EM-ANN models for microstrip vias and interconnects in dataset circuits

摘要: A novel approach for accurate and efficient modeling of monolithic microwave/millimeter wave integrated circuit (MMIC) components by using electromagnetically trained artificial neural network (EM-ANN) software modules is presented. Full-wave EM analysis employed to characterize MMIC components. Structures simulation are chosen design experiments (DOE) methodology. EM-ANN models then physical parameters as inputs S-parameters outputs. Once trained, the inserted into a commercial microwave simulator where they provide results approaching accuracy tool used characterization MIMIC without increasing time significantly. The proposed technique capable providing do not exist or over desired region operation. has been verified developing microstrip vias interconnects in dataset circuits. new hybrid (/spl Delta/S) which makes use existing approximate introduced shown be more method models. An example optimize component geometry included.