Fiber-Optic Transmission Systems Modeling
Software package modelling the single-pol and dual-pol fiber-optic channels governed by nonlinear Schrödinger equation (NLSE) and coupled NLSE, accompanied with TX and RX DSP chain implementation
Available on GitHub:
Developed by Abtin Shahkarami, FONTE ESR at the Institute Polytechnique de Paris, Telecom Paris, with contribution of Prof. Mansoor Yousefi, Prof. Yves Jaouen, and Jamal Darweesh, this library models the single-pol and dual-pol fiber-optic channels governed by nonlinear Schrödinger equation (NLSE) and coupled NLSE, accompanied with TX and RX DSP chain implementation.
NLS equation is solved using split-step Fourier method in this library.
The TX and RX DSP chain module, for single-polarization, includes the following modules:
- RRC Pulse shaping
- Matched filtering
- Chromatic dispersion compensation
- Digital back-propgation
- Carrier phase estimation
For dual-polarization systems, in addition to the modules above, the following modules are implemented:
- Polarization multiplexation and demultiplixation
- Radius-directed-equalization (RDE) -based multiple-input-multiple-output (MIMO) algorithm to compensate for the PMD
This work underlies publications by Abtin Shahkarami:
- A. Shahkarami, M. I. Yousefi, and Y. Jaouen, “Efficient Deep Learning of Nonlinear Fiber-Optic Communications Using a Convolutional Recurrent Neural Network,” in IEEE International Conference on Machine Learning and Applications (ICMLA), 2021, pp. 668-673, DOI: 10.1109/ICMLA52953.2021.00112.
- A. Shahkarami, M. Yousefi, and Yves Jaouen, “Attention-Based Neural Network Equalization in Fiber-Optic Communications.” in Asia Communications and Photonics Conference (ACP), Optical Society of America, 2021, p. M5H.3, ISBN: 978-1-957171-00-5