Research On Filter Design Method Based On Diagnosis Of Coupling Matrix

Microwave filters are important selected frequency devices in kinds of microwave systems. Currently, the structure of the filter becomes more and more complex, there are difficulties when use the existing methods to design these complex filters, which restricted the further development of the filter. Therefore, develop a design method for the filters with complex structure have a great practical significance. This study is centered on how to achieve rapid design for the complex filters, made in-depth studies on coupling matrix synthesis theory, diagnosis technique of the filter, the establishment of an equivalent circuit model of the filter, and the space mapping method based on neural network, propose a design method based on the diagnosis of the coupling matrix for the filters. The main contents of this thesis are as follows:Firstly, introduced the coupling matrix synthesis theory and the diagnosis technique of the filter. On the one hand, introduced the analytical method and optimization method used to obtain the coupling matrix, they are the basis of design coupled resonator filter. On the other hand, three kinds of analytical methods and an optimization method of the diagnosis technique are introduced, they can be used to extract the coupling matrix of the filter.Secondly, a method is proposed to establish the exact equivalent circuit model of the filter, this method is derived from the traditional equivalent circuit modeling method of the filter, added the FIR(frequency-invariant reactance) element. The proposed equivalent circuit model can be used to extract the parameters of the filter, ensure the accuracy of the parameter extraction. In this thesis, two examples are used to verify the proposed equivalent circuit model, prove the rationality of the established method.Thirdly, space mapping method based on neural network is proposed, this method can be used to design the filter, and it has significant advantages especially for designing the complex filters. Furthermore, this thesis proposes a new method of selecting sample points for how to select the sample point, and it can be used to select the training and testing sample points for training the neural network, ensure that the sample point selection is reasonable, and then able to get the correct mapping based on the selected sample points. In this thesis, an example is used to verify the reasonableness of the proposed method of selecting sample points.Finally, propose a design method based on the diagnosis of the coupling matrix for the filters, this method is the combination of diagnosis technique of the filter and space mapping method based on neural network, and it is used to design the filter. The proposed design method of the filter is based on the response curves of physical test, avoid the use of electromagnetic simulation software to optimize or verification, and there are big differences between this method and traditional design methods of the filter. Moreover, this method is a general design method, and it has significant advantages especially for designing the complex filters. In this thesis, an example of SIW(substrate integrated waveguide) filter is used to verify the proposed design method, demonstrate the effectiveness of the design method.