Significant Breakthrough in Optical Transmission: Single-Mode Fiber Exceeds 254.7 Tb/s

Recently, there was a major technological breakthrough in optical communications:

“Ultra-high-speed transmission of 254.7 Tb/s over 200 km of standard single-mode optical fiber”

This transmission metric has set a new world record for optical fiber communication.

Let’s take a closer look at the details:

• Research partners: FiberHome, in collaboration with China Mobile, the National Key Laboratory of Optical Communication Technology and Networks, Pengcheng Laboratory, and other institutions.
• Ultra-high-speed transmission of 254.7 Tb/s over 200 km of standard single-mode optical fiber, based on transfer learning neural network equalization technology.
• This technology covers a 19.8 THz spectral bandwidth and achieves a spectral efficiency of 12.86 b/s/Hz, which will have a profound impact on the architectural design, application scenarios, and industry ecosystem of future optical networks.

Traditional optical communication systems have always sought a balance within the “capacity-distance-efficiency” triangle. However, this balance is now being broken by the data deluge driven by AI.

The optical transmission system achieved three core technological breakthroughs

1. Ultra-High Capacity Transmission: By introducing dense wavelength-division multiplexing (DWDM) and multi-band collaborative transmission (S+C+L), the system covers a 19.8 THz spectral bandwidth, which is more than four times that of traditional systems. To ensure uniform signal power across all bands, the system integrates advanced spectral flattening technology and a hybrid amplification scheme, achieving a spectral efficiency of 12.86 b/s/Hz.

2. Breakthrough in Transmission Quality: The system is designed to maintain excellent signal quality even after 200 km of transmission. It achieves this by combining advanced polarization multiplexing technology with a probabilistic shaping high-order modulation scheme to precisely optimize signal entropy. Additionally, it uses a combination of hybrid amplification and Raman amplification technologies to effectively mitigate signal attenuation and nonlinear effects, ensuring transmission stability.

3. Optical Transmission + AI Breakthrough: The use of AI algorithms, specifically transfer learning neural network equalization technology, boosted the system’s total throughput by 11.7%. Traditional digital signal processing algorithms face a precision bottleneck when compensating for complex nonlinearities. However, the AI-based algorithm allows each band to have its own dedicated NN model for signal equalization, compensating for and correcting nonlinear errors while significantly reducing system power consumption.

Key Technology: AI + Optical Communications

The core of this research is using a neural network (NN) to correct errors in the optical fiber transmission system. Traditional optical fiber systems rely on mathematical models and hardware circuits to compensate for nonlinear effects, but their efficiency and precision are often limited. This AI algorithm is able to boost the system’s efficiency more effectively.

• Improved Efficiency: Traditional methods require training independent NN models for the S, C, and L bands, which demands large amounts of data and long training periods. This research uses a “train a baseline model, then transfer key parameters” strategy. This allows the neural network to apply what it learned from C-band training to the S and L bands, reducing data requirements by 70% and shortening the training period by 50%. This breakthrough solves the pain points of “difficult data collection and slow model deployment” in optical communications, accelerating the adoption of NN equalizers in live networks.
• Enhanced Generalization: After transferring C-band parameters, the S and L band models maintain stable performance even when faced with different wavelength characteristics, power distributions, and nonlinear environments. The L band, after introducing the NN equalizer, saw a 12.3% increase in net data rate, which is comparable to the performance boost in the C band.

This domestic technological breakthrough will lead optical networks into the P-bit era. This is likely more than just a technical achievement; it could also influence the future direction and landscape of network development. With the deep integration of AI technology and optical communications, continuous advancements in multi-band expansion, algorithm optimization, and hardware integration will truly usher in an “intelligent, ultra-wide, and ultra-high-speed” P-bit era for optical networks.