Model-Driven Compressive Spherical Array Sampling of Base-Station-Antenna Near Field Using Accelerated Pseudo-Skeleton Tensor Approximation

Spherical array sampling plays an important role in obtaining near-field samples (NFSs) of base-station antennas for wireless internet of things performance analysis and optimization. Due to the large sampling area that a semicircular array of probes (SAP) has to cover, spherical array sampling is often time-consuming. In order to reduce sampling time, this paper proposes a model-driven compressive spherical array sampling technique based on pseudo-skeleton tensor approximation. In the proposed method, the Huygens’ principle and the method of moments are applied to construct a mapping tensor firstly, which relates the vector consisting of Huygens’ equivalent source samples with the matrix constituted by spherical near-field samples. Then by analyzing the physical meaning of the mapping tensor, it is revealed that skeleton sampling positions of SAP could be obtained by searching skeleton slices of the mapping tensor. The skeletonization process provides the relationship between NFSs in skeleton and non-skeleton sampling positions. This means NFSs in non-skeleton sampling positions can be computed from those in skeleton sampling positions, and the sampling effort is thus reduced. Moreover, in order to efficiently search skeleton slices of the mapping tensor, a fast ordered-subset selection algorithm is developed. Numerical experiments show that the proposed sampling method can reduce the number of samples by 50%, and the proposed fast algorithm is 22 times faster than traditional algorithms from the perspective of computation efficiency. Since the proposed sampling method is model-driven, it is independent of sampling data, which makes the method attractive to multi-beam multiple-input multiple-output (MIMO) antennas.