AI-optimised tuneable sources for bandwidth-scalable, sub-nanosecond wavelength switching

Area of Science:

• Photonics and Optical Communications
• Data Centre Networking
• Artificial Intelligence in Engineering

Background:

• Wavelength routed optical switching offers potential for low power and latency in data centres.
• A critical requirement is a wideband wavelength tuneable source (WTS) with sub-nanosecond switching capabilities.
• Existing WTS technologies face challenges in achieving both wide bandwidth and rapid switching.

Purpose of the Study:

• To propose and demonstrate a novel hybrid WTS architecture.
• To achieve record-breaking wavelength switching times and wide optical bandwidth.
• To explore methods for further enhancing bandwidth scalability in optical switching.

Main Methods:

• A hybrid WTS design combining time-interleaved tuneable lasers and semiconductor optical amplifiers.
• Optimization of individual device performance using artificial intelligence (AI) algorithms.
• Experimental validation and simulation of the WTS performance, including switching time and bandwidth.

Main Results:

• Demonstrated record wavelength switch times below 900 picoseconds (ps).
• Achieved a continuously tuneable optical bandwidth of 6.05 THz (equivalent to 122 channels of 50 GHz).
• Evaluated a novel method for scaling the optical bandwidth.

Conclusions:

• The proposed hybrid WTS meets the critical requirements for next-generation data centre optical switching.
• AI-driven optimization significantly enhances WTS performance, achieving unprecedented switching speeds and bandwidth.
• The demonstrated technology and bandwidth scaling methods pave the way for more efficient and faster optical networks.