⚡ electrical engineering

Joint Subcarrier Phase Recovery for Nonlinearity Mitigation

This paper proposes a low-complexity joint subcarrier phase recovery scheme that simultaneously mitigates laser phase noise and fiber nonlinearity, achieving a 0.9 dB performance gain in long single-span Raman-amplified links with minimal computational overhead.

Original authors: Marco Secondini, Stella Civelli

Published 2026-03-02

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Marco Secondini, Stella Civelli

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 you are trying to send a high-speed text message across a very long, bumpy road. In the world of fiber optics, this "road" is a glass cable, and your "text messages" are light pulses carrying data.

The problem is that this road has two main annoyances:

The Wobbly Flashlight (Laser Phase Noise): The light source at the start isn't perfectly steady; it jitters slightly, making the message arrive a bit out of sync.
The Bumpy Road (Fiber Nonlinearity): As the light travels, it interacts with the glass itself. If you send too much power (to go faster), the road gets bumpy. This creates "noise" that scrambles the message, especially when you are sending many different channels of data at once.

For a long time, engineers had to fix these problems separately, which was slow and computationally expensive (like trying to fix a car while driving it).

This paper introduces a new, clever system called Joint Subcarrier Phase Recovery (JSCPR). Think of it as a smart navigation team that fixes the message while it's traveling, using a two-step strategy.

The Two-Step Strategy

The system splits the incoming data into several "subcarriers" (like splitting a big truckload of data into smaller, manageable delivery vans). It then uses two specialized tools to clean up the signal:

Step 1: The "Predictable Bump" Fixer (NLPC)

The Analogy: Imagine you know exactly where the potholes are on a specific stretch of road because you've driven it a thousand times. You can predict exactly how much your car will bounce and adjust your suspension before you hit the bump.
The Science: This first block handles intra-channel nonlinearity. This is the distortion caused by the signal interacting with itself. Because this distortion is "deterministic" (predictable based on the signal's own strength), the system calculates the exact "bump" and rotates the signal back to its original shape instantly. It's like a reflex action.

Step 2: The "Random Drift" Fixer (PPNC)

The Analogy: Now imagine the wind is blowing randomly, pushing your delivery vans off course in unpredictable ways. You can't predict the wind, but you have a few "checkpoints" (pilot symbols) along the route where you drop a flag. By looking at where the flags landed compared to where they were supposed to be, you can figure out the general wind pattern and steer the other vans back on track.
The Science: This second block handles inter-channel nonlinearity (signals messing with each other) and the laser jitter. These are random and unpredictable. The system uses special "pilot" symbols (known reference points) sent periodically. It compares the received pilots to the expected ones to estimate the random drift, then uses a sophisticated filter (a Wiener filter) to smooth out the noise for the rest of the data.

Why is this a Big Deal?

Usually, fixing these problems requires a supercomputer. This new method is like upgrading from a mainframe computer to a sleek smartphone app.

Efficiency: It does all this math incredibly fast. The authors calculated that it only takes about 99 simple math operations for every single piece of data it processes.
The Result: In a test simulating a 250 km (155-mile) fiber link, this system improved the signal quality by 0.9 dB. In the world of fiber optics, that's a huge win. It means you can send data faster, further, or with less error without needing to lay new cables.

The Takeaway

Think of this paper as inventing a smart autopilot for fiber optic cables. Instead of just reacting to errors after they happen, it uses a two-part team:

One part that predicts the road's bumps based on the car's own weight.
One part that learns from random wind gusts using checkpoints.

By working together across all the data "lanes" (subcarriers) at once, they keep the message clear and the connection strong, even over long distances and at high speeds. This is crucial for the future of the internet, especially as AI and data centers demand faster, more reliable connections.

1. Problem Statement

Modern optical fiber communication systems, particularly those used in Data Center Interconnects (DCIs) and long-haul links, face a critical performance bottleneck: fiber Kerr nonlinearity.

Context: As data traffic grows (driven by AI), there is a shift toward using coherent detection and advanced Digital Signal Processing (DSP) even for short-reach, unrepeatered single-span links (up to a few hundred kilometers) to maximize spectral efficiency and reach.
Challenge: Higher launch powers required for these links exacerbate nonlinear effects. Specifically, Nonlinear Interference (NLI) manifests as phase noise that limits system performance.
Limitation of Existing Solutions: Traditional Carrier Phase Recovery (CPR) schemes typically address laser phase noise but fail to effectively mitigate the complex phase fluctuations induced by fiber nonlinearity. Furthermore, existing nonlinearity mitigation techniques often lack the computational efficiency required for high-speed, multi-subcarrier transceivers.

2. Methodology: Joint Subcarrier Phase Recovery (JSCPR)

The authors propose a low-complexity, two-stage Joint Subcarrier Phase Recovery (JSCPR) scheme. This approach treats NLI as a phase noise process with limited temporal and spectral coherence and leverages subcarrier multiplexing to mitigate it.

The scheme operates jointly across all subcarriers ( $M$ ) and polarizations of an optical channel. It consists of two cascaded blocks:

Stage 1: Nonlinear Phase Compensation (NLPC)

Function: Mitigates deterministic nonlinear phase distortion arising from intra-channel NLI (interactions within the same channel/subcarrier).
Mechanism:
- Calculates subcarrier intensities $I(k)$ from the received signal.
- Processes these intensities through a MIMO filter with an $M \times M$ matrix impulse response $C(k)$ .
- Generates estimated phase rotations $\theta(k)$ which are applied to rotate the subcarrier samples.
Implementation: Operates at the full symbol rate ( $R_s$ ). Due to the long filter response ( $N_c = 250$ ), it is implemented efficiently in the frequency domain using FFT/IFFT pairs and the overlap-and-save technique.

Stage 2: Pilot-Aided Phase Noise Compensation (PPNC)

Function: Compensates for stochastic phase noise caused by inter-channel NLI (interactions between different WDM channels) and laser phase noise.
Mechanism:
- Uses pilot symbols (inserted at intervals $P$ ) to estimate preliminary phasors.
- Employs a bank of parallel MIMO Wiener filters ( $2M \times 2M$ ) to smooth and interpolate these estimates, accounting for correlations across time, subcarriers, and polarizations.
- Applies the final phase correction to the signal.
Implementation: Operates at a reduced rate ( $R_s/P$ ). Since the filter length is short ( $N_d = 2$ ), it is implemented in the time domain.

3. Key Contributions

Novel Architecture: Introduction of a joint-subcarrier CPR scheme that simultaneously handles laser phase noise and fiber nonlinearity, distinguishing between deterministic (intra-channel) and stochastic (inter-channel) components.
Low Complexity: The algorithm is designed for high efficiency. The total computational complexity is derived and optimized, making it feasible for real-time DSP implementation.
Dual-Stage Mitigation: The paper demonstrates that separating the mitigation of deterministic and stochastic phase noise allows for targeted, more effective compensation than single-stage approaches.
Optimization Potential: The authors identify that the PPNC stage (the most complex part) can be further optimized via coefficient pruning and exploiting symmetries in the Wiener filter bank.

4. Performance Results

The scheme was evaluated via numerical simulations of a 250 km unrepeatered single-mode fiber link with counter-propagating distributed Raman amplification.

System Parameters:
- Signal: 5 WDM channels, each with 4 subcarriers (Total 180 GBd per channel).
- Modulation: Dual-polarization Probabilistic-Amplitude-Shaping (PAS) 64-QAM.
- Laser Linewidth: 100 kHz (for both transmitter and local oscillator).
Complexity:
- NLPC Stage: ~16 Real Multiplications per Complex Symbol (RM/2D).
- PPNC Stage: ~83 RM/2D.
- Total Complexity: 99 RM/2D.
Performance Gain:
- Without Laser Phase Noise: The combined JSCPR scheme achieved a 0.9 dB SNR gain compared to the baseline. Individually, the NLPC and PPNC stages provided ~0.45 dB gains each, confirming they address different impairment types.
- With Laser Phase Noise: The scheme maintained a significant gain over the baseline (which included a standard CPR but no nonlinearity mitigation), demonstrating robustness against laser imperfections.

5. Significance and Conclusion

This work addresses a critical gap in optical communications: the need for high-performance, low-complexity nonlinearity mitigation in next-generation coherent systems (especially for DCIs).

Practical Impact: By achieving a 0.9 dB gain with only 99 RM/2D, the JSCPR scheme offers a viable path to extending the reach of high-capacity, unrepeatered links without compromising spectral efficiency.
Scalability: The use of subcarrier multiplexing and MIMO filtering makes the approach scalable for large-bandwidth transceivers.
Future Work: The authors note that further gains can be realized by optimizing pilot allocation strategies and refining the Wiener filter implementation (e.g., pruning coefficients), which are currently under investigation.

In summary, the paper presents a sophisticated yet computationally efficient solution that effectively decouples and mitigates the complex phase noise interactions in modern optical fiber links, enabling higher data rates and longer transmission distances.