Two-Stage Hybrid-Field Beam Training for Ultra-Massive MIMO Systems

Abstract

In this paper, by considering both the near field and far field, the beam training is investigated for an ultra-massive multiple-input-multiple-output system with a partially-connected hybrid combing structure. As the Rayleigh distance decreases quadratically with the reduction of the antenna number, the near-field effect of subarrays is much weaker than that of the whole array. Motivated by this, far-field channel steering vectors of a subarray are used to approximate its near-field channel steering vectors. Then a two-stage hybrid-field beam training scheme that works for both the near field and far field is proposed. In the first stage, each subarray independently uses multiple far-field channel steering vectors for analog combining. In the second stage, for each codeword in a predefined hybrid-field codebook, a dedicated digital combiner is designed to combine the output of the analog combiner from the first stage. Then, from the hybrid-field codebook, the codeword corresponding to the dedicated digital combiner that achieves the largest combining power is selected. Note that the dedicated digital combiners can be obtained offline before the beam training and the combining power can be computed in parallel. Simulation results show that the proposed scheme can approach the performance of the hybrid-field beam sweeping but with considerable reduction in training overhead.