Synthesis of microwave filters with non-resonating nodes using optimization

Abstract Filter coupling matrix extraction is a major issue for urgent solution in the research of high performance millimeter-wave filter based on packaging technology. For the Chebyshev prototype microwave filter equiripple transfer function, Cameron summarizes a very dedicated method for the polynomial synthesis. In this thesis, the filter transfer functions are based on the Chebyshev equiripple transfer function. Generally, two main topologies exist for microwave filter structure, one is the cross coupled resonators and the other is the extracted pole. Recently, it was shown that the introduction of non-resonating nodes (NRNs) in both of these models eliminates some of their inherent limitations. A NRN is simply a node that is connected to ground by a frequency- independent reactance, which may be an open circuit in certain cases. The total number of NRNs in a given filter is arbitrary since it does not affect the order of the filter. In extracted pole topology with NRNs, there exist an analytical way to obtain the coupling matrix. However, the analytical method cannot be applied in the cross coupling topology. Thus, the optimization by using genetic algorithm is proposed in this thesis. The genetic algorithm is introduced for the synthesis of microwave filter. In the last chapter, several examples are listed to demonstrate the availability of applying optimization in the filter parameter calculation. Using computer aided genetic algorithm optimization has both advantages and disadvantages. It is less complicated than the analytical way and appears flexible for different kinds of topologies. On the other side, sometimes, it is difficult to converge when encountering large dimensions of unknowns in the coupling matrix and the final results maybe not very accurate.