Posted in | News | Laser | Fibre Optics

Minimizing Noise Using Machine Learning

Ultra-precise lasers find their use in various applications. For example, power cable monitoring, quantum computers, optical atomic clocks and much more. However, all lasers tend to make noise, which scientists from DTU Fotonik wish to reduce with the help of machine learning.

Minimizing Noise Using Machine Learning.

Image Credit: Technical University of Denmark.

There is nothing called an ideal laser. There will always be some phase noise since the laser light frequency moves back and forth. Phase noise avoids the laser from generating light waves with the ideal steadiness that is usually a laser’s characteristic feature.

The majority of the lasers used on a day-to-day basis do not need to be entirely accurate. For instance, it is of no significance if the frequency of the red laser light in the supermarket barcode scanners changes slightly when the barcodes are being read. However, for a few applications — for instance, in optical measuring instruments and optical atomic clocks — it is completely crucial that the laser is stable so that there is no variation in the light frequency.

One method of getting closer to an ultra-precise laser is if users can find the phase noise. This might allow users to determine an approach of compensating for it so that the result is a purer, more precise laser beam.

This is exactly what Professor Darko Zibar from DTU Fotonik is working on. He guides a research group known as Machine Learning in Photonic Systems, whose aim is to develop and make use of machine learning to enhance optical systems. In recent times, scientists from the group have characterized the noise from a laser system from the Danish company NKT Photonics with unparalleled accuracy.

The question is how to measure that noise, and here weve developed the most accurate method available. We can measure much more precisely than others — our method has record-high sensitivity.

Darko Zibar, Professor and Group Leader, DTU Fotonik, Technical University of Denmark

An algorithm has been developed by Zibar that can examine and determine laser light patterns with the help of machine learning, where a model for the noise is constantly being enhanced. Based on this, the team believes to be able to design a form of intelligent filter that constantly cleans the laser beam of noise.

Quantum Mechanics Set the Limit

This is something that NKT Photonics could use in their optical measuring instruments, states Senior Researcher Poul Varming and his collaborator Jens E. Pedersen, who has worked with the DTU scientists:

We work with fiber lasers that emit constant light, and where the noise level is particularly low. Our most important task is to limit the noise, and — in terms of measuring technology — we had difficulty measuring noise at very high frequencies.

Poul Varming, Senior Researcher, Technical University of Denmark

Varming added, “But then we got in touch with Darko Zibar and his group, and we produced some lasers for them. The researchers were able to measure the noise up to very high frequencies, and the results actually contradict the established understanding of laser noise.”

With the help of a new, enhanced measuring technique, the scientists could exhibit that the theoretical basis for evaluating the noise was not in position. Having more elaborate knowledge of the noise, engineers could better determine the parts of the laser system from which the noise emanates. Thus, they are well aware of where to make enhancements. The scientists also believe that the machine learning system can be utilized to attenuate the noise in real-time.

Since the laws of quantum mechanics fix a highly fundamental limit to how good a laser could be, it is not possible to eliminate noise completely. According to Darko Zibar, it is impossible to get rid of quantum noise, but it can at least be quantified for now.

We can measure in the frequencies in which quantum noise is dominant. In this way, we can determine the fundamental noise and find out how much it contributes to the total noise. Once we know the fundamental limit for how good the laser can be, we can then figure out how to suppress the rest of the noise.

Darko Zibar, Professor and Group Leader, DTU Fotonik, Technical University of Denmark

Zibar added, “This is our next project — how we first identify and then suppress the noise, to obtain a laser that is only limited by quantum noise. This will enable us to produce some of the best lasers in the world.”

Optical Cable Feels Vibrations

When the laser noise is heard, it can be battled with the same principle as that has been utilized in noise-reducing headphones. In this context, microphones take up sound from the environment, and a signal is then transmitted in counter phase to the speakers so that the noise and the new signal remove each other, and it leads to silence.

If the method could be utilized to enhance lasers by removing a large part of the noise so that the light frequency virtually does not get altered, optical measuring instruments can exhibit a longer range and a greater sensitivity.

At NKT Photonics, the technology could originally be utilized for distributed acoustic sensing, where a fiber optic cable is utilized as a sensor for quantifying small vibrations. Distributed acoustic sensing can be utilized for several forms of monitoring.

For instance, an optical fiber could be laid ahead of an oil or gas pipeline to guarantee ultrafast detection of any ruptures. Also, the technology can be utilized to track the fence around an airport or at a border (if a hole is cut in the fence, or someone tries to climb over it) the technology can not only signal what has taken place but also locate where it has occurred.

Such an optical monitoring system works by a laser beam being transmitted into the optical fiber. At the time of the process, a bit of the light has been reflected by small impurities in the fiber. But if the fiber has been impacted along the way, the properties of the reflected light also vary, which can be measured. Even very faint vibrations could be picked up and pinpointed with great precision.

Monitoring of Cables to the Energy Islands

If the new technology from DTU offers highly effective laser light noise attenuation, distributed acoustic sensing could be utilized over somewhat longer distances compared to that of today. The range and sensitivity of distributed acoustic sensing can be raised with the more accurate lasers, and this might — for instance — be required when electricity is sent from the coming energy islands in the North Sea to the mainland.

In this context, the power cables can be tracked using the technology, so that any ruptures could be detected and repaired rapidly. At present, it is a shortcoming that the range of the current systems is restricted to a maximum of 50 km, and the distance to the energy island will seem to be a little longer.

Furthermore, Poul Varming mentions that several quantum technologies need highly accurate lasers. With noise-attenuated lasers, it turns out to be simpler to design ultra-precise optical atomic clocks and a few kinds of quantum computers, where lasers are utilized to cool individual atoms to near to absolute zero. The new generation of laser systems that might be the result of the researchers’ and engineers’ work thus provides great potential.


Tell Us What You Think

Do you have a review, update or anything you would like to add to this news story?

Leave your feedback
Your comment type