Cubic B-Splines for Hardware-Accelerated Antenna Beampattern Interpolation

For applications that require fast evaluations of an antenna's beampattern, we propose a computationally efficient interpolation technique based on cubic B-Splines. We expand the pattern into a sum of finite-support polynomial functions that allow a fixed runtime evaluation irrespective of the antenna's aperture size. Further, we show how B-Splines can be implemented efficiently on GPU hardware leveraging so-called texture memory. We outline how oversampling, i.e. decreasing the distance between adjacent B-Splines, can approximate ideal low-pass filtering. Comparison to the Effective Aperture Distribution on measured antenna calibration data shows a 10-fold speedup in runtime while attaining an accuracy of -100 dB. When used for direction estimation using antenna arrays, we show that the proposed B-Spline interpolation yields a statistically efficient estimator, as it attains the Cramer-Rao Lower Bound.