Compressive Sensing Applied to Production Testing of Array Antennas using a Robotic Arm and Very Sparsely Sampled Near-Field Measurements

Abstract

Compressive Sensing (CS) has been deployed in a variety of fields including wideband spectrum sensing, active user detection and antenna arrays. In massive MIMO arrays, CS has been applied to reduce the number of measurements required to verify the arrays excitation in a production environment. All follow the general approach of creating the sparsity needed for CS by subtracting the measured far-field or near-field of the test array from that of a ‘gold standard’ array measured under identical conditions. In a previous paper [1] the authors have applied CS to planar near-field (PNF) measurements offering a compact test facility well suited to the production environment for these antennas. In that paper the reconstruction of array excitation with a mean square error (MSE) of -30dB was achieved for a 20 × 28 element array antenna at half wavelength spacing using just 1.5% (177 samples) of the samples needed for a conventional NF measurement (12,100 samples) employing classical back projection to the aperture. Critical to the performance is the realization that the CS samples need to be confined to the central region of the NF measurement plane which for a conventional NF to FF planar antenna pattern measurement would offer a massive truncation error. In this paper we address the optimal sampling strategy needed for this NF approach to diagnose arrays with up to a 4% failure rate by employing a statistical performance analysis of the reconstruction accuracy. Previous publications concerning CS based array diagnostics have exclusively studied the reconstructed array element amplitude, in this work we consider both array element amplitude and phase reconstruction performance that is critical in applying the technique to a production environment.