Sequential HW-Aware Precoding: Over-the-air cancellation of HWI in Downlink Cell-Free Massive MIMO with Serial Fronthaul

Antoine Durant, Asma Mabrouk, and Rafik Zayani · 2025

Read Sequential HW-Aware Precoding: Over-the-air cancellation of HWI in Downlink Cell-Free Massive MIMO with Serial Fronthaul by Antoine Durant, Asma Mabrouk, and Rafik Zayani (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-2503-17562.txt.

Introduction

Antoine Durant, Asma Mabrouk, and Rafik Zayani tackle hardware flaws in a 2025 wireless paper. In cell-free networks, many small access points serve users together, which promises strong coverage. Real hardware distorts signals, though, and fixing that centrally can cost too much computing as the network grows.

Their method works step by step along a chain of access points linked in series. Each point learns what distortion earlier points added and sends a signal that cancels it in the air. A closed-form solution sets each point's weights on site. Under heavy distortion the method raised data efficiency by a large factor, while cutting computing sharply against two rival methods. It also worked well on cheap points with few antennas.

Catalog record

CREATORAntoine Durant, Asma Mabrouk, and Rafik Zayani
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE61 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-17562

Browse the full library · About the Tower of Babel

Full text

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

Sequential HW-Aware Precoding: Over-the-air cancellation of HWI in Downlink Cell-Free Massive MIMO with Serial Fronthaul Antoine Durant, Asma Mabrouk, and Rafik Zayani Abstract This paper addresses the critical challenge of mitigating hardware impairments (HWIs) in downlink cell-free massive MIMO (CF-mMIMO) networks while ensuring computational scalability. We propose a novel sequential hardware-aware (HW-aware) precoding technique that leverages the serial fronthaul topology to perform over-the-air HWI cancellation. This approach involves sequentially exchanging approximated user-perceived distortion information among successive access points (APs) for over-the-air HWI mitigation. Each AP independently computes its spatial multiplexing weights and transmits signals that counteract the distortions introduced by the preceding AP. We develop a problem formulation and present a closed-form solution for this method. For performance evaluation, we study two reference methods taken either from centralized massive MIMO literature (Tone Reservation [TR]) or tailored for CF-mMIMO networks (PAPR-aware precoding), both focusing on reducing the PAPR of OFDM signals in the downlink. Results indicate that the sequential HW-aware approach achieves a substantial increase in spectral efficiency (SE) in high-distortion scenarios, with an average SE increase factor of 1.8 under severe distortions. Additionally, the proposed method, which is executed locally, demonstrates better scalability, achieving a reduction of up to 40% and 72% in the total number of complex multiplications compared to the PAPR-aware and TR approaches, respectively. Finally, the sequential HW-aware precoder offers high performance even when applied to cost-effective APs with few antennas, presenting a promising and practical solution for HWI compensation in CF-mMIMO systems with serial fronthaul. I Introduction † † footnotetext: Antoine Durant and Asma Mabrouk are with the CEA-Leti in Grenboble, France (emails:[email protected] and [email protected]). Rafik Zayani is with the IETR laboratory at the University of Rennes, France (email: [email protected]. The growing demand for wireless data has driven the evolution of communication systems towards higher capacity and energy efficiency. Conventional approaches such as cell-densification and co-located massive MIMO (multiple-input multiple output) are known to suffer from inherent limitations, including severe inter-cell interference and limited macro-diversity, which restrict their potential to scale efficiently [ 1 ] . To address these challenges, a new paradigm termed Cell-Free massive MIMO (CF-mMIMO) has emerged, which has garnered significant attention in recent years [ 2 ] . Unlike traditional cellular networks, CF-mMIMO eliminates cell boundaries, with users being jointly served by a multitude of spatially distributed APs, providing enhanced spatial diversity and interference management. This distributed architecture offers numerous advantages, such as improved spectral efficiency and uniform service quality across the network, regardless of a user’s location. However, CF-mMIMO comes with its own set of challenges. The need to share large amounts of data, including user information, channel state information (CSI), and multi-user precoding vectors among the APs, introduces significant demands on the backhaul and fronthaul links. The sheer scale of data exchange can create scalability bottlenecks, as maintaining full cooperation among all APs is not feasible for large networks. Various strategies have been proposed to mitigate these scalability issues. For instance, local precoding schemes, where the multi-user precoders are computed locally at each AP without the need for extensive CSI exchange, have been shown to reduce backhaul traffic while maintaining a limited performance degradation compared to full cooperation scenarios [ 3 , 4 ] . Additionally, innovative network topologies such as serial fronthaul architectures and radio stripes have been proposed to reduce the complexity of inter-AP communication. In such setups, advanced precoding strategies like compute-and-forward or Team-MMSE precoding, where cooperation is limited to adjacent APs, have demonstrated promising results in improving system performance with reduced signaling overhead [ 5 , 6 ] . However, these works consider perfect hardware in the APs and are based on a network-level power constraint rather than per-AP. Hence, their computational complexity is proportional to the number of AP in the network, which leads to scalability issues . A further challenge in realizing large-scale CF-mMIMO systems lies in the hardware implementation. Given the extensive number of radio-frequency (RF) elements required, the use of low-cost and energy-efficient components is essential for economic viability. However, these cost-effective components typically introduce additional signal distortion, collectively referred to as hardware impairments (HWI). Such impairments have been extensively studied in the context of CF-mMIMO systems architectures [ 7 ] . Among the key contributors to HWI are distortions caused by the power amplifiers (PAs). The non-linearities in low-cost PAs, including non-linear compression and clipping, introduce significant additive distortion noise, which can severely degrade the system performance [ 8 ] . To mitigate these hardware-induced distortions, two critical factors can be tailored specifically to minimize the distortions when designing CF-mMIMO systems based on low-cost PAs. First, the peak-to-average power ratio (PAPR) can be reduced to enhance energy efficiency, as high PAPR forces PAs to operate inefficiently with larger power back-offs. Second, it is essential to reduce distortion caused by non-linear effects in the PA, such as compression and clipping, which can otherwise limit system capacity and spectral efficiency. Advances in signal processing techniques offer potential solutions, enabling more effective utilization of the available hardware, thus enhancing spectral efficiency while minimizing the impact of HWIs. Techniques such as digital predistortion [ 9 ] and PAPR reduction [ 10 , 11 ] can help optimize the performance of hardware-constrained systems and are crucial to offer low-cost, high performance CF-mMIMO systems. In [ 12 ] , a local PAPR-aware precoding scheme tailored for CF-mMIMO-OFDM systems is explored. While this approach achieves notable performance enhancements, it remains limited by its reliance on a high number of antenna, which can restrict practical implementation. In [ 13 ] , we introduced a team-based over the air HWI compensation solution. However, this method is developed specifically for a full-pilot zero-forcing (FZF) precoder, which is characterized by suboptimal performance in low input back-off (IBO) scenarios and incurs relatively high computational complexity. Post-distortion methods are a common approach to improving PA performance, particularly for amplifiers with memory effects [ 14 ] . Analog-based techniques involve designing auxiliary paths [ 15 ] or quasi-differential architectures [ 16 ] in the RF chain, increasing hardware cost and device size, key limitations for APs in CF-mMIMO. Digital-based techniques, on the other hand, are implemented at the receiver and thus restricted to uplink communications [ 17 ] . These methods also face computational challenges due to high-rank matrix solving, demanding high degrees of freedom for substantial performance gains [ 14 ] . In this paper, we propose an innovative over-the-air HWI compensation method, akin to digital post-distortion, implemented at the APs but in the downlink, leveraging the unique properties of CF-mMIMO with serial fronthaul. The use of a serial fronthaul architecture enables over-the-air cooperation in CF-mMIMO networks . Successive APs in the chain apply post-distortion to compensate for the HWIs of their predecessors, with compensation becoming effective at the users through coherent signal summation. This approach balances performance and flexibility. I-A Contribution Our key contributions are summarized as follows: • Development of a Sequential Hardware-Aware Precoding Scheme: We propose a novel sequential precoding technique based on a partial zero-forcing (PZF) precoder that takes into account the hardware impairments (HWI) present in low-cost RF chains. Our approach effectively mitigates distortion caused by non-linear PAs, leading to enhanced spectral efficiency. • Development of a closed-form solution

Read the complete text