Spatial Noise Shaping for Low-Resolution Massive MIMO

17 concepts8 papers1 proposal

Massive MIMO systems achieve extraordinary data rates by packing tens or hundreds of antennas onto a single base station. However, scaling up the number of antennas introduces a severe hardware bottleneck: power consumption. Running high-resolution Analog-to-Digital Converters (ADCs) and Digital-to-Analog Converters (DACs) across hundreds of RF chains is both power-prohibitive and expensive.

To build sustainable massive MIMO arrays, engineers have turned to ultra-low-resolution hardware—down to the extreme case of 1-bit quantization. While 1-bit converters are incredibly power-efficient, they introduce massive, highly non-linear quantization noise that threatens to destroy high-order modulation schemes like 16-QAM.

The solution lies in a profound conceptual shift: moving noise shaping from the time domain to the spatial domain. By treating physical antenna arrays exactly like temporal sampling circuits, we can actively steer quantization noise away from users, creating pristine, noise-free spatial channels out of low-cost hardware.

Passive Mitigation via Spatial Oversampling

Before exploring active noise shaping, it helps to understand why massive MIMO inherently tolerates quantization noise better than single-antenna systems. When an array features a massive number of antennas, it benefits from an effect known as spatial oversampling.

In a conventional system, 1-bit quantization completely distorts complex signal constellations. However, as shown by Sven Jacobsson et al. in One-bit massive MIMO: Channel estimation and high-order modulations, the sheer abundance of antennas provides a powerful averaging effect. Because the quantization noise generated at each individual antenna is largely uncorrelated with the noise at its neighbors, the noise adds up incoherently. Meanwhile, the intended communication signals combine coherently when beamformed.

This spatial averaging allows the array to overcome the strict mathematical limits of 1-bit quantization. In Throughput Analysis of Massive MIMO Uplink With Low-Resolution ADCs, Sven Jacobsson et al. demonstrated that massive MIMO systems can successfully support high-order modulations like 16-QAM even with 1-bit hardware. Similarly, Caijun Zhong et al. proved in Full-Duplex Massive MIMO Relaying Systems with Low-Resolution ADCs that deploying massive arrays at a relay station can fully compensate for low-resolution hardware noise, allowing a $1/M$ reduction in transmit power (where $M$ is the number of antennas) while maintaining a constant data rate.

The Core Mechanism: Spatial Sigma-Delta Modulation

Passive averaging helps, but it distributes quantization noise uniformly across all spatial directions. To maximize performance, we can actively manipulate where the noise goes by adapting a classic signal processing technique: Sigma-Delta ($\Sigma\Delta$) modulation.

In traditional temporal $\Sigma\Delta$ converters, a signal is sampled much faster than the Nyquist rate (temporal oversampling). A feedback loop subtracts the quantization error of the previous sample from the current sample before quantization. This mathematical subtraction acts as a high-pass filter on the noise, pushing the quantization error into high temporal frequencies, leaving a clean, low-frequency band for the actual signal.

Spatial noise shaping maps this exact principle directly onto physical space:

Concept Temporal Sigma-Delta Spatial Sigma-Delta
Oversampling Sampling faster than the Nyquist rate in time Spacing antennas closer than the standard $\lambda/2$ wave distance
Feedback Loop Delaying error by one time step ($z^{-1}$) and subtracting it from the next temporal sample Passing the quantization error of one antenna element to its physical neighbor
Frequency Effect Pushes noise to high temporal frequencies Pushes noise to high spatial frequencies (oblique angles)

By reducing the physical spacing between antenna elements below the standard half-wavelength ($\lambda/2$), the array achieves spatial oversampling. As explained by R. S. Prasobh Sankar et al. in Channel Estimation in MIMO Systems with One-bit Spatial Sigma-delta ADCs, running a feedback loop across the physical layout of the array redistributes the quantization noise.

Instead of polluting the entire environment, the noise is filtered and pushed into high spatial frequencies—which correspond to extreme, oblique angles or "invisible" regions of the spatial spectrum. This leaves a highly precise, low-noise "notch" in the spatial spectrum for high-fidelity communication, a mechanism leveraged by R. S. Prasobh Sankar et al. in Millimeter Wave MIMO Channel Estimation with 1-bit Spatial Sigma-delta Analog-to-Digital Converters to achieve unquantized-like channel estimation accuracy.

Standard 1-Bit Array (No Shaping):
[User] <--- (Signal + Uniform Quantization Noise) --- [Base Station]

Spatial Sigma-Delta Array:
[User] <--- (Clean Signal Notch) -------------------- [Base Station]
                               \__ (Shaped Noise) --> [Empty Space / High Angles]

Steering the Noise-Free Notch

A basic spatial $\Sigma\Delta$ loop naturally creates its low-noise notch at the broadside of the antenna array ($0^\circ$). If a user is located at a different angle, they will sit directly in the high-noise region of the shaped spectrum.

To make this architecture practical, the low-noise region must be dynamic. The authors achieved this by introducing a complex phase shift into the spatial feedback loop. By multiplying the quantization error by a specific phase factor before passing it to the adjacent antenna element, the system can electronically steer the zero-noise angle to match the user's position.

[Antenna n] ---> (Quantize) ---> [Output n]
     |
  (Extract Error)
     |
  [Phase Shift (e^jθ)]  <-- Controls the angle of the noise-free notch
     |
     v
[Antenna n+1] ---> (Add Error to Input) ---> (Quantize) ---> [Output n+1]

In One-Bit Sigma-Delta MIMO Precoding, Wing-Kin Ma et al. demonstrated that this steerable architecture simplifies downlink transmission. Instead of solving intractable binary optimization problems to determine 1-bit vector outputs, the spatial $\Sigma\Delta$ loop automatically handles the quantization constraints while steering a zero-noise angle directly toward target users.

This steering capability is highly versatile across different hardware form factors:

  • Receivers & Transmitters: R. S. Prasobh Sankar et al. showed that incorporating phase shifts into the receiver feedback loop ensures robust uplink channel estimation across a wide range of incident angles.
  • Reconfigurable Intelligent Surfaces (RIS): In Transmitting Data Through Reconfigurable Intelligent Surface: A Spatial Sigma-Delta Modulation Approach, Wai-Yiu Keung et al. applied this concept to passive surfaces. By steering the quantization errors resulting from discrete phase-adjustments into angular sectors where no users are present, they enabled high-order data modulation using low-resolution RIS hardware.

Repurposing the Shaped Noise

If spatial noise shaping works by dumping massive amounts of quantization noise into specific, unused angular directions, an intriguing question arises: can we use that discarded noise energy for something productive?

In Integrated Sensing and Communication (ISAC) systems, the answer is yes. In One-Bit Sigma-Delta DFRC Waveform Design: Using Quantization Noise for Radar Probing, Wai-Yiu Keung et al. designed a Dual-Function Radar-Communication system that treats the shaped quantization noise not as waste, but as a resource.

While the spatial $\Sigma\Delta$ loop maintains a pristine, low-noise notch in the direction of a communication user, the quantization noise pushed to wider angles is structured to serve as a radar probing signal. The system effectively uses the "noise" to scan the surrounding environment for radar targets, turning a hardware limitation into a dual-purpose advantage.

Key Foundational Work

One-bit massive MIMO: Channel estimation and high-order modulations (2015) — contributed 2 concept elements to this direction.
Full-Duplex Massive MIMO Relaying Systems with Low-Resolution ADCs (2017) — contributed 1 concept elements to this direction.
Throughput Analysis of Massive MIMO Uplink With Low-Resolution ADCs (2017) — contributed 1 concept elements to this direction.
History & citationVersion 1 · 2026-10-04

Current · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_spatial_noise_shaping_7c290b48,
  title = {spatial noise shaping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/spatial-noise-shaping/rdr_7c290b485b94407ab853d187bc5bb8c4},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_spatial_noise_shaping_7c290b48_v1,
  title = {spatial noise shaping},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/spatial-noise-shaping/rdr_7c290b485b94407ab853d187bc5bb8c4/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake