arxiv: 2604.06355 · v1 · submitted 2026-04-07 · 📡 eess.SP · cs.AR

Interference Suppression for Massive MU-MIMO Long-Term Beamforming with Matrix Inversion Approximation

Amirreza Kiani , Ali Rasteh , Marco Mezzavilla , Sundeep Rangan This is my paper

Pith reviewed 2026-05-10 18:34 UTC · model grok-4.3

classification 📡 eess.SP cs.AR

keywords massive MU-MIMOlong-term beamforminginterference suppressionmatrix inversionconjugate gradientnumerical stabilitysubspace nulling5G scenarios

0 comments

The pith

Projecting out the dominant interference subspace from long-term statistics reduces eigenvalue spread and cuts conjugate gradient iterations for stable long-term beamforming in massive MU-MIMO.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper establishes that dominant interferers degrade the numerical conditioning of the covariance matrix in long-term beamforming, which relies on slowly varying spatial statistics rather than instantaneous channels. The proposed subspace nulling projects the received signal onto the orthogonal complement of this interference subspace estimated solely from long-term data. This acts as an implicit preconditioner that shrinks the eigenvalue spread and improves stability for approximate inversion methods such as conjugate gradient or polynomial expansions. Ray-tracing simulations in a realistic 5G scenario confirm that the approach achieves near-optimal performance with substantially fewer CG iterations in both floating-point and fixed-point arithmetic while keeping the low-overhead advantage of long-term beamforming.

Core claim

The central claim is that projecting the received signal onto the orthogonal complement of the dominant interference subspace estimated from long-term channel statistics reduces the eigenvalue spread of the LTBF covariance matrix and serves as an implicit preconditioning step. This improves numerical stability for matrix inversion approximations. In ray-tracing simulations of a realistic 5G scenario the method substantially reduces the number of conjugate gradient iterations needed to reach near-optimal performance across floating-point and fixed-point implementations without increasing overhead.

What carries the argument

Subspace nulling projection onto the orthogonal complement of the dominant interference subspace estimated from long-term channel statistics, which functions as implicit preconditioning for the covariance matrix.

Load-bearing premise

The dominant interference subspace can be estimated accurately enough from long-term channel statistics alone that removing it does not significantly degrade the desired user signals or overall beamforming performance.

What would settle it

A ray-tracing or real-world test in which applying the subspace projection produces lower sum-rate or higher error rates than standard LTBF without projection, especially when the long-term interference subspace estimate deviates from the actual dominant directions.

Figures

Figures reproduced from arXiv: 2604.06355 by Ali Rasteh, Amirreza Kiani, Marco Mezzavilla, Sundeep Rangan.

**Figure 2.** Figure 2: Average capacity versus number of conjugate gradient [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗

**Figure 3.** Figure 3: 10th percentile capacity versus number of CG itera [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗

read the original abstract

Long-term beamforming (LTBF) is a widely-used scalable alternative to instantaneous multi-user MIMO processing that leverages slowly varying spatial channel statistics. VLSI implementations require matrix inversion that become computationally challenging for massive MIMO systems with large number of antennas. In this work, we show that dominant interferers significantly degrade the numerical conditioning of the LTBF covariance matrix, leading to severe performance loss in finite-precision implementations of polynomial and conjugate gradient (CG) based inversion methods. To address this issue, we propose a subspace nulling approach that operates solely on long-term channel statistics and acts as an implicit preconditioning step for LTBF. By projecting the received signal onto the orthogonal complement of the dominant interference subspace, the proposed method reduces the eigenvalue spread of the covariance matrix and improves numerical stability. Through ray-tracing simulations in a realistic 5G scenario, we demonstrate that the proposed method substantially reduces the number of CG iterations required to achieve near-optimal performance across floating-point and fixed-point implementations while preserving the low-overhead nature of LTBF.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The subspace nulling step improves LTBF matrix conditioning and cuts CG iterations in the reported sims, but the gains rest on untested assumptions about subspace estimation from long-term stats.

read the letter

The core idea is to project the received signal onto the orthogonal complement of the dominant interference subspace, estimated from long-term channel covariance, before doing the LTBF matrix inversion. This shrinks the eigenvalue spread and lets conjugate gradient converge in fewer steps, which the authors test in ray-tracing runs for both floating-point and fixed-point arithmetic in a realistic 5G layout. The approach stays inside the low-overhead LTBF framework and directly targets a VLSI pain point with dominant interferers.

Referee Report

2 major / 2 minor

Summary. The paper proposes a subspace nulling technique for long-term beamforming (LTBF) in massive MU-MIMO systems. Dominant interferers are shown to degrade the conditioning of the LTBF covariance matrix, impairing polynomial and conjugate-gradient (CG) inversion approximations in finite precision. The method estimates the dominant interference subspace from long-term channel statistics alone and projects the received signal onto its orthogonal complement, reducing eigenvalue spread as an implicit preconditioner. Ray-tracing simulations in a realistic 5G scenario are used to claim substantially fewer CG iterations for near-optimal performance in both floating- and fixed-point arithmetic while preserving LTBF's low overhead.

Significance. If the simulation results hold under realistic subspace estimation, the approach could provide a practical, statistics-only enhancement to numerical stability in VLSI implementations of LTBF, lowering iteration counts without added pilot or feedback overhead. This would be relevant for scalable massive MIMO where instantaneous CSI is infeasible.

major comments (2)

[§4] §4 (Proposed subspace nulling): the central claim that projection onto the orthogonal complement of the dominant interference subspace (extracted solely from long-term covariance) reliably shrinks eigenvalue spread without material signal loss for desired users rests on an unverified assumption of negligible overlap and accurate subspace estimation. No analysis or bounds are given for finite-sample estimation error or slow channel evolution, which directly affects whether CG iteration reduction is achieved in practice.
[§5] §5 (Ray-tracing simulations): the reported reduction in CG iterations for near-optimal performance lacks quantitative baselines (standard LTBF without projection), exact iteration counts, error bars across channel realizations, and details on fixed-point word-length effects or subspace estimation sample size. Without these, the performance claims cannot be verified and the robustness concern remains unaddressed.

minor comments (2)

[Abstract] Abstract: the phrase 'substantially reduces the number of CG iterations' should be accompanied by concrete numbers or ranges drawn from the simulation results.
Notation: ensure consistent use of symbols for long-term covariance matrices and subspace projectors across sections to avoid reader confusion.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments. We address each major comment below and will incorporate revisions to strengthen the manuscript.

read point-by-point responses

Referee: [§4] §4 (Proposed subspace nulling): the central claim that projection onto the orthogonal complement of the dominant interference subspace (extracted solely from long-term covariance) reliably shrinks eigenvalue spread without material signal loss for desired users rests on an unverified assumption of negligible overlap and accurate subspace estimation. No analysis or bounds are given for finite-sample estimation error or slow channel evolution, which directly affects whether CG iteration reduction is achieved in practice.

Authors: We agree that a more explicit discussion of the assumptions would improve the paper. The approach relies on the practical observation that dominant interferers are typically angularly separated from desired users in massive MIMO deployments, allowing long-term covariance estimates to identify the interference subspace with limited overlap. Our ray-tracing results, based on extended observation periods for covariance estimation, demonstrate consistent CG iteration reductions. In the revision we will add a dedicated paragraph in §4 discussing robustness to finite-sample estimation error and slow channel evolution, including a sensitivity study varying the number of samples used to form the long-term covariances. revision: yes
Referee: [§5] §5 (Ray-tracing simulations): the reported reduction in CG iterations for near-optimal performance lacks quantitative baselines (standard LTBF without projection), exact iteration counts, error bars across channel realizations, and details on fixed-point word-length effects or subspace estimation sample size. Without these, the performance claims cannot be verified and the robustness concern remains unaddressed.

Authors: We accept that additional quantitative detail is required for verifiability. The revised §5 will include direct numerical comparisons against standard LTBF without subspace nulling, reporting the exact CG iteration counts needed to reach near-optimal performance, error bars computed over multiple independent channel realizations, the fixed-point word length employed, and the number of samples used to estimate each long-term covariance matrix. These changes will allow readers to assess the claimed iteration reductions and numerical stability improvements. revision: yes

Circularity Check

0 steps flagged

No circularity: proposal and simulation results are independent of inputs

full rationale

The paper introduces a subspace nulling projection operating on long-term covariance statistics to precondition the LTBF matrix inversion, then validates the reduction in CG iterations via ray-tracing simulations. No equations or claims reduce by construction to fitted parameters, self-definitions, or self-citation chains; the eigenvalue-spread improvement is presented as a direct consequence of the orthogonal projection step, which is independently motivated and tested against external benchmarks rather than tautologically derived from the same data or prior self-referential results.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that long-term statistics suffice to identify a dominant interference subspace whose nulling improves conditioning without harming desired signals; no free parameters or invented entities are introduced in the abstract.

axioms (1)

domain assumption Dominant interference subspace can be reliably estimated from long-term channel statistics
Invoked as the basis for the subspace nulling projection step.

pith-pipeline@v0.9.0 · 5486 in / 1360 out tokens · 98268 ms · 2026-05-10T18:34:12.711139+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/AlexanderDuality.lean alexander_duality_circle_linking unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Pv = I - Hv (Hv^H Hv)^-1 Hv^H

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

15 extracted references · 15 canonical work pages · 2 internal anchors

[1]

Massive mimo for next generation wireless systems,

E. G. Larsson, O. Edfors, F. Tufvesson, and T. L. Marzetta, “Massive mimo for next generation wireless systems,”IEEE communications magazine, vol. 52, no. 2, pp. 186–195, 2014

work page 2014
[2]

Massive MIMO evolution toward 3GPP release 18,

H. Jin, K. Liu, M. Zhang, L. Zhang, G. Lee, E. N. Farag, D. Zhu, E. Ong- gosanusi, M. Shafi, and H. Tataria, “Massive MIMO evolution toward 3GPP release 18,”IEEE Journal on Selected Areas in Communications, vol. 41, no. 6, pp. 1635–1654, 2023

work page 2023
[3]

Extreme Massive MIMO for Macro Cell Capacity Boost in 5G-Advanced and 6G,

H. V . Harri Holma and P. Mogensen, “Extreme Massive MIMO for Macro Cell Capacity Boost in 5G-Advanced and 6G,” Nokia, White Paper, 2025. [Online]. Available: https://www.nokia.com/asset/210786/

work page 2025
[4]

Joint Detection, Channel Estimation and Interference Nulling for Terrestrial-Satellite Downlink Co-Existence in the Upper Mid-Band,

S. Jia, M. Ying, M. Mezzavilla, D. Calin, T. S. Rappaport, and S. Ran- gan, “Joint Detection, Channel Estimation and Interference Nulling for Terrestrial-Satellite Downlink Co-Existence in the Upper Mid-Band,” arXiv preprint arXiv:2510.08824, 2025

work page arXiv 2025
[5]

Bandwidth Gain: The Missing Gain of Massive MIMO,

M. Akrout, V . Shyianov, F. Bellili, A. Mezghani, and R. W. Heath, “Bandwidth Gain: The Missing Gain of Massive MIMO,” inICC 2023- IEEE International Conference on Communications. IEEE, 2023, pp. 5997–6003

work page 2023
[6]

A scal- able generator for massive mimo baseband processing systems with beamspace channel estimation,

Y . Dai, H. Liew, M. E. Rasekh, S. H. Mirfarshbafan, A. Gallyas- Sanhueza, J. Dunn, U. Madhow, C. Studer, and B. Nikoli ´c, “A scal- able generator for massive mimo baseband processing systems with beamspace channel estimation,” in2021 IEEE Workshop on Signal Processing Systems (SiPS). IEEE, 2021, pp. 182–187

work page 2021
[7]

Long-term transmit beamforming for wireless multicast- ing,

A. Lozano, “Long-term transmit beamforming for wireless multicast- ing,” in2007 IEEE International Conference on Acoustics, Speech and Signal Processing-ICASSP’07, vol. 3. IEEE, 2007, pp. III–417

work page 2007
[8]

Scalable Long-Term Beamforming for Massive Multi-User MIMO

A. Rasteh, A. Kiani, M. Mezzavilla, and S. Rangan, “Scalable long- term beamforming for massive multi-user mimo,”arXiv preprint arXiv:2511.09464, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[9]

Vlsi design of large- scale soft-output mimo detection using conjugate gradients,

B. Yin, M. Wu, J. R. Cavallaro, and C. Studer, “Vlsi design of large- scale soft-output mimo detection using conjugate gradients,” in2015 IEEE International Symposium on Circuits and Systems (ISCAS). IEEE, 2015, pp. 1498–1501

work page 2015
[10]

A fast-convergent pre- conditioned conjugate gradient detection for massive mimo uplink,

Y . Xue, C. Zhang, S. Zhang, and X. You, “A fast-convergent pre- conditioned conjugate gradient detection for massive mimo uplink,” in2016 IEEE International Conference on Digital Signal Processing (DSP). IEEE, 2016, pp. 331–335

work page 2016
[11]

Low complexity linear detectors for massive mimo: A comparative study,

M. A. Albreem, W. Salah, A. Kumar, M. H. Alsharif, A. H. Rambe, M. Jusoh, and A. N. Uwaechia, “Low complexity linear detectors for massive mimo: A comparative study,”IEEE Access, vol. 9, pp. 45 740– 45 753, 2021

work page 2021
[12]

Finite-Precision Conjugate Gradient Method for Massive MIMO Detection

Y . Fang, L. Chen, C. You, D. Wen, and P. Zhu, “Finite-precision conjugate gradient method for massive mimo detection,”arXiv preprint arXiv:2504.09820, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[13]

A spatial array for spectrally agile wireless processing,

A. Rasteh, A. Hennessee, I. Shivhare, S. Garg, S. Rangan, and B. Reagen, “A spatial array for spectrally agile wireless processing,” arXiv preprint arXiv:2512.04182, 2025

work page arXiv 2025
[14]

On the convergence rate of variants of the conjugate gradient algorithm in finite precision arithmetic,

A. Greenbaum, H. Liu, and T. Chen, “On the convergence rate of variants of the conjugate gradient algorithm in finite precision arithmetic,”SIAM Journal on Scientific Computing, vol. 43, no. 5, pp. S496–S515, 2021

work page 2021
[15]

Sionna rt: Differentiable ray tracing for radio propagation modeling,

J. Hoydis, F. A ¨ıt Aoudia, S. Cammerer, M. Nimier-David, N. Binder, G. Marcus, and A. Keller, “Sionna rt: Differentiable ray tracing for radio propagation modeling,” in2023 IEEE Globecom Workshops (GC Wkshps). IEEE, 2023, pp. 317–321

work page 2023