Computationally efficient design closure of miniaturized impedance matching transformers using response features

In this article, we describe a procedure for reliable and computationally efficient design optimization of miniaturized impedance matching transformers. Our approach exploits a concept of feature-based optimization (FBO). According to FBO, considerable reduction of the computational cost of the simulation-driven design process can be achieved—compared to conventional methods—by reformulating given performance requirements (typically, minimization of reflection over a frequency range of interest) in terms of suitably defined response features. For impedance transformer circuits, the feature points are defined as local maxima of the reflection characteristic, as well as the points defining the −20 dB bandwidth. As the feature point coordinates (i.e., their frequencies and levels) depend on the geometry parameters of the structure in less nonlinear manner than the original responses (S-parameters versus frequency), the optimization algorithm exhibits faster convergence. Further reduction of the optimization cost is obtained by utilization of variable-fidelity electromagnetic simulations. Our technique is demonstrated using two design cases of an example miniaturized three-section 50-to-100 ohm microstrip transformer.