Abstract
This paper investigates the performance of two distinct joint antenna selection and precoding techniques for downlink multi-user massive multiple-input single-output (MISO) systems. We examine systems constrained by limited dynamic range power amplifiers, comparing a standard $\ell_1$-norm regularized precoder against a proposed thresholded $\ell_1$-norm precoder. To evaluate these methods, we utilize a novel extension of the Convex Gaussian Min-Max Theorem (cGMT) that accounts for non-linear post-processing operations. Our analysis demonstrates that the thresholded $\ell_1$-norm precoder provides superior signal-to-noise and distortion ratios (SINAD) and lower bit error rates (BER) compared to the one-shot $\ell_1$ approach. Theoretical asymptotic predictions are validated through Monte Carlo simulations, showing high accuracy for large-scale antenna arrays.
1. Introduction
In the evolution toward 6G wireless communication, Massive MISO technology stands as a primary pillar for increasing spectral efficiency. By deploying hundreds of antennas at a base station, it is possible to serve multiple users simultaneously with high precision. However, the practical implementation of such systems is hindered by the high cost and power consumption associated with dedicated Radio Frequency (RF) chains for every antenna.
Antenna selection offers a solution by activating only a subset of available antennas. While finding the optimal subset is an NP-hard combinatorial problem, sparse optimization techniques—specifically $\ell_1$-norm regularization—have emerged as a computationally feasible alternative. This study analyzes the trade-offs between signal fidelity and hardware efficiency in these sparse systems, particularly under the constraints of real-world power amplifiers.
2. Precoding Methodologies
We analyze two primary strategies for achieving sparsity in the precoding vector, which directly corresponds to antenna selection:
2.1 The $\ell_1$-norm Precoder
This method integrates an $\ell_1$-norm penalty directly into the precoding optimization problem. The penalty encourages the optimizer to produce a sparse solution where several antenna weights are exactly zero. While mathematically elegant as a "one-shot" solution, this approach acts as a blunt instrument: to achieve high levels of sparsity, the regularization parameter must be increased significantly, which often leads to substantial signal distortion and a reduction in the received SINAD.
2.2 The Thresholded $\ell_1$-norm Precoder
To address the distortion caused by heavy regularization, we propose a two-step process. First, the system solves the precoding optimization with a light $\ell_1$ penalty to maintain high signal quality. Second, an entry-wise thresholding operation is applied to the resulting vector, zeroing out the weakest entries to meet the specific RF chain budget. This "scalpel-like" approach allows for high sparsity without the collateral damage to signal integrity seen in the pure $\ell_1$ method.
3. Theoretical Framework: Extending the GMT
A central contribution of this work is the development of a rigorous mathematical framework to predict system performance. While the Convex Gaussian Min-Max Theorem (cGMT) is a standard tool for analyzing high-dimensional optimization, it is typically limited to the direct output of an optimizer.
Analyzing the thresholded precoder is more complex because thresholding is a non-linear post-processing step. We developed a novel GMT framework that tracks the statistical distribution of the signal after this non-linear operation. This allows for the precise calculation of asymptotic performance metrics—such as SINAD and BER—as the number of antennas ($n$) and users ($m$) grow large at a fixed ratio.
4. Performance Analysis and Results
The theoretical analysis was validated using synthetic Rayleigh fading channels and Monte Carlo simulations. For a system configuration of $n=256$ antennas and $m=128$ users, the following insights were observed:
- Asymptotic Accuracy: The mathematical predictions derived from our novel GMT framework perfectly matched the empirical simulation results, confirming the robustness of the theoretical model.
- Superiority of Thresholding: The thresholded $\ell_1$ precoder consistently outperformed the standard $\ell_1$ approach. For a fixed number of active RF chains, the thresholded version maintained a significantly higher SINAD, leading to a lower BER.
- Optimization of Parameters: Our analysis provides a deterministic method for selecting the optimal threshold parameter ($t_x$). This allows engineers to balance the number of active antennas against the resulting signal distortion without relying on exhaustive trial-and-error simulations.
5. Conclusion and Future Work
This work demonstrates that thresholding is a more efficient mechanism for antenna selection in massive MISO systems than pure sparse regularization. By providing a novel mathematical framework to analyze post-processing non-linearities, we enable more precise design of energy-efficient 6G hardware.
Future research will extend this analysis to more complex scenarios, including correlated fading channels and higher-order modulation schemes like 64-QAM. Additionally, incorporating more detailed hardware models—such as the power consumption of the switching networks—will further bridge the gap between theoretical analysis and practical deployment.
Summary updated