Space mapping with co-simulation coarse model for accurate modeling of microwave structures

Abstract

Accurate modeling of microwave devices is important in many areas including signal processing, wireless communication and biomedical engineering. Although high accuracy can be obtained using full-wave EM simulation, CPU cost of EM-based models is usually unacceptable, particularly for direct statistical analysis and parametric optimization. Approximation techniques, e.g., kriging [1] and neural networks [2] are fast but can only ensure reasonable accuracy if a large number of training points is involved. Space mapping (SM) [3] allows an engineer to create fast ad hoc surrogate models suitable for design optimization and statistical analysis. SM surrogate is constructed using a simplified representation (coarse model) of the structure under consideration (fine model) and auxiliary mappings that “reshape” the parameter space of the coarse model. Accuracy of SM is limited by the quality of the coarse model, typically an equivalent circuit.