COMPRESSIVE SENSING RECONSTRUCTION OF WIDEBAND ANTENNA RADIATION CHARACTERISTICS

Characterization measurements of wideband antennas can be a time intensive and an expensive process as many data points are required in both the angular and frequency dimensions. Parallel compressive sensing is proposed to reconstruct the radiation-frequency patterns (RFP) of antennas from a sparse and random set of measurements. The modeled RFP of the dual-ridge horn, bicone, and Vivaldi antennas are used to analyze the minimum number of measurements needed for reconstruction, the difference in uniform versus non-uniform reconstruction, and the sparsity transform function used in the compressive sensing algorithm. The effect of additive white Gaussian noise (AWGN) on the minimum number of data points required for reconstruction is also studied. In a noise-free environment, the RFP of the antennas were adequately reconstructed using as little as 33% of the original data points. It was found that the RFPs were adequately reconstructed with less data points when the discrete cosine transforms (DCT), rather than the discrete Fourier transforms (DFT) was used in the compressive sensing algorithm. The presence of noise increases the number of data points required to reconstruct an RFP to a specified error tolerance, but the antenna RFPs can be reconstructed to within 1% root-mean-square-error of the original with a signal to noise ratio as low as −15 dB. The use of compressive sensing can thus lead to a new measurement methodology whereby a small subset of the total angular and frequency measurements is taken at random, and a full reconstruction of radiation and frequency behavior of the antenna is achieved during post-processing.