Surrogate Model-Based Space Mapping in Postfabrication Bandpass Filters’ Tuning

Abstract

A procedure for diagnosis and tuning of fabricated bandpass filters is presented. A surrogate model consisting of a mapped coarse model at a basis point and a mapping of measured response with respect to the change of tuning elements from the basis point is established. To reduce the possibility of nonuniqueness, an implicit multipoint parameter-extraction technique is exploited to match both the response and the first-order derivative of response in terms of tuning parameters in the two models. With this approach, the robustness of the surrogate model and the iteration convergence are significantly improved. To verify the proposed method, two screw-tuned waveguide filters, including an ${X}$ -band filter fabricated by computerized numerical control (CNC) technology and a 3-D-printed metallic air-filled filter with large alignment tolerance and surface roughness, are given. To test the robustness of this approach, an FR-4 microstrip combline filter with uncertain dielectric properties is tuned by active varactors for different specifications utilizing the same surrogate model. All testing results achieve sufficient satisfaction to expectation.