Design optimization of interdigital filters using aggressive space mapping and decomposition

Abstract

This paper presents a new electromagnetic (EM) design methodology which combines two powerful techniques in a coherent strategy: space mapping (SM) and decomposition. An accurate but computationally intensive fine-resolution EM model is used sparingly only to calibrate a less accurate, but computationally much more efficient "coarse model." Applying this new approach to interdigital filter design, the authors exploit structural decomposition to construct a highly efficient coarse model using a combination of EM models with a coarse grid and empirical models for the noncritical substructures. The authors employ the aggressive SM optimization technique to obtain a rapidly improved design after each fine-model simulation while the bulk of the computation is carried out using the coarse model. To avoid possible oscillation in the iterative process, a penalty function is introduced. Fast and stable convergence to a desirable interdigital filter design is achieved after only three EM fine-model simulations.