Modal decomposition in multimode optical fibres by solving the challenges posed by phase ambiguity and large datasets

Abstract

Listen

Translate article

Mode decomposition (MD), which can obtain the amplitude and phase of each propagating mode in multimode optical fibres, provides important information and unleashes enormous potential applications in the fields of optical communication and optical sensing. Phase ambiguity and large datasets are two big challenges for mode decomposition based on deep-learning numerical methods. In this paper, the influence of phase ambiguity is eliminated by combining near-field and far-field intensity patterns. By designing a two-step hybrid process, the accuracy of mode decomposition is improved significantly. Compared with previous reported methods, our results demonstrate that even for the complicated 10-mode case, the average correlation coefficient (CC) between the reconstructed and the true intensity pattern can be optimized to over 0.99 using much smaller datasets with 10,000 stacked images.