Research On Pre-distortion Model Of Satellite Communication Power Amplifier

With the development of the communication industry,the importance of satellite communication is becoming more and more obvious.Power amplifiers in satellite communication system have a great impact on the quality of the signal.Excessive input signal power will produce nonlinear distortion.The low efficiency will also affect the adjacent frequency-domain,so the technique of the power amplifier linearization is imperative.In this paper,the importance of digital pre-distortion technology in improving the nonlinear distortion of power amplifier is introduced.The main work and innovation of this paper are as follows:(1)Based on the adaptive theory,the Volterra series pre-distortion system is established,and the least mean square algorithm is used to extract the parameters.After the signal is pre-distorted,the adjacent channel interference is improved by about 12dB.With the establishment of the memory polynomial pre-distortion system,and the use of recursive least squares algorithm for parameter extraction,the adjacent channel interference is improved about 14dB after the signal is pre-distorted.(2)The BP neural network and RBF neural network are studied in detail.The RBF neural network model is established.The K-means clustering method is used to obtain the radial basis center,and the weights are updated by using the LMS algorithm.Experimental results show that the RBF neural network has higher accuracy than Volterra Series and memory polynomials.By analyzing the characteristics of the complex domain function,the limitation of the neural network is expounded,the DUAF structure based on BP neural network is proposed and the pre-distortion system is established.The simulation results show that the pre-distortion of the signal is improved by about 7dB,and its pre-distortion effect is not ideal,which verifies the limitation of the DUAF structure to deal with the complex number.(3)Focus on the neural network model of complex domain.A fully connected recurrent neural network(FCRNN)pre-distortion system is established,and the parameters are extracted by RTRL algorithm.The simulation results show that FCRNN has higher accuracy compared with the DUAF structure when dealing with complex signals.(4)An improved short-term memory recurrent neural network(STMRNN)is proposed.The FCRNN model uses fully connected neurons feedback signal which increases complexity.In the STMRNN model,the feedback signal of the output layer in FCRNN model is changed to the short-term memory,so the model complexity can be reduced while the accuracy is guaranteed..(5)An improved all feedback short-term memory recurrent neural network(AFSMRNN)is proposed.The AFSMRNN model transforms the hidden layer feedback signals in the STMRNN model into short-term memory,which further reduces the model complexity and achieves the same accuracy as the memory polynomial.