Nonlinear Behavior Modeling Of Receiver With Neural Network

With the development of wireless communication technology,the variety and quantity of electronic equipment are increasing,which leads to the increasingly complex electromagnetic environment.As the core equipment of electronic communication system,receiver contains various nonlinear devices such as amplifier and mixer.The interference signal will make the receiver produce non-linear effects such as gain compression and intermodulation;The modulation signal with wideband and non-constant envelope makes the memory effect of the receiver more significant,which affects the nonlinear effect of the receiver and leads to the distortion of the received signal.Therefore,the study of receiver nonlinear effect,the construction of receiver nonlinear behavior model and the prediction of receiver nonlinear response have certain reference value for the performance evaluation of receiver in complex electromagnetic environment.The topic of this thesis originated from the project of "Nonlinear Effect Measurement and Evaluation Technology".Based on the basic theory of receiver and ADS simulation,the nonlinear effect and memory effect of receiver are analyzed in detail.Based on the neural network,the nonlinear behavior models of the receiver under the input of multi-tone signal and wideband modulation signal are constructed respectively,and the model is verified by the simulation and measured data.The main work of this thesis is as follows:1.The nonlinear characteristics of receiver are analyzed by combining theory with simulation.The source of nonlinear characteristics of receiver is analyzed,and the mechanism of nonlinear generation is expounded.The simulation link of receiver system is built by ADS,and the nonlinear phenomenon and memory effect are analyzed.2.A receiver nonlinear behavior modeling method based on generalized regression neural network is proposed for single-tone and multi-tone signal input.The nonlinear behavior model of receiver is composed of frequency conversion module,nonlinear module and frequency response module,which is realized by MATLAB algorithm.The frequency conversion module down converts the received signal,and the nonlinear module and frequency response module are used to characterize the nonlinear characteristics and memory effect of the receiver.The neural network is trained by ADS simulation data,and the optimal smoothing factor ? is sought by cross-validation.The model is verified by singletone signal and two-tone signal.The results of time domain and frequency domain show that the accuracy of the model is good.3.A receiver nonlinear behavior modeling method based on radial basis function neural network is proposed for wideband modulated signals input.The memory effect of the receiver can not be ignored when the wideband modulation signal is input.The memory effect of the receiver is simulated by adding a delay tap at the input of the model.The center of hidden layer and weight of the model were selected and learned by K-means clustering algorithm and orthogonal least squares algorithm respectively.The model is trained by the input and output measured data of the receiver,and verified by the in-phase and orthogonal components of the wideband signal.The simulation results of the model are consistent with the measured data,and the normalized mean square error of the model can reach-41 dB.The results show that the neural network model has fast convergence speed,good modeling accuracy and generalization ability.