This paper describes an efficient and robust approach for the computer diagnosis and tuning of RF and microwave filters relying upon model-based parameter estimation (MBPE) and multilevel optimization. The frequency sampled S-parameters are obtained from the measurement, and then an MBPE procedure based on adaptive sampling is employed to approximate the frequency-domain behavior of S-parameters in terms of rational functions. This approach uses a reduced-order system model. The effect of measurement noise is also considered. The approach is applied to coupled resonator filters that are modeled by a general equivalent circuit. The loss of each resonator is included in the model by a series resistor. A simple and efficient error function is used to reduce the computational effort of the optimization while improving the speed and robustness of diagnosis process for lossy filters. This approach can be applied to many classes of filters. The proposed approach is demonstrated through numerical examples and application to the manufactured filter.
展开▼
