Linear Receive Beamforming for CAPA Systems

Chongjun Ouyang, Zhaolin Wang, Xingqi Zhang, Yuanwei Liu

TL;DR

As wireless technology moves toward 6G, we are shifting from discrete antenna arrays to Continuous-Aperture Arrays (CAPA)—essentially turning entire surfaces into electromagnetic sensors. This paper provides the first comprehensive mathematical framework for linear receive beamforming in CAPA systems, deriving closed-form solutions for MRC, ZF, and MMSE techniques and proving that these continuous systems significantly outperform traditional discrete antenna arrays.

What Problem Does This Paper Address?

For decades, Multiple-Input Multiple-Output (MIMO) technology has relied on Spatially-Discrete Arrays (SPDAs)—groups of individual antenna elements separated by specific distances. Our current signal processing math is built on this "discrete" reality, using vectors and matrices to represent signals.

However, the "endgame" of antenna evolution is the Continuous-Aperture Array (CAPA), where the entire surface has a continuous current distribution. While CAPAs offer superior spatial resolution and flexibility, they break our traditional math. Instead of simple matrix multiplications, we now have to deal with continuous integral operators. Standard operations like matrix inversion or eigen-decomposition don't easily translate to this infinite-dimensional space. This paper bridges that gap by providing the theoretical tools and closed-form expressions needed to design and analyze beamforming for these next-generation surfaces.

Methods

The authors tackle the problem using the formalism of integral operators and functional analysis. They propose three linear receive beamforming strategies:

  1. Maximum-Ratio Combining (MRC): Designed to maximize the signal power for each individual user by aligning the beamformer with the user's spatial channel response.
  2. Zero-Forcing (ZF): Aimed at completely nullifying inter-user interference (IUI). The authors use a "function space interpretation" to prove that their proposed ZF beamformer is the optimal way to maximize SNR while ensuring zero interference in a continuous domain.
  3. Minimum Mean-Squared Error (MMSE): The most sophisticated approach. The authors mathematically prove that in CAPA systems, maximizing the data rate is equivalent to minimizing the estimation error (MSE). They derive a closed-form MMSE beamformer that acts as a two-stage filter: one part whitens the noise and interference, and the second part performs matched filtering.

Crucially, they prove that despite the "infinite" nature of the continuous aperture, the optimal beamformers always lie within the finite-dimensional signal subspace spanned by the users' spatial responses.

Datasets

This is a theoretical and analytical paper, so it does not use external training datasets. Instead, the authors validate their findings through numerical simulations using:

  • Deterministic Channels: A free-space line-of-sight (LoS) model.
  • Statistical Channels: An isotropic Rayleigh fading model to account for multipath scattering.
  • Benchmarks: The performance is compared against conventional SPDAs (discrete arrays) with equivalent physical dimensions and aperture areas.

Key Insights and Ideas

  • The "Signal Subspace" Shortcut: One of the most important "aha" moments is the proof that optimal beamforming in an infinite-dimensional continuous space can be reduced to a finite-dimensional problem. Because the beamformers only need to care about the specific spatial "signatures" of the $K$ users, the complexity remains manageable.
  • The MRC Paradox: The study reveals a surprising result: in high-interference scenarios, a CAPA using simple MRC can actually perform worse than a discrete array. This is because the larger continuous aperture is so sensitive that it amplifies interference just as much as the signal if not managed correctly.
  • Unified Framework: The paper shows that CAPA beamformers share a similar structural form with SPDA beamformers (weighted sums of spatial responses), providing a comforting bridge between classical MIMO theory and the future of continuous EM surfaces.

Key Results

  • Superiority of CAPA: Under ZF and MMSE beamforming, CAPAs consistently achieve higher sum-rates and lower sum-MSE than discrete arrays.
  • Effective Channel Gain: CAPAs leverage spatial resources more effectively. For example, in the low-SNR regime, CAPAs provide a significant boost in effective channel gain compared to SPDAs due to their larger "effective" capture area.
  • Asymptotic Convergence: The authors demonstrate that the MMSE beamformer is the "universal" solution—it naturally converges to the MRC solution when noise is high (low SNR) and to the ZF solution when interference is high (high SNR).

Limitations and Future Work

  • Perfect CSI Assumption: The paper assumes the Base Station has perfect Channel State Information (CSI). In practice, estimating a continuous channel response is much more difficult than estimating a discrete vector, and the impact of "channel aging" or estimation error remains an open question.
  • Hardware Complexity: While the math works, the paper notes that CAPAs require a limited number of RF chains. The mapping between the continuous aperture and the discrete RF ports (the "analog-to-digital" interface of the antenna) is a hardware challenge not fully explored here.
  • Near-Field Effects: As apertures get larger, users are more likely to be in the "near-field" where wavefronts are spherical rather than planar. While the LoS model used here touches on this, more complex non-stationary channel models are needed for future 6G deployments.

Summary updated

Abstract

The performance of linear receive beamforming in continuous-aperture array (CAPA)-based uplink communications is analyzed. Three continuous beamforming techniques are proposed under the criteria of maximum-ratio combining (MRC), zero-forcing (ZF), and minimum mean-squared error (MMSE). i) For <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MRC beamforming</i>, a closed-form expression for the beamformer is derived to maximize per-user signal power. The achieved uplink rate and mean-squared error (MSE) in detecting received data symbols are analyzed. ii) For <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">ZF beamforming</i>, a closed-form beamformer is derived based on channel correlation to eliminate interference. As a further advance, its optimality in maximizing effective channel gain while ensuring zero inter-user interference is proven. iii) <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">MMSE beamforming</i> is established as the optimal linear receive approach for CAPAs in terms of maximizing per-user rate and minimizing MSE. Closed-form expressions are derived for the MMSE beamformer and the achievable sum-rate and sum-MSE. It is mathematically proven that all proposed beamformers lie within the signal subspace spanned by users’ spatial responses. Numerical results demonstrate that CAPAs outperform conventional spatially-discrete arrays (SPDAs) by achieving higher sum-rates and lower sum- MSEs under the proposed linear beamforming techniques.
BibTeX
@article{ouyang2026linear,
  author = {Chongjun Ouyang and Zhaolin Wang and Xingqi Zhang and Yuanwei Liu},
  title = {{Linear Receive Beamforming for CAPA Systems}},
  journal = {IEEE Trans. Wireless Communications},
  year = {2026},
  doi = {10.1109/twc.2025.3588626},
}

Loading related content...

Figure