| 英文摘要 |
A two-phase optimization approach is proposed for the waveguide filter design. In the RF front-end design, a high pass filter with very sharp cutoff is commonly employed to suppress image frequency. Conventional filter design is optimized only at the center frequency so that use of the design approach makes it difficult to achieve the desired wide-band performance. In the past, the mode-matching technique had been employed for analysis and the descent method for optimization. The full-wave analysis, though, is computationally intensive. In addition, use of the descent method requires an initial educated guess and usually ends with a local minimum. In order to avoid the dependence on a starting point or the need for gradient information in optimization process, stochastic methods, such as genetic algorithm(GA) , are more suitable. To begin, the waveguide filter is modeled as cascaded transmission lines and then GA optimizes the H-plane dimensions of wavguide. The E-plane dimensions of the waveguide sections are assumed to be in a linear variation. The best linear coefficient is determined from full wave analysis with effects of step junctions that are rigorously included. The approach is applied to design a high pass waveguide filter at W-band and the numerical results are verified by measurement. The numerical and measurement data are consistent with the simulation results obtained by the commercial software Ansoft HFSS. |
本站僅提供期刊文獻檢索。
