Research On Key Technology Of Broadband Satellite Polarization Interference Cancellation

With the proposal of global seamless coverage,the broadband satellite communication is developing rapidly,the number of satellites is increasing,thus the limited spectrum resources become more and more precious.In order to meet the requirements of system capacity while saving spectrum resources,dual-polarization frequency reuse(DPFR)is often adopted in broadband satellite communication systems.However,the depolarization of channel and the imperfection of devices will cause cross-polarization interference between two orthogonal channels.Therefore,cross-polarization interference cancellation(XPIC)is an indispensable research topic in the DPFR system.In order to increase the power efficiency,the high-power amplifier(HPA)usually works near the saturation region,and thus the severe nonlinear characteristics are existed between the input signal and the output signal of HPA.The nonlinear characteristics of satellite channel will further increase the difficulty of eliminating the cross-polarization interference at the receiver,and then reducing the system performance.In order to ensure the validity and reliability of the broadband satellite DPFR transmission system,the XPIC technology under the broadband satellite channel is studied in this thesis.First of all,the broadband satellite channel is described by two models.One is a HPA model with both nonlinear effects and memory effects,which is composed of a linear filter and a traveling wave tube amplifier(TWTA).The TWTA model and the group delay model are described in detail,and the characteristic curves of the TWTA model and the group delay model can be obtained through simulations.The other is the cross-polarization interference model.The cause of cross-polarization interference is explained and the index used to measure the degree of cross-polarization interference is introduced.Next,the XPIC technology based on training sequences is studied.Considering the strong nonlinear feature extraction ability of convolutional neural network(CNN)and the good time sequences processing ability of long short-term memory network(LSTM),a XPIC neural network architecture is proposed and compared with the traditional XPIC algorithms.The simulation results show that,compared with the traditional XPIC algorithms,the proposed neural network architecture can overcome the nonlinear effects of broadband satellite channel,and effectively eliminate the cross-polarization interference of the received signal,thus can achieve more superior XPIC performance.Then,considering the waste of channel resources caused by periodically sending training sequences,the blind XPIC technique is studied.In order to improve spectrum utilization,high-order modulation is often used in satellite communications.Thus,the traditional constant modulus algorithm(CMA)is modified to apply to the high-order modulation signal with multiple amplitudes.Then,inspired by the traditional dual-mode CMA and the decision directed-LMS(DD-LMS)algorithm,a transversal filter architecture that combines the modified CMA algorithm and the DD-LMS algorithm is proposed,and compared with the conventional blind XPIC algorithms.Simulation results show that the proposed transversal filter architecture can effectively eliminate the cross-polarization interference of the received signal without the training sequences and has better XPIC performance compared with the conventional blind XPIC algorithms.