News

More efficient data transfer in data centers

28.06.2024 - Researchers combine nonlinear predistortion and digital resolution enhancement to overcome DAC challenges.

Data centers stand as the cornerstone of modern information technology infra­structure. These centralized faci­lities are designed to store, process, and distribute vast amounts of data and applications, serving as the nerve center for digital services and businesses worldwide. From ensuring data availa­bility to supporting scala­bility, disaster recovery, and maintaining robust security measures, data centers play a critical role in enabling the seamless functioning of today's inter­connected digital landscape.

At the heart of data center operations lie data center interconnects, the vital networking infra­structure responsible for facilitating communi­cation between various components within and across data centers. Digital-to-analog converters (DACs) are indispensable components within these interconnects, tasked with converting digital signals into analog signals for transmission over copper cables. Their role in enabling high-speed, cost-effective, and low-latency connec­tivity cannot be overstated. However, the challenge arises from the requirement for high-resolution DACs, which poses a signi­ficant bottleneck due to the associated increase in costs of optical modules.

Addressing this challenge head-on, researchers have presented a groundbreaking solution that combines a look-up-table-based nonlinear predistortion technique with digital resolution enhancement. This innovative approach aims to alleviate the limitations imposed by high-resolution DACs while maintaining efficient data transfer and communi­cation within data center interconnects. The proposed technique has yielded remarkable experimental results, pushing the boundaries of what is achievable in terms of data transmission rates. By employing look-up-table-based predistor­tion to mitigate nonlinear impairment, and digital resolution enhancement to reduce the demand for DAC resolution, the research team has achieved record-breaking data transmission performance.

Notably, the digital signal processing technique enabled the transmission of signals at rates exceeding 124 GBd PAM-4/6 and 112 GBd PAM-8 over 2 kilometers of standard single-mode fiber using 3/3.5/4-bit DACs. Addi­tionally, it faci­litated the transmission of 124 GBd PAM-2/3/4 signals over 40 km of standard single-mode fiber using 1.5/2/3-bit DACs. These results represent a significant advancement in data center inter­connect technology, demonstrating the feasibility of supporting the next generation of ethernet links targeting speeds of up to 800-GbE or potentially even 1.6-TbE.

Zhaopeng Xu of Peng Cheng Laboratory underscores the signi­ficance of these findings, highlighting that they demonstrate “the transmission of the highest data rates with the lowest-cost digital-to-analog converters for data center inter­connects.” Beyond revo­lutionizing data center inter­connects, these advancements hold promise for transforming various appli­cations across 6G access networks and passive optical networks. By overcoming the challenges associated with high-resolution DACs, this innovative approach also paves the way for more cost-effective and efficient data transmission. (Source: SPIE)

Reference: Q. Wu et al.: Beyond 200-Gb/s O-band intensity modulation and direct detection optics with joint look-up-table-based predistortion and digital resolution enhancement for low-cost data center interconnects, Adv. Phot. Nexus 3, 036007 (2024); DOI: 10.1117/1.APN.3.3.036007

Link: Peng Cheng Laboratory, Shenzhen, China

Top Feature

Digital tools or software can ease your life as a photonics professional by either helping you with your system design or during the manufacturing process or when purchasing components. Check out our compilation:

Proceed to our dossier

Top Feature

Digital tools or software can ease your life as a photonics professional by either helping you with your system design or during the manufacturing process or when purchasing components. Check out our compilation:

Proceed to our dossier