Modeling And Computer Tuning Strategy For Microwave Cavity Filter

Microwave filters have been widely used to separate spectrum information,improve communication quality and prevent signal crosstalk.Most microwave filters are designed by using electromagnetic simulation technology,the difference of tolerance and material characteristics makes the output frequency response inconsistent with the theoretical results.Therefore,the microwave filters after production must be tuned to meet the requirements,however,the tuning work is extremely complex.Simply relying on manual tuning not only has low tuning efficiency,but also has poor product consistency.To improve the tuning efficiency of microwave filters,an intelligent tuning method for cavity filters is proposed in this dissertation.The main work as follows.(1)Parameter extraction of microwave cavity filter under different modes.This dissertation presents a method of parameter extraction for microwave cavity filters in different modes.To extract admittance poles and residues of cavity filters,a improved vector fitting method is used to realized by considering the influence of phase shift,resonator loss and initial zero-poles on the accuracy of parameter extraction;When dealing with the inconsistent port phase and resonator loss of the filter,the inconsistent phase and attenuation factor are eliminated as optimization variables,and the loss is removed by introducing the real part into the admittance pole.When the filter detuning is large,the coupling matrix is extracted by directly constructing the relation between polynomial coefficients and admittance functions,which overcomes the application of lossless coupling matrix synthesis method in lossy coupling matrix synthesis.(2)Modeling for microwave cavity filter.This dissertation proposes a modeling method of cavity filters based on different relational data.For the admittance poles and residues extracted problem from low-order simulated filters,the relation model between the tuning height of resonant bar,coupling bar and admittance poles,residues is established by using Gauss kernel clustering and sub-model probabilistic fusion,which solves the problem of poor accuracy and low efficiency of single model;Facing the coupling matrix extracted from inconsistent port phase and resonant cavity loss,a relationship model between the elements in coupling matrix and the tuning height of screws based on distributed extreme learning machine,which solves the problems of parallel computation of the model and random selection of hidden layer nodes;Facing the multi-stage process tuning data extracted under severe detuning state,the relation model of the nine-order cross-coupled filter under severe detuning is established by combining the fuzzy C-means clustering algorithm with the T-S fuzzy neural network,and the effectiveness of the method is verified by experiments.(3)Computer tuning strategy for microwave cavity filter.This dissertation presents a tuning method based on different relational models(surrogate models).For low-order simulated filters,the parameter optimization of surrogate model under non-linear mapping is realized by using implicit space mapping based on poles and residues of Y-parameters.To tackle the problem of tuning of multivariable cavity filters,the tuning is realized by combining distributed extreme learning machine with space mapping algorithm,and the difficulty of convergence and dependence on initial value are avoided by setting convergence radius and update step size.For the tuning problem of cavity filters with complex topological structures under large detuning,the the phase-by-phase tuning of nine-order cross coupled filters is realized by combining multi-objective particle swarm algorithm with T-S fuzzy neural network space mapping.(4)Design of tuning system for microwave cavity filter.First,This dissertation realizes the visual human-machine interface and the storage and display of dynamic data by MATLAB GUI based on the basic idea of modular programming.The three dimensional modeling and acquisition of tuning data for cavity filter are completed by HFSS-MATLAB-API.The communication between vector network analyzer and PC and other equipments can be connected by communication program.Finally,the dynamic simulation of microwave filter is realized by calling parameter extraction algorithm and optimization strategy.In conclusion,the computer-aided tuning strategy for microwave cavity filters are proposed,and the methods of parameters extraction under different modes are developed.The different relation models are established by using electromagnetic simulation software and field data.The computer-aided tuning is realized by building the mapping relation between the surrogate model and the actual model,and the effectiveness of the proposed method is verified by simulation experiments.