| dc.description.abstract | Microwave filters and multiplexers are widely used in all types of electronic systems. This research focuses on advanced filters and multiplexers for applications in communication satellites with stringent requirements and specifications. Conventional modeling techniques are typically either based on fast but inaccurate circuit models, or based on full-wave electromagnetic (EM) simulation, which is accurate yet time-consuming. To overcome the limitations of conventional methods, this thesis presents two modeling techniques for microwave filters to enhance the efficiency and accuracy in both near-band and wide-band simulations.
In the first method, a hybrid model is constructed with an EM analyzed I/O coupling junction, which is close to the multiplexer manifold, and a circuit model representing the rest of a channel filter. Since only one junction is EM simulated, neural networks (NNs) can be applied in the hybrid models to facilitate efficient modeling of the filter over a large frequency range. As demonstrated with a temperature compensated Ku-band multiplexer and a high power dielectric resonator filter, both in-band response and out-of-band spurious mode behavior of channel filters are captured in the hybrid models.
In the second modeling method, the entire filter is segmented and each junction is EM simulated. By only preserving the necessary information of the generalized scattering matrix (GSM) of each critical junction of the filter, the data size of each junction is dramatically reduced, which makes training and testing of a NN model possible. All the well-trained NN models are connected through transmission lines. The resulting filter model consists of only NNs and transmission line models, and is therefore very fast. Conventional dual mode cylindrical waveguide filter and high-power dielectric resonator filters are used to demonstrate this method. The efficiency and accuracy of the modeling method are proved by comparing the simulation results of the proposed model to the results from full-wave simulation and measurements. A good agreement is observed for both near-band and wide-band results.
Both methods present in the thesis are proven to be more efficient than conventional circuit models or full-wave EM models. | en |
