Single-channel 1.61 Tb/s Optical Coherent Transmission Enabled by Neural Network-Based Digital Pre-Distortion

Abstract

We propose a novel digital pre-distortion (DPD) based on neural networks for high-baudrate optical coherent transmitters. We demonstrate experimentally that it outperforms an optimized linear DPD giving a 1.2 dB SNR gain in a 128GBaud PCS-256QAM single-channel transmission over 80km of standard single-mode fiber resulting in a record 1.61 Tb/s net data rate.

54.5 Tb/s WDM Transmission over Field Deployed Fiber Enabled by Neural Network-Based Digital Pre-Distortion

“…In this paper, we extend our previous work [12] by developing a NN-based DPD technique to jointly precompensate for the distortions on the I and Q tributaries of the transmitter. The joint DPD is learned from "scratch" on a field-deployed 48 km link connecting industrial, business and academic parks in a metropolitan area.…”

“…DPD techniques based on neural networks (NN) [5] have recently received more attention [6][7][8][9][10], with notable simulation results in [8] by implementing NN-based DPD using direct learning architecture (DLA) for optical coherent systems. We recently demonstrated a NN-based DPD technique that independently mitigates the distortions of the "I" and "Q" tributaries on a single channel fiber-optic transmission system where an SNR improvement of up to 1.6 dB was achieved [12]. Another recent in-lab demonstration was of a total C-band capacity of 52.1 and 51.6 Tb/s over 80 km of TXF™ fiber and standard single mode fiber (SSMF), respectively [11].…”

Performance Versus Complexity Study of Neural Network Equalizers in Coherent Optical Systems

“…Convolution layers have also been found efficient in the analysis of temporal 1D sequences with several applications to time series sensors, audio signals, and natural language processing [45], [46]. For longer sequences, the CNN layer can be used as a pre-processing step due to its ability to reform the original sequence and extract its high-level features used for further processing cycles [24].…”

“…To that end, in the current study, we analyze how the combined NN architectures work for the optical channel equalization task. A simplified CNN+MLP combination was already successfully used in [24] at the transceiver for the high-baud-rate 80 km system.…”

“…a simple densely connected feedforward NN architecture), convolutional NNs (CNN) [13], [14], echo state networks (ESN) [15], and long short-term memory (LSTM) NNs [16], is efficient in improving optical system-level performance. However, the test of similar NN architectures in coherent optical systems has been carried out, mainly, numerically [17]- [20], or in short-haul experiments [21]- [24]. It is worth noticing that some very recent works evaluated the functioning of NN-based equalizers in metro/long-haul trials [4], [5], [8]- [10].…”