🔬 optics

An orthogonal-to-non-orthogonal multiplexing format converter

This paper proposes and demonstrates a versatile Talbot-based converter that bridges orthogonal and non-orthogonal multiplexing formats by coherently transferring and superposing distinct wavelength channels via cross-phase modulation, enabling flexible spectral allocation and high-capacity transmission in next-generation agile optical networks.

Original authors: Zijian Li, Chen Ding, Zixian Wei, Qiarong Xiao, Ka-Suen Lee, Chaoran Huang, Changyuan Yu, Chester Shu

Published 2026-07-07

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Zijian Li, Chen Ding, Zixian Wei, Qiarong Xiao, Ka-Suen Lee, Chaoran Huang, Changyuan Yu, Chester Shu

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 a busy highway where every car (data) is assigned its own specific lane (wavelength). This is how current optical networks work: they use Orthogonal Multiplexing. It's very orderly, but it has a problem. If a lane is empty, no one else can use it, even if they are right next to it. It's like having a 10-lane highway where only two lanes are open; the rest sit empty, wasting space.

This paper introduces a clever new "traffic converter" that changes the rules. Instead of forcing every car to stay in its own lane, it allows multiple cars to share the same lane by driving at different speeds and volumes. This is called Non-Orthogonal Multiplexing.

Here is how the researchers built this converter, explained through simple analogies:

1. The Problem: The "Rigid Lane" System

In current networks, if you have two users sending data, they need two separate lanes (wavelengths) and two separate receivers at the destination. If the traffic is light, those lanes sit half-empty. If the traffic is heavy, you run out of lanes. It's rigid and inefficient.

2. The Solution: The "Super-Lane" Converter

The authors built a device that takes two separate lanes of traffic and merges them into a single "super-lane."

The Magic Trick: They don't just smash the data together. They use a special technique called Talbot Processing (think of this as a "time-magnifying glass") and a Nonlinear Fiber (a special glass tube that reacts strongly to light).
How it works: Imagine a rhythmic drumbeat (a pulsed laser) hitting the two separate data streams. Because of the "time-magnifying glass," the drumbeat speeds up. When this fast drumbeat hits the data streams inside the special glass tube, it acts like a translator. It takes the data from Lane A and Lane B and "transfers" them both onto the same frequency, but with different volumes (power levels).
The Result: You now have one lane carrying two streams of data. One stream is loud (high power), and the other is quiet (low power).

3. The Receiver: The "Noise-Canceling" Decoder

You might ask, "If two streams are in one lane, how do we separate them?"
The paper explains that the receiver uses a digital trick called Successive Interference Cancellation (SIC).

Step 1: The receiver listens to the "loud" stream first. Because it's loud, it's easy to hear and decode, even with the quiet stream buzzing in the background.
Step 2: Once the loud stream is decoded, the computer subtracts it from the signal (like using noise-canceling headphones to remove a specific sound).
Step 3: Now that the loud noise is gone, the "quiet" stream becomes clear and can be decoded easily.

4. The Experiments: Proving it Works

The researchers tested this in two real-world scenarios:

Scenario A: The Campus Uplink (The "Office Building" Test)
They simulated a scenario where two different users (like two offices) send data back to a central server.
- Before: They needed two separate receivers.
- After: They used their converter to merge the two signals into one. A single receiver successfully decoded both users' data. They even tested this over 20 kilometers of real fiber optic cable, and it worked perfectly.
Scenario B: The Dense Edge Network (The "Crowded Stadium" Test)
They took this a step further. Instead of just two users, they had two "channels," where each channel carried four smaller sub-channels (like four different people in a group). That's 8 users total.
- They merged all 8 users into a single lane.
- They sent this signal over a 3-kilometer field-deployed fiber (real cables running across a university campus, not just a lab table).
- Result: The single receiver successfully separated and decoded all 8 users.

Why This Matters

The paper claims this technology allows networks to:

Save Space: You can reuse wavelength slots that would otherwise sit empty.
Save Hardware: You need fewer expensive receivers at the central hub (cutting the number of receivers in half in their tests).
Be Flexible: The system can be reprogrammed to handle different distances between channels, making it adaptable to various network layouts.

In short, the authors have built a "traffic manager" that turns a rigid, lane-based highway into a flexible, high-capacity super-highway, allowing more data to flow through existing infrastructure without needing to lay new cables or buy more expensive equipment.

Technical Summary: An Orthogonal-to-Non-Orthogonal Multiplexing Format Converter

Problem Statement
Current optical communication networks rely heavily on orthogonal multiplexing principles, such as Dense Wavelength Division Multiplexing (WDM) and Time-Frequency Division Multiple Access. While these methods effectively suppress inter-channel interference, they impose rigid constraints on spectral allocation. In dynamic, bursty traffic scenarios typical of next-generation networks—such as coherent Passive Optical Networks (PONs) and dense edge access networks—strict orthogonality leads to spectral fragmentation and suboptimal resource utilization. Each channel occupies a dedicated spectral slot until decoded, preventing dynamic reuse of spectrum across users. Furthermore, the requirement for dedicated coherent receiver ports for every wavelength channel creates hardware bottlenecks and high capital costs at aggregation nodes. There is a need for a mechanism to bridge orthogonal WDM infrastructures with non-orthogonal multiple access (NOMA) paradigms, which offer higher spectral efficiency and flexible power-domain superposition, without sacrificing compatibility with existing coherent detection systems.

Methodology
The authors propose and demonstrate an all-optical, reconfigurable converter that transforms orthogonal WDM channels into a non-orthogonal, power-division multiplexing format. The core mechanism relies on programmable temporal Talbot processing combined with Kerr-induced cross-phase modulation (XPM).

Talbot-Processed Pump Generation: A pulsed optical pump, derived from a parametric Optical Frequency Comb (OFC), is routed through a programmable dispersive module. By satisfying specific temporal Talbot conditions, the module multiplies the pump's repetition rate by an integer factor ( $m$ ). This allows the system to adapt to various channel spacings ( $\Delta f_c$ ) by ensuring the condition $\Delta f_c = q \cdot m f_r$ is met, where $f_r$ is the base repetition rate.
Nonlinear Signal Aggregation: The rate-multiplied pump is combined with the input WDM signal channels and launched into a third-order ( $\chi^{(3)}$ ) Kerr nonlinear medium. Through XPM, the pump induces coherent phase shifts on the signal channels.
Coherent Transfer: The XPM effect generates sidebands that coherently transfer the distinct frequency channels onto a single target wavelength. The signals from multiple users are superimposed in the power domain, creating a composite waveform.
Reception and Recovery: The aggregated signal is received by a single coherent receiver. Digital Signal Processing (DSP) employs Successive Interference Cancellation (SIC) to recover individual user data. The stronger-power user is decoded first, its contribution is subtracted, and the weaker-power user is subsequently recovered.

Key Contributions

Novel Conversion Mechanism: The paper introduces a method to perform orthogonal-to-non-orthogonal conversion entirely in the optical domain, avoiding electronic throughput bottlenecks. Unlike conventional all-optical modulation format converters (MFCs) that synthesize new constellations, this approach operates in the multiplexing domain, preserving separable channel information while relaxing phase stability requirements between users.
Reconfigurability: The system is programmable; by adjusting the Talbot multiplication factor ( $m$ ) via the dispersive module, the converter can accommodate different WDM channel spacings (demonstrated for 200 GHz and 250 GHz separations) and signal formats.
Hardware Consolidation: The architecture reduces the number of required coherent receiver ports at the aggregation node by half (e.g., converting two channels to one), thereby addressing the hardware scaling bottleneck in PONs and edge networks.

Experimental Results
The authors validated the concept through two proof-of-concept demonstrations:

Coherent PON Uplink Scenario: Two 80-Gbps WDM QPSK channels (separated by 200–250 GHz) were aggregated into a single 160-Gbps non-orthogonal channel.
- Performance: With an optimal power allocation ratio of 7.5 dB, both users achieved Bit Error Rates (BER) below the soft-decision Forward Error Correction (SD-FEC) threshold ( $2.2 \times 10^{-2}$ ) after 20 km of fiber transmission.
- Robustness: The system maintained performance within 25% SD-FEC limits for received optical powers (ROP) between 4 dBm and 10 dBm, demonstrating tolerance to fiber dispersion and nonlinearities.
Dense Edge Access (DSCM) Scenario: The converter was applied to a more complex scenario involving two 80-Gbps Digital Subcarrier Multiplexing (DSCM) channels, each containing four subcarriers (totaling eight users).
- Field Trial: The system was tested over a 3-km field-deployed campus fiber link at The Chinese University of Hong Kong.
- Performance: All eight sub-channels were successfully recovered using a single coherent receiver and SIC-based DSP. At an ROP of 7 dBm, all users achieved BERs below the 25% SD-FEC threshold. Constellation diagrams confirmed the successful recovery of both high-power and low-power users.

Significance and Claims
The paper claims that this orthogonal-to-non-orthogonal multiplexing format converter provides a practical pathway to enhance spectral efficiency and network flexibility in next-generation optical systems. By migrating part of the multiuser resource management from the electrical layer to the physical optical layer, the technology enables:

Spectral Reuse: Consolidating orthogonal wavelength slots into non-orthogonal channels frees up spectrum for additional traffic.
Scalability: Reducing receiver port counts and hardware complexity at aggregation nodes (OLTs and edge hubs) facilitates cost-effective scaling in dense network topologies.
Compatibility: The approach maintains interoperability with existing OFC sources, fiber links, and standard coherent detection systems, requiring no upgrades to user-end transmitters.

The authors position this work as a foundational step toward agile optical networks capable of handling complex topologies and dense populations, potentially extending beyond communications into optical computing and pattern recognition where rethinking channel orthogonality constraints may enable novel functionalities.