Hardware-Efficient Duobinary Neural Network Equalizers for 800 Gb/s IM/DD PAM4 Transmission Over 10 km SSMF

Article Engineering, Electrical & Electronic

Feedforward and Recurrent Neural Network-Based Transfer Learning for Nonlinear Equalization in Short-Reach Optical Links

Zhaopeng Xu et al.

Summary: In this article, transfer learning (TL)-aided feedforward neural networks (FNN) and recurrent neural networks (RNN) are proposed for nonlinear equalization in short-reach direct detection optical links, allowing for a fast transition to new equalizers when the channel condition changes. TL significantly reduces the training requirements of FNNs/RNNs in the target system, with a substantial decrease in epochs and training symbols. Additionally, it is found that FNNs can be effectively transferred to their corresponding RNNs for equalization in the target system.

JOURNAL OF LIGHTWAVE TECHNOLOGY (2021)

Add to Collection

Article Engineering, Electrical & Electronic

Transfer Learning for Neural Networks-Based Equalizers in Coherent Optical Systems

Pedro J. Freire et al.

Summary: In this work, the adaptability of artificial neural networks (NNs) used for impairments mitigation in optical transmission systems is addressed through transfer learning techniques. The study demonstrates the effectiveness of retraining NN-based equalizers to adapt to changes in the transmission system with just a fraction of the initial training data or epochs. Transfer learning proves to be efficient in adapting NN architectures to different transmission regimes and scenarios, showing promise for engineering flexible and universal solutions for nonlinearity mitigation in coherent optical communication systems.

JOURNAL OF LIGHTWAVE TECHNOLOGY (2021)

Add to Collection

Article Engineering, Electrical & Electronic

Nonlinear Differential Coding for Spectral Shaping of PAM Signal in High-Baudrate Short-Reach Optical Transmission

Shuto Yamamoto et al.

Summary: This article discusses a spectral-shaping technique for PAM signals using nonlinear differential coding to enhance signal transmission performance and tolerance to bandwidth limitations, with experimental verification of signal transmission performance.

JOURNAL OF LIGHTWAVE TECHNOLOGY (2021)

Add to Collection

Article Engineering, Electrical & Electronic

Transmission Beyond 200 Gbit/s With IM/DD System for Campus and Intra-Datacenter Network Applications

Md Sabbir-Bin Hossain et al.

Summary: This study compares experimental performance of PAM-4, PAM-6, and PAM-8 modulation formats with and without THP and VNLE in the transmitter-side and receiver-side. The results demonstrate that these signal processing schemes can increase tolerance against ISI, which is critical for achieving high-speed data rates.

IEEE PHOTONICS TECHNOLOGY LETTERS (2021)

Add to Collection

Article Engineering, Electrical & Electronic

Soft-Demapping for Short Reach Optical Communication: A Comparison of Deep Neural Networks and Volterra Series

Maximilian Schaedler et al.

Summary: This study explores the use of soft deep neural network as an alternative for nonlinear equalization in optical fiber communication, showing a 65% reduction in computational complexity compared to Volterra series at equal performance. At equal complexity, there is a 0.35 dB improvement in optical signal-to-noise ratio (OSNR).

JOURNAL OF LIGHTWAVE TECHNOLOGY (2021)

Add to Collection

Proceedings Paper Computer Science, Information Systems

Multi-Symbol Output Long Short-Term Memory Neural Network Equalizer For 200+Gbps IM/DD System

Bohan Sang et al.

Summary: The proposed single-lane 212Gbps IM/DD PAM-4 system, equipped with a novel Multi-Symbol Output LSTM equalizer, outperforms FFE & VNE and single-symbol output LSTM, reducing complexity by 49.85% while achieving similar performance with around 1/4 complexity of Bi-directional LSTM.

2021 EUROPEAN CONFERENCE ON OPTICAL COMMUNICATION (ECOC) (2021)

Add to Collection

Proceedings Paper Optics

Machine-learning-based equalization for short-reach transmission: neural networks and reservoir computing

F. Da Ros et al.

Summary: The substantial increase in communication throughput driven by machine-to-machine communication within and between data centers is straining short-reach communication links. Researchers are investigating equalization techniques for intensity-modulation/direct-detection transmission, focusing on addressing issues like inter-symbol interference introduced by fiber dispersion. Complex approaches involving maximum likelihood sequence estimation and machine-learning tools, such as neural networks, are being explored to improve performance.

METRO AND DATA CENTER OPTICAL NETWORKS AND SHORT-REACH LINKS IV (2021)

Add to Collection

Article Engineering, Electrical & Electronic