Measurement-Based Statistical Modeling of Spatially Selective ELAA Channel in UMi Scenario

Abstract

In recent years, extremely large antenna arrays (ELAAs) have garnered significant attention in both academia and industry. However, channel modeling efforts for ELAA in outdoor scenarios with long-range propagation remain limited. As the array aperture increases, spatial selectivity becomes increasingly pronounced, making it difficult for conventional approaches that uniformly distribute power across multipath components (MPCs) within each cluster to accurately capture the true spatial variations in the channel. To analyze channel characteristics and establish an accurate channel model based on real measurement data, this article proposes a time-efficient frequency-domain channel sounder for ELAA, which combines switch-based and virtual array methods with radio-over-fiber (RoF) to enable long-range measurements. Extensive channel measurements were conducted in an urban microcell (UMi) scenario using a 32-element Tx linear antenna array and a $32 \times 32$ -element Rx planar antenna array at the carrier frequency of 10 GHz. Significant power variations across the antenna array are observed in the measurement data, and to accurately model the constructive and destructive addition of MPCs, a new parameter referred to as the intracluster K factor is introduced in this article. The optimized searching strategy-aided space-alternating generalized expectation-maximization (OSS-SAGE) algorithm, under the spherical-wave assumption, is proposed and utilized to extract the near-field characteristics. By exploiting the orthogonality among multiple parameter domains, the algorithm compresses the data dimensionality, thereby reducing computational complexity arising from the increased number of resolvable MPCs enabled by the high spatial resolution of ELAAs. The large-scale channel characteristics are presented, such as path-loss and composite parameter spreads. A clustering algorithm is exploited to group the extracted MPCs with similar parameters. Finally, a measurement-based statistical channel model for ELAA in the UMi scenario has been established. This work provides an accurate channel model considering near-field and spatial selectivity for ELAA in UMi scenario.