The Latest Trends in Backbone Network Optical Communications

NADDOD Peter Optics Technician May 23, 2024

400G, it’s really here.

 

Starting last year, Chinese telecommunications operators began a full-scale rollout of 400G in backbone networks. Following extensive commercial validation in 2023 and the initiation of large-scale procurement, 2024 marks the official commencement of widespread commercial deployment. In March 2024, China Mobile Communications Group Co., Ltd. (CMCC) launched the world's first 400G all-optical inter-provincial backbone line between Beijing and Inner Mongolia, a milestone event.

 

The reasons for upgrading the backbone network to 400G are clear. On one hand, the growth in consumer internet traffic driven by digital lifestyles—such as high-definition video, remote conferencing, online streaming, and gaming—continues unabated. On the other hand, industry-wide digital transformation is causing a surge in traffic from industrial digital systems, further straining the backbone network.

 

A significant contributor to this increased pressure on backbone networks is the AI boom. The rise of generative AI has created a surge in AI-driven applications. To support these AI operations, numerous intelligent computing centers need to be built. As models scale from billions to trillions of parameters, GPU clusters are also expanding from thousands to tens of thousands of GPUs.

 

GPU clusters are essentially arrays of GPU servers interconnected via high-performance networks like InfiniBand or RoCEv2. These networks demand high performance and reliability, directly impacting training efficiency and costs. Current GPU server network port rates start at 400G, with 800G or higher becoming increasingly necessary.

 

Network ports of GPU servers

 

Previously, GPU clusters were confined to data center networks (DCN). Now, as clusters grow, there's a move towards using distributed intelligent computing centers for model training, necessitating more robust data center interconnect (DCI) networks. The backbone optical network must meet these technical requirements.

 

China’s strategic approach to computational power involves a coordinated, nationwide plan. Initiated in February 2022, the "East-to-West Computing" project aims to create an integrated national computing power system. This involves building numerous data centers (akin to power plants) and a robust backbone transmission network (akin to the power grid) to ensure the seamless flow of computational power across industries.

 

To address these evolving demands, NADDOD offers advanced 800G/400G solutions that provide the necessary bandwidth and reliability for modern backbone networks. These solutions are critical for supporting both current and future data transmission needs in high-performance computing and AI-driven environments.

 

How is 400G Achieved?

The current backbone optical network, as the foundation of our digital society, must offer ultra-high bandwidth (400G, and in the future, 800G or even 1.6T), ultra-low latency, large-scale networking, high stability, reliability, security, flexible deployment, and intelligent operation and maintenance.

 

Today's focus is on the most critical aspect: speed and bandwidth. Achieving higher speeds in optical communications involves improvements in a few key areas:

 

  • Baud Rate

 

Transmission speed, or bit rate, is the number of bits transmitted per unit time, measured in bits per second (bit/s).

 

Bit rate = Baud rate × Number of bits per modulation symbol

 

Baud rate is the number of symbols transmitted per unit time. The higher the baud rate, the more symbols transmitted per second, resulting in higher information throughput.

 

Baud rate is determined by the capabilities of optical components. Advances in chip manufacturing processes have increased the baud rate from 30+ Gbaud to 64+ Gbaud, 90+ Gbaud, and now 128+ Gbaud. The commercial viability of 400G is attributed to achieving a baud rate of 128Gbaud.

 

  • Modulation Methods

 

The "number of bits per modulation symbol" in the formula is determined by the modulation method.

 

Main Parameters

QPSK

16QAM

64QAM

256QAM

1K QAM

4K QAM

Number of bits per symbol

2

4

6

8

10

12

Total points in the constellation diagram

4

16

64

256

1024

4096

Multiplier of QPSK capacity

-

2x

3x

4x

5x

6x

Modulation Schemes Comparison

 

For 400G technology, the main modulation schemes are 16QAM, 16QAM-PCS (probabilistic constellation shaping), and QPSK, each suitable for different application scenarios.

 

400G Modulation Methods

 

Unlike wireless communications, optical communications do not always pursue higher-order modulation. Lower-order modulation reduces the complexity and cost of the network. Initially, long-distance backbone networks focused on 16QAM and QPSK. With the advent of 16QAM-PCS, it also became a viable option.

 

Earlier, operators believed 400G wouldn't require long-distance transmission, so using lower-cost, mature low-baud rate components with higher-order modulation like 16QAM was the industry consensus. However, the need for longer transmission distances and the rapid maturation of 128Gbaud components (driven by the rise of 800G in DCN scenarios) have made QPSK increasingly viable.

 

QPSK is more tolerant to nonlinearity and can increase input power compared to 16QAM-PCS. Additionally, QPSK has a better OSNR threshold and a channel spacing of 150GHz, minimizing filtering penalties during transmission. These advantages have made QPSK the preferred choice for backbone and DCI networks.

 

 

Modulation Scheme

Channel Spacing

Baud Rate

Transmission Distance

16QAM 400G

75GHz

64GBd

~600km

16QAM-PCS 400G

100GHz

90GBd

~1000km

QPSK 400G

150GHz

128GBd

~1500km

Transmission Parameters for 400G

 

  • Extending wavelength Bands

The baud rate and modulation mainly impact single-wavelength rates. A single optical fiber can carry multiple wavelengths, as long as the spectral range is sufficient.

 

Total fiber bandwidth = Single wavelength bandwidth × Number of wavelengths per fiber

 

QPSK 400G has a channel spacing of 150GHz. Traditional C-band and extended C-band alone are insufficient to meet the spectral bandwidth requirements.

 

Therefore, a C6T+L6T approach is increasingly adopted, providing a total spectral bandwidth of 12THz. With 80 wavelengths at 400G each, the total capacity per fiber reaches 32T. For provincial networks, sacrificing some distance and deploying QPSK or 16QAM-PCS can further increase the capacity to 48T.

 

The main challenge with extending wavelength bands is whether the components can support it and whether the costs are manageable. These components include ITLA (integrable tunable laser assembly), CDM (coherent detection module), ICR (integrated coherent receiver), EDFA (erbium-doped fiber amplifier), and WSS (wavelength selective switch), which are crucial for optical transmission, switching, and amplification.

 

Another issue with extending wavelength bands is integration. Currently, band expansion resembles two independent systems (C and L) simply combined. These systems operate independently, transmitted together by wavelength multiplexing, and then separated at the receiving end for further processing.

 

Using two independent systems increases the size, power consumption, and complexity. The industry needs to research how to integrate these components, allowing a single system to support different extended wavelength bands—achieving true integration.

 

  • Fiber Optical Technology

 

Besides optical modules and equipment, attention should be given to fiber optics. The current mainstream fiber is G.652D. With EDFA amplification, 400G QPSK can be transmitted up to 1500km on G.652D fiber.

 

Extensive industry validation has identified G.654E fiber as the new successor. Using G.654E fiber, which offers superior performance, can increase the transmission distance of 400G QPSK by over 30% under the same conditions.

 

G.654E fiber is ready for large-scale deployment on long-haul routes and is preferred for transoceanic submarine cables due to its low-loss characteristics.

 

In addition to traditional fiber, the industry sees broad prospects for multi-core fiber and hollow-core fiber. Multi-core fiber significantly increases capacity by packing more cores into a single fiber using fewer modes.

 

Hollow-core fiber, which replaces the glass core with air, is even more advanced. It offers greater capacity, lower latency, reduced transmission loss, and ultra-low nonlinearity, making it one of the most promising technologies in optical communications.

 

Beyond 400G: 800G or 1.6T?

With the commercial deployment of 400G, the industry's focus shifts to technologies beyond 400G. The debate continues on whether the next step should be 800G, 1.2T, or 1.6T. Achieving higher speeds will require advancements in modulation methods and baud rates, with 130GBd or even 260GBd being the likely targets. Higher baud rates necessitate the development of mature components and a robust supply chain.

 

Beyond 400G, QPSK is no longer viable. The industry favors 16QAM modulation as the preferred choice. Further expansion of the spectral band is needed, considering S-band, U-band, and E-band in addition to C and L bands. Combining C+L+S bands provides a spectral bandwidth of 17THz, making 100Tbps transmission over a single fiber achievable.

 

In data centers, 800G (based on a 100GBd baud rate and 100G per channel) is already commercially available. Single-channel rates of 200G, 400G, and 800G are only a matter of time.

 

As capacity increases, so do technical challenges. The development of optical communications relies on advancements in components, chips, manufacturing processes, and materials. Meeting requirements for power consumption, security, and operations will depend on innovations in processes, architecture, packaging, artificial intelligence, and digital twins. There is still much work to be done across the supply chain, and the journey ahead is long.

 

NADDOD's Solution

NADDOD, a professional provider of innovative optical networking solutions to HPC, AI, data centers, enterprises, and telecom customers, is also a leading global transceiver manufacturer. We offer a comprehensive range of 800G/400G solutions, including various fiber cables, OSFP transceivers, and more.

 

Our cutting-edge solutions are designed to meet the growing demands of modern network infrastructures and to support the transition to higher speeds and capacities beyond 400G.

 

NADDOD 400G Optical ModuleNADDOD's 400G/800G OSFP Optical Transceivers with Dual Broadcom Chips

 

Final Thoughts

Optical communications are the digital arteries of our society. Despite skepticism about other technologies (including 5G), the necessity of optical communications has never been questioned, as it is essential for societal development.

 

The trend of increasing data traffic will continue for decades, amplified by the rapid rise of artificial intelligence. Current developments in optical communications cannot meet this growing demand, driving companies to invest more in research and development for profitable returns.

 

NADDOD's advanced 400G/800G solutions are at the forefront of this technological evolution, ensuring that the backbone of our digital society remains robust, efficient, and capable of handling future demands. Check NADDOD's website to explore our cutting-edge products and stay ahead in the rapidly evolving world of optical communications.

 

NADDOD Brand