Space Mapping Method Research And Its Application In Microwave Circuits

Optimization technologies have been used in electro-magnetic (EM) modeling andComputer aided design (CAD) for decades. Traditional optimization techniques exploitmainly two kinds of CAD tools: circuit-theory based tools and EM simulators.Circuit-theory based tools are fast but not accurate. Therefore, EM simulators, thoughCPU-intensive and time-consuming, are widely used due to their high fidelity. Servingto expedite the optimization process of EM designs, the space mapping (SM) techniquewas introduced in1994. By exploiting both kinds of tools, the SM method manages tocombine the speed of circuit models and the high accuracy of EM models, facilitatingthe optimization process of EM designs.In this thesis, novel SM techniques are proposed and their application tomicrowave circits is studied.First, an extension of the Aggressive Space Mapping (ASM) method has beenproposed as an amelioration of the ASM method, which we call Aggressive SpaceMapping based on Responses Error (ASMRE). This method replaces the original stopcriterion of the iteration of the ASM technique with the error between the fine modelresponse and the wanted response. A microstrip low-pass filter is taken as an example inthree situations. The results show that the ASMRE algorithm could effectively improvethe robustness and convergence of the ASM method.Then we propose a novel space mapping algorithm called space mapping based onthe particle swarm optimization (SMPSO), which combines the coarse model and theparticle swarm optimizer as the surrogate model. The optimizer carries out theparameter extraction (PE) evolutionarily. Correspondingly, the PE only performs simplearithmetic update operations rather than the heavy circuit model optimization. Thus thePE becomes much faster in the SMPSO algorithm. A microstrip hairpin-line bandpassfilter is employed to illustrate the SMPSO algorithm.At last, a Matlab-based program for the SM technique is introduced. We present the algorithm of the core control Matlab program, the technique and method used to callAnsoft HFSS and Agilent ADS in Matlab, and the construction of the platform.