🔬 optics

Analysis of Nonlinear Phase Noise in Coherent Fiber-Optic Systems Based on Phase Shift Keying

This paper develops analytical expressions for phase variance in coherent phase-shift keying fiber-optic systems by utilizing Gauss-Hermite functions as an orthogonal basis to model how amplifier noise excites higher-order modes that enhance both nonlinear and linear phase noise, ultimately deriving an optimum launch power that balances these effects.

Original authors: Shiva Kumar

Published 2026-08-06

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Shiva Kumar

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine the internet as a massive, high-speed highway made of glass threads called optical fibers. Instead of cars, we send tiny, invisible pulses of light carrying our emails, videos, and cat memes. To make these pulses travel thousands of miles without fading, we use "amplifiers" along the road, like gas stations that boost the signal. But here's the catch: these amplifiers aren't perfect. They add a little bit of static noise, like the hiss on an old radio. In a perfect world, this static would just be a tiny background hum. However, glass fibers have a weird quirk: they are slightly "nonlinear," meaning the light pulses can interact with each other and with the static noise in unexpected ways.

When a light pulse gets too strong, or travels too far, this static noise gets twisted by the fiber's nonlinearity into a new kind of problem called "nonlinear phase noise." Think of it like trying to read a message written on a spinning top. If the top spins perfectly, you can read it. But if the static noise makes the top wobble and spin erratically, the message becomes a blur. In digital communications, this wobble is measured as "phase noise." If the wobble gets too big—specifically, if the signal spins more than 90 degrees off course—it causes a "bit error," meaning your message gets garbled. Scientists have long known that two specific types of noise movements (one pushing the signal forward, one pushing it sideways) are the main culprits, but they wondered if other, more complex movements were also sneaking in to cause trouble.

This paper dives deep into that mystery, asking: "Are we ignoring other types of noise movements that might be making the signal wobble even worse?" The author, Shiva Kumar, uses a clever mathematical toolkit involving "Gauss-Hermite functions"—which are like a set of perfectly shaped, nested building blocks—to describe the noise field. Instead of just looking at the two main noise movements, the paper explores what happens when we count all the possible noise movements, or "degrees of freedom" (DOFs).

The study finds that while the two main noise movements are indeed the heavy lifters, the other, higher-order movements aren't entirely harmless. When the signal pulse mixes with these higher-order noise modes, it creates new, sneaky noise fields that can boost the phase noise. The paper uses computer simulations to test its math, and the results show a fascinating nuance: if the receiver uses a "matched filter" (a very precise tool that only listens to the exact shape of the signal), ignoring the extra noise movements causes only a small error—about 10% in the scenarios tested. However, if the receiver uses a broader filter that lets in more noise, or if the signal travels a very long distance at high power, those extra noise movements become much more significant, and ignoring them leads to bigger mistakes.

The paper also calculates the "sweet spot" for how much power to send into the fiber. Sending too little power makes the signal weak and vulnerable to linear noise (the basic static), while sending too much makes the nonlinear wobble explode. The author derives a formula for this optimum launch power, which matches up well with their simulations. They confirm that even in complex, dispersive systems (where the fiber stretches the light pulses), the total phase noise is minimized when the average nonlinear phase shift is around 1 radian. Ultimately, the paper doesn't say the old two-movement model is wrong, but it warns that for very long, high-power systems, we need to be more careful and count more of the noise's "degrees of freedom" to keep our digital messages crystal clear.

Technical Summary: Analysis of Nonlinear Phase Noise in Coherent Fiber-Optic Systems Based on Phase Shift Keying

Problem Statement
In fiber-optic systems utilizing Phase-Shift Keying (PSK) or Differential Phase-Shift Keying (DPSK), performance is significantly degraded by nonlinear phase noise. This phenomenon arises from the coupling between the fiber's Kerr nonlinearity and Amplified Spontaneous Emission (ASE) noise introduced by inline amplifiers. While ASE causes amplitude fluctuations, the fiber nonlinearity translates these into phase fluctuations. If the phase deviation exceeds $\pi/2$ , it results in bit errors.

Previous analytical models, notably by Gordon and Mollenauer, Mecozzi, and Ho, have largely relied on the assumption that only two degrees of freedom (DOFs) of the noise field—specifically, the in-phase and quadrature components matching the signal pulse shape—are significant. These models generally assume that other noise modes are negligible, particularly when a matched filter is used at the receiver. However, these expressions are often valid only for zero-dispersion systems or specific conditions. There is a need to understand the impact of higher-order noise modes and arbitrary filter bandwidths in dispersive, nonlinear systems to accurately model phase variance and determine optimum launch powers.

Methodology
The paper develops analytical expressions for phase variance in a coherent PSK system by representing the noise field as a superposition of Gauss-Hermite functions. This basis set is chosen because Gauss-Hermite functions are the exact modes of the paraxial Helmholtz equation and serve as exact solutions to the linear part of the nonlinear Schrödinger (NLS) equation.

The analysis proceeds under the following assumptions:

The system operates in a pseudo-linear regime where nonlinear effects are treated as a first-order perturbation on the linear system.
Signal power is significantly larger than the mean ASE power.
The interaction between dispersion, Self-Phase Modulation (SPM), and ASE on a single pulse is considered, while Inter-Channel Cross-Phase Modulation (IXPM) is excluded.

The methodology involves two main analytical approaches:

Two DOFs: The noise field is restricted to the fundamental mode (matching the Gaussian signal pulse). This recovers and extends previous results to include dispersive effects and matched filtering.
Arbitrary DOFs: The noise field is expanded to include higher-order Gauss-Hermite modes. The study investigates how the nonlinear mixing of the signal pulse with these higher-order noise modes generates new noise fields. It also analyzes the impact of electrical Gaussian filters with arbitrary bandwidths at the receiver.

The analytical results are cross-validated against Monte-Carlo simulations of the NLS equation using the split-step Fourier technique.

Key Contributions and Results

Extension to Arbitrary DOFs: The paper demonstrates that while two DOFs are sufficient for linear systems with matched filters, they are not strictly sufficient for nonlinear systems. The nonlinear mixing of the signal with higher-order noise modes generates new noise fields that occupy the same spectral region as the signal, thereby enhancing nonlinear phase noise.
Impact of Filter Bandwidth: The study reveals that when the receiver filter bandwidth exceeds the signal bandwidth, higher-order noise modes are no longer orthogonal to the signal. Consequently, these modes contribute to both linear and nonlinear phase noise.
Quantitative Error Analysis: In the specific parameter space analyzed (using a matched filter), the maximum error introduced by considering only two DOFs is approximately 10% compared to using 14 DOFs. However, the paper notes that this error may increase with longer transmission distances or higher launch powers. When using a Gaussian filter with twice the matched bandwidth, the error for a 2-DOF model rises significantly (up to 30% in the tested scenario).
Dispersion Effects: The analysis confirms that as dispersion increases, the variance of nonlinear phase noise due to SPM decreases, consistent with prior findings. The variance of nonlinear phase noise grows cubically with the number of amplifiers, whereas linear phase noise grows linearly, making nonlinear noise dominant in ultra-long-haul systems.
Optimum Launch Power: An analytical expression for the optimum launch power is derived by minimizing the total phase variance (sum of linear and nonlinear components). The derived expression assumes two DOFs and matched filters. The results indicate that the total phase variance is minimized when the mean nonlinear phase shift of the signal sample is approximately 1 radian, a condition that holds true even for dispersive systems when calculated correctly. The calculated optimum launch power (approx. 1.8 mW in the simulation) aligns well with numerical simulations.

Significance and Claims
The paper claims to provide a more comprehensive analytical framework for understanding nonlinear phase noise in coherent optical systems. By utilizing Gauss-Hermite basis functions, the author offers a method to systematically include arbitrary degrees of freedom, moving beyond the limitations of the traditional two-DOF approximation.

The author modestly states that their analytical expressions are derived under first-order perturbation theory. They acknowledge that at very high launch powers or extremely long distances, where second-order terms become significant, the accuracy of the first-order approximation may degrade. They suggest that for such cases, multiple-scale approaches or second-order perturbation theories could be employed.

Ultimately, the work provides a tool to estimate the number of degrees of freedom required for accurate modeling and offers a derived expression for optimum launch power that accounts for both linear and nonlinear phase noise in dispersive environments. The findings suggest that while the two-DOF model is a reasonable approximation for matched-filtered systems, the contribution of higher-order modes cannot be ignored when filter bandwidths are wide or system parameters push the limits of the first-order perturbation regime.

More like this