| MS TSA Meeting

Name of the Speaker: Mr. S VISWANATHAN (EE21S075)
Guide: Prof. Balaji Srinivasan
Venue: ESB-244 (Seminar Hall)
Online meeting link: https://meet.google.com/nsd-mmnv-uxa
Date/Time: 17th May 2024 (Friday), 10:00 AM
Title: Scalable Phase Control Algorithm for Tiled Aperture Coherent Beam Combining

Abstract :

The pursuit of high-power continuous wave (CW) laser sources has spurred extensive research endeavors, aiming to achieve remarkable output levels while upholding stringent beam quality. This pursuit faces a significant hurdle: maintaining diffraction-limited beam quality amidst thermal fluctuations and optical non-linearities in the gain medium, particularly when targeting power levels in the order of 100 kWs from a single emitter. Coherent Beam Combining (CBC) emerges as a transformative solution, offering a pathway to enhance output power efficiency while preserving beam quality. Unlike conventional incoherent beam combining, which linearly scales beam intensities, CBC leverages the intricate combination of beam fields, resulting in a remarkable quadratic scaling of intensity with the number of beams.

At the heart of CBC's effectiveness lies the challenge of implementing active phase control. The coherent combination of the optical beams is fraught with perturbations from various factors, impeding their ability to coherently interfere and compromising combining efficiency. Dynamic phase control emerges as a crucial strategy, enabling real-time compensation for phase fluctuations and thereby enhancing combining efficiency. The seminar initially delves into modeling and simulation of phase control using conventional Stochastic Parallel Gradient Descent (SPGD) for a Tiled Aperture CBC system. It further develops an adaptive SPGD version, showcasing improved performance with increasing channel numbers. Subsequently, practical implementations of these algorithms within an experimental setup provide valuable insights into real-world performance and applicability. Lastly, the thesis explores Reinforcement Learning (RL) as an alternative to SPGD for phase control optimization, aiming to address challenges of phase noise, scalability, and efficiency in CBC systems and foster more robust and adaptable control strategies.