On Efficient Representations of Frequency Dependent Far-Field Information for Array Antennas

This paper investigates the number of required data points to accurately represent the vector far-field information over the radiation sphere and its frequency variation for array antennas. The method needs to be explicit in terms of the excitation of the elements. We compare representations of far-field data on the sphere with two different approaches to spherical wave expansion representations and apply a data-driven model-order reduction to find a small-size representation. As is well known, a spherical wave expansion strongly reduces the number of required data points. Here, we show that accounting for the phase center of the elements further strongly reduces the number of required coefficients in the tested cases. The Loewner framework datadriven model-order reduction shrinks the required far-field data by an additional factor of 8–14 per mode coefficients, depending on the acceptable level of error. This reduction works well for the two investigated cases.