Field measurements for the validation of antennas are, in current practice, obtained in the near-field region (nearer to the antenna than where the ultimate angular intensity pattern is apparent) to improve measurement quality and reduce cost. The large-scale linear transformation between the near-field and far-field patterns is typically accomplished using the fast Fourier transform (FFT). The restriction the FFT places on the measurement locations (to a regular planar grid), however, is often not realized in practice. As measurement frequencies increase and mobile measurement platforms with less stable scanners are increasingly employed, this problem becomes more severe.
This project was formed with the goal of creating a practical, accurate method for computing the near-field to far-field transformation for data taken at non-ideal measurement locations. The investigators have designed and implemented an algorithm that meets this goal. It combines the recently-developed nonequispaced fast Fourier transform, interpolation, and conjugate gradient iteration to achieve any prescribed accuracy in nearly optimal computational complexity. The method requires order operations per iteration, where n is the number of measurements; in current practice, n is typically between and .
This work has generated interest among researchers working with other measurements, including plane polar data, and further work will address mathematical issues that arise in related application areas.