Abstract
Previous studies have confirmed the adverse impact of fading correlation on the mutual information (MI) of two-dimensional (2D) multiple-input multiple-output (MIMO) systems. More recently, the trend is to enhance the system performance by exploiting the channel's degrees of freedom in the elevation, which necessitates the derivation and characterization of three-dimensional (3D) channels in the presence of spatial correlation. In this paper, an exact closed-form expression for the Spatial Correlation Function (SCF) is derived for 3D MIMO channels. This novel SCF is developed for a uniform linear array of antennas with nonisotropic antenna patterns. The proposed method resorts to the spherical harmonic expansion (SHE) of plane waves and the trigonometric expansion of Legendre and associated Legendre polynomials. The resulting expression depends on the underlying arbitrary angular distributions and antenna patterns through the Fourier Series (FS) coefficients of power azimuth and elevation spectrums. The novelty of the proposed method lies in the SCF being valid for any 3D propagation environment. The developed SCF determines the covariance matrices at the transmitter and the receiver that form the Kronecker channel model. In order to quantify the effects of correlation on the system performance, the information-theoretic deterministic equivalents of the MI for the Kronecker model are utilized in both mono-user and multi-user cases. Numerical results validate the proposed analytical expressions and elucidate the dependence of the system performance on azimuth and elevation angular spreads and antenna patterns. Some useful insights into the behaviour of MI as a function of downtilt angles are provided. The derived model will help evaluate the performance of correlated 3D MIMO channels in the future.
Introduction
MIMO technology is a cornerstone of modern wireless communications, relying on the multipath richness of the environment to achieve high data rates. However, spatial correlation—where signals at different antennas are too similar—can significantly degrade performance. Historically, engineers relied on 2D models focusing solely on the horizontal (azimuth) plane. With the advent of 5G and small-cell deployments, the vertical (elevation) dimension has become critical for beamforming and spatial multiplexing.
Existing 3D models often suffer from being too specific to certain signal distributions or requiring computationally intensive numerical integration. This paper addresses the need for a generalized, mathematically efficient model that accounts for arbitrary antenna patterns and complex 3D propagation environments.
Methodology
The authors derive an exact closed-form expression for the Spatial Correlation Function (SCF) using a sophisticated mathematical framework:
- Spherical Harmonic Expansion (SHE): Plane waves are represented using spherical harmonics to naturally capture 3D spatial properties.
- Legendre Polynomials: Trigonometric expansions of Legendre and associated Legendre polynomials are used to model the elevation components.
- Fourier Series (FS) Coefficients: This core innovation expresses the Power Azimuth Spectrum (PAS) and Power Elevation Spectrum (PES) as Fourier Series. This allows the model to handle any arbitrary angular distribution without assuming a specific statistical shape (e.g., Gaussian).
- Kronecker Model: The derived SCF builds covariance matrices for a Kronecker channel model, analyzed via Random Matrix Theory (RMT) to determine Mutual Information (MI).
Results and Validation
The proposed analytical model was validated against Monte-Carlo simulations using parameters from industry standards like 3GPP SCM and ITU-R M.2135.
Key findings include:
- High Accuracy: The closed-form expressions match Monte-Carlo results almost perfectly. The infinite series converges rapidly, with only 14 to 15 terms required to achieve an error rate below 0.5%.
- 2D Model Limitations: As shown in Figure 3, 2D models tend to overestimate spatial correlation. By ignoring the elevation dimension, they assume energy is more concentrated than it is in reality, leading to pessimistic performance predictions.
- Efficiency: The closed-form nature of the solution allows for rapid optimization of antenna parameters, such as downtilt, which previously required hours of simulation.
Discussion
The "plug-and-play" nature of the Fourier-based model makes it highly versatile for researchers. By measuring the Fourier coefficients of a specific environment's power spectrum, the correlation function can be immediately determined.
However, the authors note the "pinhole effect" where low correlation does not necessarily guarantee high capacity if physical paths are limited. While this model excels at characterizing correlation, parametric models are still useful for capturing such specific physical constraints. Additionally, the study highlights that antenna downtilt is a critical factor in multi-user scenarios, with optimal throughput achieved when the antenna boresight aligns with cell-edge users.
Conclusion
This work provides a robust and efficient tool for evaluating 3D MIMO channels. While currently focused on Uniform Linear Arrays (ULAs) and stationary environments, it lays the groundwork for extensions into Massive MIMO and non-stationary dynamic scenarios. Future research will likely explore more complex scattering environments where azimuth and elevation distributions are statistically dependent.
Summary updated