Machine learning reduces noise in optical systems

DTU Fotonik uses machine learning techniques to reduce optical noise.

Black light

Optical applications such as fiber-optic sensing systems, gravitational wave detection, optical space communication, and optical fiber communication need high-power, narrow-linewidth lasers.

High power is obtained by amplifying the output power of a low-noise laser. However, the amplifiers induce fluctuations in the phase of the incoming optical signal, which causes spectral broadening and leads to poor system performance.

At DTU Fotonik, at the Technical University of Denmark, Darko Zibar and his group use machine learning techniques to significantly reduce noise in optical systems and develop a novel method to measure phase noise close to the quantum limit. They have described their work in “Approaching the optimum phase measurement in the presence of amplifier noise” in Optica.

For the experiment, the researchers chose a Koheras DFB fiber laser with narrow linewidth and ultra-low noise. And their new techniques are relevant in all applications where it is vital to maintain the low phase noise of the seed laser, such as quantum applications.

“At NKT Photonics, we look forward to working closely with Darko Zibar’s group and incorporate their work into our lasers to deliver even better products,” says CTO, Christian Vestergaard Poulsen.

Read the paper on Optica’s website.