Research On Filter Network Transformation Method Based On Inverse Eigenvalue Problem

One of the bottlenecks of high-performance 5G network coverage is how to produce and adjust base station filters on a large scale and at a low cost,the key problem is how to quickly deduce filter networks with specific topologies(application requirements),which are mathematically represented as similar transformation problems in which coupling matrices meet specific constraints.Since the similar transformation paths of many topologies cannot be obtained,optimization methods become an important method for the synthesis and debugging of arbitrary filter networks.Firstly,this paper reviews how to extract the N+2 initial coupling matrix in the dual-port filter network,describes the research problem of dual-port mathematically,introduces the optimization model of dual-port filter network based on the inverse eigenvalue problem,and reproduces the fast and efficient Levenberg-Marguardt(LM)algorithm.Based on the analysis of the experimental results,the existing problems in LM optimization algorithm are summarized,such as sensitive initial value,limited order of application and how to select parameters.Secondly,aiming at the optimization problem of three-port filter network,an optimization model based on inverse eigenvalue problem is proposed for the first time,and the effectiveness of the loss function in the model is proved theoretically,and a variety of loss functions are solved by Quasi-Newton algorithm and LM optimization algorithm respectively,according to the experimental results,the two optimization algorithms can get the ideal topology structure,the best loss function of the two optimization algorithms is analyzed,and the conclusion is given at last.