Study On Doppler Frequency Offset Estimation Of OFDM/MIMO System In High-speed Railway Scenario

In recent years, with the rapid development of wireless communication technology, LTE (Long Term Evolution) technology has developed more mature and been used in a large-scale commercial application. LTE technology not only promotes the innovation of the mobile intelligent terminal equipment and hardware, but also accelerate the development of mobile internet. At the same time, Chinese high-speed railway construction is also carried out like a raging fire, exported overseas, becoming a new name card of Chinese technology. Mobile users in the high-speed rail scenario are growing more and more and the data service requirements are also getting higher and higher. Original GSM-R (Global System for mobile Communication-Railway) can’t meet the demand of rapid development of the high-speed railway communication business diversity and high quality communication, therefore the implementation of LTE-R (LTE for Railway) is imperative.In high-speed mobile scenario, the relative speed between the user and the base station is very high, which results in a large Doppler frequency offset. The downlink of LTE system uses orthogonal frequency division multiplexing (Orthogonal Frequency Division Multiplexing, OFDM) technology which is sensitive for the frequency offset. Smaller frequency offset will affect the orthogonality of sub-carrier and destruct system performance. Therefore, in high-speed mobile scenario, Doppler frequency offset estimation has become an important issue.In terms of the current study, carrier frequency offset estimation in OFDM systems (Carrier Frequency Offset, CFO) has been very mature in normal cases. The Doppler frequency estimation model in high-speed mobile scenario is similar to the CFO estimation model of OFDM systems. Therefore the mature methods of OFDM frequency offset estimation can be applied to the Doppler offset estimation in the high-speed mobile scenario. This paper mainly studies three algorithm:Moose algorithm, S&C algorithm and algorithm based on the cyclic prefix. The principles of the three algorithms are all to calculate partial correlation to get frequency offset, but the structures of symbols are different. In terms of computational complexity, Moose algorithm is higher than the other two algorithms. On the simulation performance, the estimation error of the three algorithms decrease with SNR increasing, Moose algorithm has the best performance under low SNR, but the error floor occurs earlier, the error floor of S&C algorithm and the algorithm based on CP occurs later.Due to the high-speed railway channel meeting the features of Rice channel, this paper adopts Rice channel to model the high-speed time-varying multi-path channel, and evaluates the performance of Moose algorithm, S&C algorithm and the algorithm based on CP in this channel. The simulation results show that estimation error decreases with the SNR increasing, and the greater K factor, the better estimation performance. Considering that MIMO can reduce the randomness of the channel, using multiple antennas to receive signals and then calculating the average frequency offset estimation can improve the estimation performance. This paper compares the estimation performance by using two or three receive antennas to using one receive antenna. The result shows that the estimation performance can be improved by adding the number of receive antenna. When the SNR is low, the SNR can improve 2-3dB by adding one receive antenna at the same estimation performance. In multipath scenarios, signal transmission affected by multipath effects, the received signal is affected by a direct path and scattering path together. However, the high-speed railway has a very strong direct path, the received signal is mainly coming from the transmission of direct path, and the effect of scattering path is weak. Therefore, this paper proposes a new method by using only received signal from direct path to estimate the Doppler offset, which reduces the complexity of frequency estimation. The simulation uses the result of the total received signal subtracting the received signal from all scattering paths as the received signal from the direct path to estimate the frequency offset. Compared with the original method, the new one has a significant improved performance. The paper compensates the frequency offset of the received signal after the estimation, the simulation shows that the BER after compensation by the frequency offset estimation using LOS path is smaller than the BER after compensation by the frequency offset estimation using multipath.