Accelerating Assessments of Optical Components Using Machine Learning: TDECQ as Demonstrated Example

Transmitter and dispersion eye closure quaternary (TDECQ) penalty has replaced eye mask and transmitter dispersion penalty (TDP) methodologies for qualifying PAM-4 transmitters. TDECQ correlates better than the eye mask test with BER, and its implementation of receiver equalizers abstracts the metric from receiver features, enabling robust and systematic component assessments. However, assessing TDECQ is computationally intensive and time-consuming. Here, we present the use of machine learning (ML) to dramatically accelerate TDECQ assessments of PAM-4 transmitter signals. Explored techniques include convolutional neural networks and long short-term memory. We demonstrate that these methods provide comparable assessment accuracies compared to the conventional method, while tremendously reducing computational time. Some ML methods were similar to 4500 times faster than the conventional method. Described architectures are generic and can be modified to accelerate any class of optical component assessments.

Automated summary

TDECQ outperforms eye mask and TDP methodologies in evaluating PAM-4 transmitters, and using machine learning techniques can significantly enhance assessment speed while maintaining accuracy.