EM-driven tuning of substrate integrated waveguide filters exploiting feature-space surrogates

A fast and accurate tuning technique for substrate integrated waveguide microwave filters is proposed. A set of filter response features is identified and used to determine satisfaction/violation of given performance requirements. A response surface model of the features is constructed using the feature gradients estimated through coarse-discretization EM simulations. The model serves as a prediction tool, guiding towards an optimum design. Due to almost linear dependence of the response features on designable parameters, the optimum design can be obtained at the cost of a few high-fidelity EM-model evaluations. Validation example is provided.