Fast microwave filter optimization using adjoint sensitivities and variable-fidelity electromagnetic simulations

Low-cost optimization of microwave filters using adjoint sensitivities is presented. In our approach, a set of electromagnetic (EM)-simulated models of increasing fidelities is sequentially optimized using a trust-region-based algorithm. Computational efficiency is obtained by executing the major search steps at the coarse-discretization EM simulation level. The technique is demonstrated using the fifth-order waveguide filter example.