR in the time domain. As a result of the different points of IDCT and DCT operation, another gain factor need to be multipled in order to compensate the unmatched transform before, however the estimated consequence is still not ideal. For channel estimation algorithm based on the DCT/IDCT, pad zeros at the end of"spectrum sequence"in the transform domain after implementing the conventional DCT operation towards CFR derived from LMMSE criteria, and then transform the new"spectrum sequence"into frequency domain by extendible IDCT (EDICT) to get the whole CFR. Owing to the shift of data sequence after DCT operation, we cannot adopt IDCT directly but EDICT to fulfill the compensation. So the estimation performance is improved greatly with the system BER being reduced.Simultaneously, the performance of the modified channel estimation algorithms are simulated and compared under the different system and channel parameters. The simulation results illustrate that the modified approaches can increase the channel estimation accuracy and improve the system communications quality by taking advantage of small amplitude and phrase modulation, increasing the number of pilots, selecting the large number of carriers in DFT, reducing the Doppler shift, cutting down the number of wireless channel paths and choosing more precise interpolation algorithms.