Computational Optimizations towards an Accurate and Rapid Electromagnetic Emis- Sion Modeling

In this paper, a parametric electromagnetic radiated emission model has been explored. Several mathematical improvements with respect to its extraction and computational performance have been deployed. The model, represented with an array of radiating electric dipoles, predicts the electromagnetic emission of components and systems. Core-level changes have been made in order to extract the model parameters: the dipole positions, their orientations and currents, and the effective relative permittivity from near-field measurements. Thresholding and windowing techniques are used to detect and optimize dipole positions, directly from the field data. A fast and memory efficient two-level optimization algorithm based on the LevenbergMarquardt non-linear least squares technique is implemented for parametric extraction. All the constraints of the previous model have been overcome and the system is validated for mono-substrate and multisubstrate devices from measurements and/or simulations, with promising results. A tremendous improvement in modeling capability and performance has been obtained when compared with that of its erstwhile counterpart.