Orthogonality among channels is a canonical basis for optical multiplexing, but it limits capacity. This paper reports on non-orthogonal optical multiplexing over a multimode fiber (MMF) using a deep neural network (SLRnet) to learn the mapping between multiple non-orthogonal input light fields and a single intensity output. Experiments show SLRnet effectively retrieves multiple non-orthogonal input signals with high fidelity (~98%), suggesting a path towards high-capacity optical multiplexing.

Tuqiang Pan, Jianwei Ye, Haotian Liu, Fan Zhang, Pengbai Xu, Ou Xu, Yi Xu, Yuwen Qin