CNN-based eye pattern analysis and BER prediction in PAM4 inter-datacenter optical connections impaired by intercore crosstalk

• Autor(a) principal: Esteves, Sofia Pérsio Eugénio
• Data de Publicação: 2021
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10071/24298
• topic: Bit error rate; Convolutional neural network; Inter-core crosstalk; Machine learning; Multicore fiber; Aprendizagem automática; Crosstalk entre núcleos; Fibras multinúcleo; Rede neuronal convolucional; Taxa de erros binários

To meet the required future challenge of providing enough bandwidth to achieve high data traffic rates in datacenter links, four-level pulse amplitude modulation (PAM4) signals transmission in short-haul intensity modulation-direct detection datacenters connections supported by homogeneous weakly-coupled multicore fibers has been proposed. However, in such fibers, a physical effect known as inter-core crosstalk (ICXT) limits significantly the performance of short-reach connections by causing large bit error rate (BER) fluctuations that can lead undesirable system outages. In this work, a convolutional neural network (CNN) is proposed for eye-pattern analysis and BER prediction in PAM4 inter-datacenter optical connections impaired by ICXT, with the aim of optical performance monitoring. The performance of the CNN is assessed using the root mean square error (RMSE). Considering PAM4 interdatacenter links with one interfering core and for different skew-symbol rate products, extinction ratios and crosstalk levels, the results show that the implemented CNN is able to predict the BER without surpassing the RMSE limit. The CNNs trained with different optical parameters obtained the best performance in terms of generalization comparing to CNNs trained with specific optical parameters. These results confirm that the CNN-based models can be able to extract features from received eye patterns to predict the BER without prior knowledge of the transmitted signals.