Reaseach On Modeling And Optimization Techniques Of Radio Frequency Devices Based On GaN Technology

In recent years,the development of radio frequency technology has been very rapid,and the application of radio frequency microwave devices in engineering technology has become more and more extensive.In order to improve efficiency and design accuracy of chips and related modules,it is necessary to accurately characterize the radio frequency characteristics of radio frequency devices at different frequency.On the one hand,the above tasks can be described by behavior-level modeling,on the other hand,they can be described by analysis and calculation through accurate threedimensional full-wave electromagnetic simulation methods.However,as the frequency increases,the equivalent circuit model parameters of the RF devices will show a certain deviation in different application scenarios,and the traditional empirical model will no longer be applicable.It is timeconsuming and labor-consuming to obtain device characteristics through electromagnetic simulation.With the improvement of computer hardware,the emergence of big data and the improvement of algorithms,artificial intelligence technology has gradually begun to be applied to circuit design relying on its powerful nonlinear mapping capabilities.The rapid development of artificial intelligence technology in recent years has also laid a solid foundation for RF device modeling and optimization.In this article,combined with the specific third-generation compound semiconductor Ga N process,an artificial intelligence-based radio frequency modeling method is proposed.Specifically,using the microstrip line model as the basic model,and on the basis of empirical model modification,an artificial intelligence(AI)-based modeling optimization method was introduced to develop a hybrid microstrip line model,which effectively improved the accuracy and Scope of application.The main content and innovation of this article are:First,the characteristic parameters of radio frequency devices are simulated by three-dimensional electromagnetic simulation software as a data set,and artificial intelligence algorithm training is used to generate AI models that can replace simulation tools.This model can accelerate the process of generating radio frequency characteristic parameters and replace electromagnetic simulation tools to improve radio frequency.Circuit design iteration efficiency;Secondly,taking the microstrip line of the specific process as the basic device sample,based on the traditional empirical model,a new modified model for the influence of frequency,multilayer dielectric and substrate loss is proposed,and its accuracy is verified;Finally,a hybrid model combining modified model,genetic algorithm and AI model is proposed to realize the overall optimization from design target parameters to physical parameters.This modeling method can speed up the process of designing microstrip line components,and accurate correction models and genetic algorithm optimization methods with small errors can simplify the long iterative process of traditional multiple design simulations.