Research On The Modeling And Predistortion Technology Of Navigation Signal Radio Frequency Channel Based On Neural Network

The high-precision applications of satellite navigation imposes high requirements on the quality of navigation signals,while the non-ideal characteristics of satellite payload components will induce signal distortion,degraded signal quality,and affect the positioning accuracy of navigation signals.Therefore,it is necessary to study the non-ideal properties on board.At present,the research on the non-ideal characteristics of on-board payload mostly analyzes the influence of filter and high-power amplifier on signal quality independently,and their compensations are also performed separately.However,when the components of the satellite payload RF channel are assembled and integrated,it will be difficult to study each component separately.Especially when the environmental factors cause the channel characteristics to change,it is cumbersome and impractical to perform measurement and analysis on each component separately,so it is of great significance and necessary to study the non-ideal characteristics of the satellite payload RF channel and compensate it as a whole.In this thesis,based on the structural block diagram of navigation payload,a general satellite payload baseband equivalent model of radio frequency channel is established.It is difficult to separately extract the parameters of each sub-component in the model based on the overall input and output.Therefore,this thesis uses neural network methods to study the RF channel baseband equivalent model.This thesis expands the number of hidden layer in the RVTDNN model and proposes a new neural network model for modeling and predistortion of RF channels.This thesis introduces normalized mean squared errors to assess modeling accuracy.Moreover,four parameters related to the quality of navigation signals,such as out-of-band power loss,correlation loss,zero-crossing of the discriminant function,and slope distortion of the discriminant function zero-crossing,are introduced to evaluate the pre-distortion effect.In addition,amplitude and frequency response,group delay,AM-AM,and AM-PM are introduced to evaluate the linear and nonlinear characteristics of the channel to assist in evaluating the improvement of channel characteristics caused by predistortion.The simulation experiments show that the proposed neural network can accurately model the baseband equivalent model of the RF channel,and the predistortion of the baseband equivalent model of the RF channel improves the navigation signal quality overall.Compared with the RVTDNN model,the proposed model has higher modeling accuracy and better predistortion performance.