Performance Prediction Recipes for Optical Links
Erik Agrell, Marco Secondini, Alex Alvarado, Tsuyoshi Yoshida
TL;DR
This tutorial tackles the challenge of predicting end-to-end optical-link performance with forward error correction from data collected without FEC. It develops practical, implementable recipes across hard-decision, soft-decision, and probabilistic shaping scenarios, grounded in mismatched decoding and information-theoretic metrics such as BER, SER, AIR, and ASI. By enabling accurate pre-FEC predictions and RSNR estimation without actual FEC processing, the methods offer a scalable way to compare link configurations and forecast post-FEC performance, including thresholds and capacity-like bounds. The work emphasizes extending these metrics to more general channels, memory effects, and optimized receivers, providing a bridge between theory and deployable optical-system design. The practical impact is a suite of tools and thresholds that reduce the need for exhaustive FEC experiments while still delivering reliable performance predictions for modern optical links.
Abstract
Although forward error-correction (FEC) coding is an essential part of modern fiber-optic communication systems, it is impractical to implement and evaluate FEC in transmission experiments and simulations. Therefore, it is desirable to accurately predict the end-to-end link performance including FEC from transmission data recorded without FEC. In this tutorial, we provide ready-to-implement "recipes" for such prediction techniques, which apply to arbitrary channels and require no knowledge of information or coding theory. The appropriate choice of recipe depends on properties of the FEC encoder and decoder. The covered metrics include bit error rate, symbol error rate, achievable information rate, and asymptotic information, in all cases computed using a mismatched receiver. Supplementary software implementations are available.