Adaptive windowed estimation of electromagnetic frequency spectra and radiation patterns to a specified uncertainty

Common to areas involving waves and fields like acoustics and electromagnetics is the need to evaluate frequency spectra and far-field radiation patterns. This has usually been done on a point-wise basis, sampling a first-principles or generating model (GM) finely enough to ensure that important features such as narrow resonances and nulls are not missed. This can lead to many more samples, and consequently greater computational cost, than is really needed. Described below is an approach that employs a fitting model (FM) derived from model-based parameter estimation together with multiple, overlapping frequency- or angle-observation windows. Comparing FM samples in the overlapping windows shows where additional GM samples are needed to achieve a desired uncertainty in the computed results while minimizing the number of GM samples required.