This code package is related to the following scientific article:
Özgecan Özdogan and Emil Björnson, "Massive MIMO with Dual-Polarized Antennas," IEEE Transactions on Wireless Communications, vol. 22, no. 2, pp. 1448-1463, February 2023, doi: 10.1109/TWC.2022.3205471.
The package contains a simulation environment, based on Matlab, that reproduces some of the numerical results and figures in the article. We encourage you to also perform reproducible research!
This paper considers a single-cell massive MIMO (multiple-input multiple-output) system with dual-polarized antennas at both the base station and users. We study a channel model that includes the key practical aspects that arise when utilizing dual-polarization: channel cross-polar discrimination (XPD) and cross-polar correlations (XPC) at the transmitter and receiver. We derive the achievable uplink and downlink spectral efficiencies (SE) with and without successive interference cancellation (SIC) when using the linear minimum mean squared error (MMSE), zero-forcing (ZF), and maximum ratio (MR) combining/precoding schemes. The expressions depend on the statistical properties of the MMSE channel estimator obtained for the dual-polarized channel model. Closed-form uplink and downlink SE expressions for MR combining/precoding are derived. Using these expressions, we propose power-control algorithms that maximize the uplink and downlink sum SEs under uncorrelated fading but can be used to enhance performance also with correlated fading. We compare the SEs achieved in dual-polarized and uni-polarized setups numerically and evaluate the impact of XPD and XPC conditions. The simulations reveal that dual-polarized setups achieve 40-60% higher SEs and the gains remain also under severe XPD and XPC. Dual-polarized also systems benefit more from advanced signal processing that compensates for imperfections.
The article contains 9 simulation figures, numbered 1-9. Figure X is generated by the Matlab script mainFigX.m for X=1,...,9. The package also contains 16 Matlab functions that are used by some of the scripts. You can run the code in MATLAB online without a license by clicking on the link above. For the optimization problem, the convex programming solver CVX http://cvxr.com/cvx/ is used. See each file for further documentation.
This paper was supported by the Grant 2019-05068 from the Swedish Research Council.
This code package is licensed under the GPLv2 license. If you in any way use this code for research that results in publications, please cite our original article listed above.