Research On Nonlinear Network Model Of Microwave Semiconductor Circuit

With the rapid development of the electronic communication industry,the third-generation semiconductor devices represented by Ga N(gallium nitride)have become a research hotspot.Because the application environment of the device gradually develops toward high frequency and high power,the nonlinearity of the device must be paid enough attention.At the same time,because current communication systems mostly use multi-frequency signals,the establishment of dual-frequency or even multi-frequency nonlinear models of microwave semiconductor circuits has broad prospects in guiding device production and circuit design.Traditional S parameters can't represent devices under high power nonlinear characteristics,so the theory of X-parameter has become the focus of nonlinear research on microwave semiconductor devices as soon as it is proposed.As a classic representative of behavioral model,X-parameter does not need to understand the physical characteristics of the device,and at the same time it can well describe the nonlinear characteristics of the device.Since the domestic research on nonlinear modeling started late,it is very important to study nonlinear models of microwave semiconductor devices.Based on the study of the nonlinear characteristics and modeling methods of the device,this article mainly focuses the following work.1.Firstly,the related theories of PHD(Multi-Harmonic Distortion)model and X-parameter model are introduced,as well as the important characteristics and correlation of the two models.At the same time,the dual frequency X-parameter theoretical model is analyzed.In addition,two methods of parameter extraction by ADS software and nonlinear vector network analysis are introduced.2.The neural network algorithm of Extreme Learning Machine is applied to the study of nonlinear behavior model.Powerful Learning and generalization ability of the algorithm just makes up for the lack of large amount of data and long measurement time in the measurement system of X-parameter.The research on single-frequency and dual-frequency nonlinear prediction models adopts the same establishment method.After data is extracted from ADS,the obtained data is divided into training set and test set.Then,after model training with training set,the test set is used to detect the accuracy of the obtained model.In this paper,on the basis of single frequency model,the main research double-frequency X-parameter prediction model,the model simulation result shows that in 96% of the data,the nonlinear model to predict the maximum relative error is less than 0.02,and the single,double-frequency X-parameter prediction model and circuit model,the harmonic balance simulation comparison,the results show that the nonlinear prediction model has good accuracy,At the same time,the prediction model is used to design the actual power amplifier circuit,and the measured results are basically consistent with the simulation results.All these results indicate that the X-parameter model based on ELM neural network can accurately characterize the characteristics of nonlinear networks.3.The load related X-parameters were simulated under different impedances.The simulation results show that different harmonic impedances have different effects on device characteristics.Finally,based on the nonlinear study of the device in this paper,a reasonable X-parameter measurement hardware platform is established,the measurement device and experimental target are determined,and the measurement process is refined,forming a complete test scheme.