Information Transmission in Signal Dependent Noise Channels

18 concepts10 papers1 proposal

In classical information theory, communication channels are frequently analyzed through the lens of Additive White Gaussian Noise (AWGN). In this traditional paradigm, noise is treated as an independent, omnipresent background hum—like static on a radio—that degrades a signal equally regardless of the message being sent.

However, in many physical, biological, and optical systems, this assumption fails. In molecular communication (where biological agents exchange chemical concentrations) or optical networks (where photon counts carry data), the noise is intrinsically tied to the signal itself. This is the domain of signal-dependent noise, where the variance of the noise fluctuations changes dynamically based on the intensity or state of the transmitted signal. Understanding these channels requires a fundamental shift in how we calculate channel capacity, shape input signals, and decode information.

The Non-Linear Profile of Biological Noise

To build intuition for why signal-dependent noise changes the rules of communication, consider a molecular receiver. In Molecular Communication Between Two Populations of Bacteria, Arash Einolghozati et al. model how populations of engineered bacteria exchange information via chemical concentrations. When a transmitter releases signaling molecules into a medium, they diffuse and bind to receptors on a receiver node.

The noise in this channel is not a constant background factor; instead, it arises from the stochastic, probabilistic nature of ligand-receptor binding.

  • If the signal concentration is zero, no molecules bind, and the variance drops to zero.
  • If the signal concentration is extremely high, the receptors become completely saturated, meaning every receptor is occupied, and the variance drops to zero again.
  • The uncertainty peaks exactly in the middle—at 50% receptor occupancy—where the system is most volatile and unpredictable.

As noted in Collective Sensing-Capacity of Bacteria Populations, this creates a noise profile that fluctuates based on the probability of receptor occupancy. This non-additive behavior means that the reliability of a message depends entirely on the specific concentration level chosen to represent it.

Shaping Inputs to Avoid High-Variance States

Because noise is concentrated at intermediate signal levels, maximizing the rate of data transmission requires a radical departure from traditional signaling strategies. In standard Gaussian channels, a smooth, continuous Gaussian input distribution is mathematically optimal. In contrast, when dealing with signal-dependent noise that peaks at intermediate activation states, the optimal strategy is to avoid the middle entirely.

In Decode and forward relaying in diffusion-based molecular communication between two populations of biological agents, Arash Einolghozati et al. demonstrate that maximizing mutual information requires shaping the input distribution to place higher probabilities near the minimum and maximum concentration boundaries.

By forcing the system to operate primarily at the extremes (near zero or at maximum power), the transmitter prioritizes states where the noise variance naturally diminishes. This non-uniform signaling distribution leverages the "quiet zones" of the physical medium, effectively filtering out individual randomness by design.

The Paradox of Helpful Noise

Intuitively, one might assume that increasing the noise variance in a communication system always degrades its capacity. However, signal-dependent noise introduces a fascinating, counter-intuitive phenomenon: a channel with higher noise variance can sometimes achieve a greater transmission capacity than one with lower variance.

In On the Capacity of Signal Dependent Noise Channels, Hamid Ghourchian et al. uncover this non-monotonic capacity behavior. Because the noise variance changes predictably with the signal intensity, the pattern of the noise fluctuations itself carries information. A receiver can analyze the variance of the received signal to help infer exactly what the transmitter sent.

As further explored by the authors in On the Capacity of a Class of Signal-Dependent Noise Channels, reducing noise variance can paradoxically hurt capacity if it smooths out these distinct, signal-revealing fluctuation signatures. The receiver utilizes the specific structure of the noise to distinguish between different inputs more effectively.

Super-Exponential Identification over Poisson Channels

When noise scales directly with signal intensity—such as in discrete-time Poisson channels common in optical and molecule-counting receivers—traditional Shannon transmission capacity is not the only metric of interest. If the communication goal shifts from standard message decoding to Deterministic Identification (DI)—where the receiver merely needs to identify whether a specific, pre-arranged trigger message was sent—signal-dependent noise unlocks extraordinary scaling laws.

In Deterministic Identification Over Poisson Channels, Mohammad J. Salariseddigh et al. prove that the number of identifiable messages over Poisson channels grows super-exponentially ($2^{(n \log n)R}$) with the codeword length $n$. This stands in stark contrast to Gaussian channels, where identification capacity scales exponentially ($2^{nR}$).

This phenomenon, also detailed in Deterministic Identification for Molecular Communications over the Poisson Channel by Mohammad J. Salariseddigh et al., provides a theoretical explanation for the extraordinary discriminatory power of natural systems like olfaction. Because the noise variance is intrinsically linked to the signal mean, the geometric space of distinguishable noise profiles expands rapidly, allowing an observer to reliably identify vastly more distinct states than traditional coding frameworks would predict.

Exploiting and Reshaping Noise in Physical Systems

Understanding the mathematical structure of signal-dependent noise allows researchers to identify how natural systems have evolved to handle it, and how synthetic systems can be engineered to mitigate it.

In biological sensory processing, Boris S. Gutkin et al. demonstrate in Sensory noise predicts divisive reshaping of receptive fields that neural circuits use "divisive normalization" as an optimal strategy to cope with intensity-dependent uncertainty. By selectively dividing inputs by their predicted values, neurons dynamically reshape their receptive fields, maintaining a consistent and invariant code despite noise that scales with the magnitude of environmental stimuli.

In synthetic molecular communication, the fact that noise scales proportionally with molecule concentration presents a major hurdle for maintaining high signal-to-noise ratios, as noted in Physical Layer Network Coding in Molecular Two-Way Relay Networks by Maryam Farahnak-Ghazani et al.

To combat this, the authors propose leveraging chemical reactions directly within the communication medium. As expanded in On Medium Chemical Reaction in Diffusion-Based Molecular Communication: A Two-Way Relaying Example by Maryam Farahnak-Ghazani et al., allowing information-carrying molecules to react and cancel each other out in the environment performs physical-layer network coding (such as simultaneous XOR operations). This "computation in the medium" actively reduces inter-symbol interference and clears away excess molecules, mitigating the compounding effects of signal-dependent noise before it ever reaches the receiver.

Key Foundational Work

Molecular Communication Between Two Populations of Bacteria (2012) — contributed 5 concept elements to this direction.
Decode and forward relaying in diffusion-based molecular communication between two populations of biological agents (2014) — contributed 3 concept elements to this direction.
Collective Sensing-Capacity of Bacteria Populations (2012) — contributed 3 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_signal_dependent_noise_ee0c3c11,
  title = {signal-dependent noise},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/signal-dependent-noise/rdr_ee0c3c11ea844436bc46a35c588e0d77},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-19},
}

This exact version (version 1, frozen text):

@misc{lacuna_signal_dependent_noise_ee0c3c11_v1,
  title = {signal-dependent noise},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/signal-dependent-noise/rdr_ee0c3c11ea844436bc46a35c588e0d77/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-19},
}
About

These directions overlap with this one through shared papers.

See something wrong? Report a mistake