The capacity crunch and the evolution towards autonomous and plug-and-play optical systems require of research, development and implementation for the requirements of this paradigm shift.

This PhD thesis addresses the problem of noise estimation in optical communication systems. This topic is of critical importance since a better understanding of the optical link is required to optimize the transmission parameters with the final goal of maximizing capacity and enabling autonomous networks.

The focus on this work is online metrology, where measurements are taken in conjunction with live traffic enabling to detect malfunctioning links and discrepancies between theoretical models and measured results. After covering the theoretical and experimental background, and a review of the state of the art, several noise estimations methods are described classified as frequency domain and time domain approaches: depending on their description domain; and transparent and invasive methods: whether they affect performance or not.

The original contributions to knowledge of this work are the development of a time domain transparent method based on noise correlation to estimate noise requiring a translation layer (in this case a neural network), a frequency domain invasive method for the estimation of linear and nonlinear noises based on a least-square fitting approach which does not need of a translation layer, and a method to calibrate transceiver noise contributions based on perturbations was also developed, capable of the estimating uncompensated transceiver impairments. Different measurement interfaces are possible, allowing to estimate transceiver noise contributions in different locations.

The ideas developed in this work were conceptualized and described by the author, and verified by simulations and experiments to support the claims of this work. A variant of the time domain noise estimates is currently implemented in Ciena’s WaveLogic Family. The frequency domain noise separation idea and transceiver calibration idea are currently being patented and begin considered for implementation by the Industrial sponsor. Help and support from the industrial sponsor was provided during the development of this work.