On the Spatial-Wideband Effects in Millimeter-Wave Cell-Free Massive MIMO

Seyoung Ahn, Soohyeong Kim, Yongseok Kwon, Joohan Park, Jiseung Youn, Sunghyun ChoarXiv 2023 · DOI · arXiv

Abstract

Cell-Free (CF) Massive MIMO is a promising architecture for millimeter-wave (mmWave) communications, offering improved coverage and reliability through distributed access points (APs). However, the large physical separation between APs introduces significant timing delays that, at high bandwidths, lead to spatial-wideband effects. This paper investigates these effects, characterizing the resulting inter-symbol interference (ISI) and beam squint. We derive a mathematical framework for the uplink channel and provide a closed-form expression for the minimum cyclic prefix (CP) length required to maintain system performance in these distributed environments.

Introduction

Traditional cellular networks rely on centralized base stations. In contrast, Cell-Free Massive MIMO distributes antennas across a geographic area as independent APs, all coordinated by a Central Processing Unit (CPU). While this architecture effectively mitigates blockages common in mmWave bands, it introduces a critical synchronization challenge. Because APs are located at varying distances from a User Equipment (UE), signals arrive with significant relative delays. In high-bandwidth mmWave systems, these delays are no longer negligible compared to the symbol period, giving rise to "spatial-wideband effects."

System Model and Spatial-Wideband Effects

The study develops a rigorous mathematical model for uplink spatial-time and spatial-frequency channels. A key contribution is the explicit modeling of time delay as a combination of two components:

  1. Macro-delay: The propagation time determined by the large-scale distance between the UE and a specific AP.
  2. Micro-delay: The relative delay between individual antenna elements within a single AP's array, typically associated with the angle of arrival.
Figure 1: Illustration of a Cell-Free Massive MIMO system showing the distributed architecture and the physical paths leading to macro and micro delays.

These delays manifest as frequency-dependent phase shifts in the Orthogonal Frequency-Division Multiplexing (OFDM) domain. In CF-mMIMO, beam-squint is particularly complex because both the "macro-steering vector" (related to AP positions) and the "phase-shift matrix" (related to antenna spacing) are frequency-dependent. If not properly addressed, these factors cause inter-symbol interference (ISI) and beam squint, where the effective beam direction shifts across the signal bandwidth.

Methodology

The authors utilize the 3GPP TR 38.901 UMi-Street Canyon NLOS channel model to simulate realistic urban environments. The analysis involves:

  • Channel Decomposition: Decomposing the spatial correlation matrix into micro-correlation (intra-AP) and macro-correlation (inter-AP) components.
  • Virtual Angle Domain Analysis: Using Discrete Fourier Transforms (DFT) to analyze beam-squint effects in the angular domain.
  • OFDM Characterization: Determining how the frequency-phase domain impacts signal integrity across the wide bandwidths typical of mmWave.

Key Results: Cyclic Prefix Requirements

A primary finding of the research is the derivation of the minimum Cyclic Prefix (CP) length ($CP_{min}$) necessary to mitigate ISI in a cell-free environment. The authors show that the CP must be sized to accommodate the maximum delay spread across all serving APs:

$$CP_{min} \approx W \frac{d_{k,L-1} - d_{k,0}}{c}$$

where $W$ is the bandwidth, $c$ is the speed of light, and $d_{k,L-1} - d_{k,0}$ represents the distance difference between the furthest ($L-1$) and nearest ($0$) access points serving user $k$. This formula highlights that in CF-mMIMO, the "array" is effectively kilometers wide, necessitating a much more robust CP than traditional co-located MIMO systems.

Conclusion and Future Work

This work provides a foundational roadmap for understanding spatial-wideband effects in distributed mmWave networks. While it establishes the physical requirements for synchronization, several challenges remain. Future research should address the computational overhead at the CPU for managing these frequency-dependent correlations and develop beamforming algorithms that can "pre-compensate" for these effects at the AP level. Additionally, the impact of timing jitter at the reference AP warrants further investigation to move beyond the assumption of perfect initial synchronization.

Summary updated

BibTeX
@misc{ahn2023spatial,
  author = {Seyoung Ahn and Soohyeong Kim and Yongseok Kwon and Joohan Park and Jiseung Youn and Sunghyun Cho},
  title = {{On the Spatial-Wideband Effects in Millimeter-Wave Cell-Free Massive MIMO}},
  year = {2023},
  eprint = {2307.02784},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2307.02784},
}
Figure