Research On Demodulation Technology For Short-CP OFDM

In Orthogonal Frequency Division Multiplexing(OFDM)technology,cyclic prefix(CP)suppresses the impact of multipath channels at the expense of spectral efficiency loss.However,in the fifth generation of mobile communication system(5G),in some channel environments where the channel impulse response is quite long,CP with sufficient length would lead to a considerable reduction in spectral efficiency.If the CP length is insufficient,the delay spread of the wireless channel will cause intersymbol interference(ISI)and intercarrier interference(ICI).In this case,the performance of the traditional single-tap equalization degrades seriously,which results in a poor bit error rate level.Therefore,this paper studies a series of demodulation methods for OFDM in short CP scenes to improve the poor bit error rate level of the traditional single-tap equalization.First,this paper introduces the OFDM system model in a short CP scenario.The formula expression of the received signal in the frequency domain of OFDM in short CP scenes is obtained by detailed formula derivation,and the equivalent channel matrix of OFDM in short CP scene is derived,and the element expressions of ISI and ICI interference matrix are calculated.Then,the wireless channel model adopted in this paper is introduced to analyze the power of the interference matrix elements,and the relationship between ISI/ICI and subcarrier distribution is simulated.It is concluded that for a given subcarrier in OFDM system with short CP scenes,the strongest ISI/ICI contribution comes from adjacent subcarriers of the previous symbol and the current symbol,respectively.Finally,the BER performance of OFDM systems using single-tap equalization technique in short CP scenarios is discussed.Subsequently,this paper studies the interference cancellation detection method of OFDM in short CP scenes.Considering the disadvantages of high complexity or large detection delay for parallel interference cancellation(PIC)and serial interference cancellation(SIC),a reduced complexity interference cancellation(RCIC)algorithm is proposed.The algorithm reduces the computational complexity and realizes the combination of serial detection and parallel detection,which compromises the detection delay and hardware complexity,and the performance of these interference cancellation detection methods has been compared in the corresponding channel conditions.Then,considering the interference cancellation still has the poor bit error rate level in some short CP scenarios,we study the maximum likelihood detection method for OFDM in short CP scenes.Because maximum likelihood sequences estimation(MLSE)has extremely high complexity,a reduced state maximum likelihood(RSML)detection algorithm is proposed.Simulation shows that when ISI/ICI exists,the performance of RSML detection is better than that of interference cancellation detection,and its detection performance is even better than that of single-tap equalization with sufficient CP.Finally,in order to further improve the RCIC bit error rate performance and reduce the excessive online computational complexity of RSML,this paper proposes a neural network structure suitable for OFDM demodulation in short CP scenes.The simulation verifies that in the typical short CP channel scenario,the performance of the proposed deep learning network structure is significantly better than that of interference cancellation detection,and its performance can even approach the performance of RSML.