Model inversion of erbium-doped fiber amplifiers employing artificial intelligence

• Grant number: 23/00591-0
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): August 01, 2023
• Effective date (End): April 14, 2024
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Ivan Aritz Aldaya Garde
• Grantee: Ana Júlia Nora Francisco
• Host Institution: Faculdade de Engenharia. Universidade Estadual Paulista (UNESP). Campus Experimental São João da Boa Vista. São João da Boa Vista , SP, Brazil

Abstract

The need for greater data transmission capacity on the Internet can be alleviated with the deployment of infrastructures based on optical fibers. Optical fibers have a high link length per bandwidth product, but still require the use of optical amplifiers, essential devices in lightwave systems, to achieve the desired lengths. One such device is the erbium-doped fiber amplifier, which plays a central role in optical communication systems operating in the 1550 nm window. The fiber core of these amplifiers is doped with the element erbium during the manufacturing process and its fibers are optically pumped in order to excite its ions and promote optical gain. In wavelength division multiplexed systems, the signal beams compete for the excited ions and the gain experienced by each of them depends not only on its own intensity but also on the intensities of the other channels leading to so-called cross-gain modulation. One of the consequences of this characteristic is the difficulty of controlling the power of the channels at the output of the amplifiers, mainly in systems with heterogeneous power levels. Modeling this effect mathematically is complex and requires knowledge of many parameters. One solution is to use an artificial neural network that considers the interaction between the different channels of the system, as already studied during a previous scientific initiation research. In this project, this network model will be applied inversely in order to achieve constant output powers from the input powers of the amplifier. It will also be directly applied together with heuristic optimization algorithms, with the same objective, and the results obtained by both tests will be compared.