Accelerated design optimization of miniaturized microwave passives by design reusing and Kriging interpolation surrogates

Electromagnetic (EM) analysis has become ubiquitous in the design of microwave components and systems. One of the reasons is the increasing topological complexity of the circuits. Their reliable evaluation—at least at the design closure stage—can no longer be carried out using analytical or equivalent network representations. This is especially pertinent to miniaturized structures, where considerable EM cross-coupling effects occurring in densely arranged layouts affect the performance in a non-negligible manner. Although mandatory, EM-driven design is normally associated with significant computational expenses. Consequently, expediting the procedures that require massive simulations, such as parametric optimization, is a practical necessity. In this paper, a framework for accelerated parameter tuning is proposed. The keystones of our methodology are a set of pre-existing designs optimized for various design objectives, as well as kriging interpolation surrogates. The latter are constructed to yield—for a given set of performance specifications—a reasonably good starting point and to enable rapid optimization by providing the initial approximation of the Jacobian matrix of the circuit outputs. The proposed approach is validated using two compact impedance matching transformers designed within the objective spaces defined by wide ranges of operating bandwidths. As demonstrated, the average tuning cost corresponds to a few EM simulations of the respective circuit despite large numbers of adjustable parameters.