Optimal Distortion-Aware Multi-User Power Allocation for Massive MIMO Networks

Siddarth Marwaha, Pawel Kryszkiewicz, and Eduard Jorswieck · 2025

Read Optimal Distortion-Aware Multi-User Power Allocation for Massive MIMO Networks by Siddarth Marwaha, Pawel Kryszkiewicz, and Eduard Jorswieck (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-06491.txt.

Introduction

Power amplifiers that distort signals are put at the center of a 2025 wireless engineering paper by Siddarth Marwaha and two co-authors. Massive MIMO networks use many antennas at once to serve many users. Real amplifiers clip signals, yet most power-sharing plans ignore that flaw and so miss the best performance.

Working with a soft-limiter amplifier model and fading radio channels, the authors derive a measure that counts both noise and distortion. A surprise follows: distortion creates a best operating point on its own, with no need for an added power cap. They split the hard optimization into two parts and alternate between them to find the optimum. Simulations show large sum-rate gains over plans that ignore distortion, especially with fewer antennas serving many users.

Catalog record

CREATORSiddarth Marwaha, Pawel Kryszkiewicz, and Eduard Jorswieck
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-06491

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Optimal Distortion-Aware Multi-User Power Allocation for Massive MIMO Networks Siddarth Marwaha, Pawel Kryszkiewicz, and Eduard Jorswieck Abstract Real-world wireless transmitter front-ends exhibit certain nonlinear behavior, e.g., signal clipping by a Power Amplifier (PA). Although many resource allocation solutions do not consider this for simplicity, it leads to inaccurate results or a reduced number of degrees of freedom, not achieving the global performance. In this work, we propose an optimal PA distortion-aware power allocation strategy in a downlink orthogonal frequency division multiplex (OFDM) based massive multiple-input multiple-output (M-MIMO) system. Assuming a soft-limiter PA model, where the transmission occurs under small-scale independent and identically distributed (i.i.d) Rayleigh fading channel, we derive the wideband signal-to-noise-and-distortion ratio (SNDR) and formulate the power allocation problem. Most interestingly, the distortion introduced by the PA leads to an SNDR-efficient operating point without explicit transmit power constraints. While the optimization problem is non-convex, we decouple it into a non-convex total power allocation problem and a convex power distribution problem among the users (UEs). We propose an alternating optimization algorithm to find the optimum solution. Our simulation results show significant sum-rate gains over existing distortion-neglecting solutions, e.g., a median 4 times increase and a median 50% increase for a 64-antenna and 512-antenna base station serving 60 users, respectively. I Introduction Massive multiple input multiple output (M-MIMO) has been shown to increase throughput and enable massive multiple access [ 1 ] , however, under the assumption that hundreds of parallel high-quality front-ends are employed. Not only does this require expensive hardware, but it also results in high energy consumption, which can be justified by considering a single Power Amplifier (PA) in the M-MIMO transmitter. Typically, a multi-carrier waveform is used for M-MIMO processing, however, this results in high variations of the instantaneous signal power [ 2 ] , measured typically by the Peak-to-Average-Power Ratio (PAPR). Unfortunately, as such a signal passes through a practical PA, i.e., a signal with limited maximal output power, some of its samples can be distorted, introducing signal distortion in each transmission chain. From the Signal-to-Distortion Ratio (SDR) perspective, even for the optimal PA, i.e., a soft-limiter that provides linear amplification till its clipping level, all the samples exceeding the clipping threshold will generate a non-linear distortion signal. Fortunately, the clipping problem can be reduced by means of certain transmitter-side signal processing, e.g., reduction of PA input signal variation [ 3 ] , or advanced reception algorithms [ 4 ] . Additionally, this effect can be reduced by optimizing the transmit power in relation to the PA clipping level. A higher signal-to-noise ratio (SNR) can be achieved by allocating higher transmit power, however, this results in significant signal clipping, thereby reducing the SDR. Therefore, considering the PA model, there is significant potential for wireless link optimization by distortion aware power allocation. A few solutions have been proposed to address this issue in Single Input Single Output (SISO) multi-carrier systems. In [ 5 ] , the authors focused on optimizing total transmit power in a cognitive radio system based on orthogonal frequency division multiplexing (OFDM), where the PA is modeled using a third-order memoryless polynomial. Moreover, equal power is allocated on each utilized sub-carrier. The authors assumed that the non-linear distortion power is equally distributed among sub-carriers, which is only approximately true as demonstrated in [ 6 ] . A similar approach has been applied to a cognitive raido system based on generalized frequency division multiplexing (GFDM) in [ 7 ] . Furthermore, in [ 8 ] , a similar third-order polynomial-based PA model—fitted to real-world measurements—was adopted to optimize transmission power for maximizing throughput in the uplink of a Non Orthogonal Multiple Access (NOMA) system. However, the third-order polynomial PA model is relatively simple and thus does not completely reflect the non-linear distortion power in the entire range of PA back-off values [ 6 ] . Typically maximum transmission power constraint is required. A more suitable option in this matter is a soft-limiter PA model. For a digital M-MIMO system, where each front-end is equipped with a PA and transmits multi-carrier waveform, a similar phenomenon as in the SISO system occurs. If the PA-distorted signal is decomposed into the desired signal and distortion signal, the key issue lies in how the distortion signals from multiple transmitting antennas are added at the reception point. Initially, it was assumed that non-linear distortions from multiple front-ends are uncorrelated, allowing them to be treated as additional white noise emitted from each transmitter antenna [ 9 ] . However, further studies revealed that this is not always the case. The generalized analytical framework for power spectral density (PSD) derivation for M-MIMO, combined with the discussion in [ 10 ] , show that in some cases, e.g., Line-of-Sight (LoS) channel, the non-linear distortion can be directed towards the intended receiver, preventing the SDR to increase with the number of antennas. Fortunately, as recently shown in [ 11 ] , this problem vanishes as the channel impulse response prolongs. Therefore, for small-scale Rayleigh fading channel, we assume uncorrelated noise-like distortion from each front-end. Until now, there is a limited set of power allocation solutions for M-MIMO systems. In [ 12 ] , the energy efficiency of an M-MIMO system was maximized considering hardware impairments. However, it was assumed that a narrowband signal is transmitted and that nonlinear distortion power is linearly dependent on the allocated power, similar to [ 9 ] . As discussed in [ 9 ] (Sec. VIIB) the distortion power proportional to the wanted signal power is just a local approximation in a linear range of the PA. These assumptions are far from practical applications of multi-carrier system or full utilization of the PA properties, e.g., operation close to the clipping region. In [ 13 ] , power allocation per antenna is optimized for a single user equipment (UE) considering the power consumed by PAs. Unfortunately, this does not consider OFDM signal transmission, or signal distortion on a PA. In [ 14 ] , power allocation in the downlink (DL) of multiple Distributed MIMO base stations (BSs) in a LoS environment under non-linear PAs is considered. Unfortunately, the study utilizes a particular and limited-order memoryless PA model that requires a total power consumption constraint in the optimization problem definition. Moreover, this study only considers the presence of a single UE in the system. In [ 15 ] , transmit power of a M-MIMO system under analog beamforming and soft limiter PA was optimized. In this setup, distortion is directed towards the intended UE, resulting in significant performance degradation. However, while the system serves a single UE at a time, it is not optimizing inter-UE power distribution. Finally, in [ 16 ] the authors consider power allocation and precoder design in multi-user DL MIMO transmission. Although the paper provides interesting results, the nonlinearity is modeled only by third-order intermodulation, which is a significant simplification [ 6 ] because higher-order intermodulations are not considered. Probably for this reason, the authors had to add a constraint on the total allocated power. Moreover, a single-carrier transmission, though of complex Gaussian signal sample distribution, is considered omitting all difficulties related to a typical scenario of multicarrier MIMO system design. In this work, the main contributions

Read the complete text