Mitigating nonlinear interference through constellation shaping
Sebastiaan Goossens defended his PhD thesis at the Department of Electrical Engineering on January 22nd.
Optical fiber communication systems are the backbone of the internet as we know it today. Whether streaming video, calling friends, or texting, nearly all data travels through optical fibers at some point in its journey. The demand for online services has driven an exponential increase in global communication capacity, necessitating continuous research into new technologies and improvements. While adding more optical fiber links is an option, it is expensive and time-consuming. A more cost-efficient solution involves upgrading existing fiber connections with improved transmitters and receivers. Increasing optical power at the transmitter is one way to allow increasing the performance of the system. However, for higher powers, nonlinear effects occur in the optical fiber, degrading the system performance. One promising technique to address this challenge is constellation shaping, which optimizes the transmitted light to a specific channel, increasing power efficiency. For his PhD research Sebastiaan Goossens investigated whether constellation shaping can also mitigate nonlinear effects encountered during optical fiber transmission. By enabling communication at higher optical powers, this approach could significantly improve system performance.
Nonlinear interference noise (NLIN), caused by the optical Kerr effect and chromatic dispersion, is one of the primary limitations in optical fiber communication systems. Although computationally intensive digital signal processing (DSP) techniques, such as digital backpropagation (DBP), can partially compensate for NLIN, they cannot fully address the issue due to model inaccuracies and unpredictable signal-noise interactions. To mitigate NLIN, optimizing the transmitted modulation format through constellation shaping is a promising strategy. In his research Sebastiaan Goossens focused on two forms of constellation shaping: probabilistic shaping (PS), which adjusts the probabilities of constellation points, and geometric shaping (GS), which alters the geometry of the points.
Key contributions
Four main contributions are presented in his thesis. First, the performance of two finite-blocklength PS algorithms was experimentally evaluated in a long-haul wavelength division multiplexing (WDM) setup, confirming a theoretical observation about blocklength dependency on nonlinear tolerance. Second, a framework was developed to optimize 4D GS constellations for nonlinear optical channels, reducing the optimization space through energy shell discretization and symmetry constraints. Third, GS techniques were optimized for high cardinality (up to 17 bit/4D symbol) in finite signal-to-noise ratio (SNR) additive white Gaussian noise (AWGN) channels. Finally, the thesis optimized 4D GS constellations (up to 12 bit/4D symbol) for realistic optical communication system models, addressing current and future standards for transceivers supporting 1.6 Tbps and 3.2 Tbps links.
Title of PhD thesis: . Supervisors: Prof. Alex Alvarado, and Dr. Chigo Okonkwo.