Study On Method Of RF Devices And Circuit Modeling And Design Optimization

With the rapid development of wireless communication technology,human society will realize a leap in the era of intelligence,and the 5G/6G technology will be widely used in people’s lives.The rapid growth of the demand for intelligent terminal products,such as mobile communication,wireless WIFI,wireless Bluetooth earphones,will promote the increasing demand for transmitter system in the wireless communication system.In addition,with the speed of product iteration and standard update,shortening the design cycle of wireless communication system will be more conducive to the development of wireless communication technology.As an important part of transmitter system design,device/circuit modeling and design optimization can realize efficient communication between foundry and circuit designers.Accurate device and circuit models can achieve the system’s expected design objectives,and efficient circuit design optimization methods can shorten the design cycle.Artificial neural network has strong non-linear regression ability.Combining it with traditional modeling theory can effectively improve the modeling efficiency and accuracy.Based on the above analysis,the main content of this article is as follows:1.An efficient and accurate device modeling method is realized,and the proposed method is illustrated by taking coplanar waveguide as an example.Firstly,we have analyzed the realization method of calibrating HFSS simulation model with test data to obtain more data,and solved the key problems of insufficient test data and unstable simulation data.Then,based on the semi-empirical model of coplanar waveguide,a two-step genetic algorithm is proposed to accelerate parameter extraction,which significantly improves the precision and efficiency of component parameter extraction in equivalent circuits.Secondly,BP neural network is used to fit the nonlinearity between the dimensional structure parameters and the electrical characteristic parameters of coplanar waveguides,which solves the key problem of building a scalable model of the dimensional structure parameters of coplanar waveguides.In order to further improve the modeling accuracy,genetic algorithm is proposed to optimize the BP neural network,optimize the search process of the optimal weight and threshold,improve the fitting accuracy of the nonlinear relationship between the dimensional structure parameters and the electrical characteristics parameters,and thus achieve a high-precision scalable model,which lays a foundation for the realization of high-reliability and highprecision circuits and systems.2.A high-precision circuit behavioral modeling method is completed,and a Class-F power amplifier with high efficiency is taken as an example.Firstly,a method of dividing training data and validation data in the case of multidimensional data is proposed,which solves the problem of inaccurate nonlinearity fitting caused by uneven division of training data and validation data when the dimension of input and output parameters increases.In this paper,genetic algorithm(GA)and particle swarm optimization(PSO)are used to optimize BP neural network to fit the nonlinearity between the frequency and input power of power amplifier and the S-parameter and power parameter respectively,and the behavioral model of power amplifier is realized.On this basis,a method of optimizing BP neural network for power amplifier modeling based on NSGA-Ⅱ is proposed,which solves the problem of precision deterioration caused by the disunity of output node dimension in the process of establishing power amplifier behavioral model.The modeling accuracy can be further improved by using this method,thus providing a reliable scheme for system analysis and design.Finally,the device model and circuit behavioral model proposed in this paper are used to realize the application program Device & Circuit Model.The application program is not only user-friendly,but also realizes the data interface function between MATLAB and circuit design software.Thus,the device model and circuit behavioral model can be compatible with multiple platforms.3.A method for design optimization based on efficient and high-precision circuit behavior level models is presented,and a Sub-6G charge pump phase-locked loop is taken as an example to explain the specific implementation method.Firstly,it is analyzed that the tradeoff relation of performance index and the difficulty of establishing nonlinearity between objective optimization parameter and variable parameter during circuit design optimization process.Combined with the idea of multi-objective optimization proposed in Chapter 4 and the modeling method of circuit behavioral model proposed in Chapter 3,a general method of design optimization is proposed.Based on the modeling method of circuit behavioral and the idea of multi-objective optimization,a general method of circuit optimization is proposed.Based on the method,the problem of difficulty in establishing nonlinear relationships between objective optimization parameters and variable parameters has been solved,and breaking the bottleneck of compromise between performance indicators.Then,taking the charge pump phase-locked loop as an example,the influence of voltage-controlled oscillator on phase-locked loop circuit is analyzed.And the performance of voltage-controlled oscillator is optimized by using the proposed method.Finally,by comparing the circuit performance,the optimized circuit using the proposed method reduces the phase noise by2.1dBc/Hz@1MHz and improves the maximum oscillation frequency by 0.093GHz@4V,and the tuning range is increased by 0.123 GHz.The phase noise of charge pump phaselocked loop is reduced by 2.21 dBC/Hz@1MHz and the bandwidth is increased by 0.029 GHz.