13 August 2018 in Journal of Lightwave Technology
In this paper, we implement an optical fiber communication system as an end-to-end deep neural network, including the complete chain of transmitter, channel model, and receiver. This approach enables the optimization of the transceiver in a single end-to-end process. We illustrate the benefits of this method by applying it to intensity modulation/direct detection (IM/DD) systems and show that we can achieve bit error rates below the 6.7% hard-decision forward error correction (HD-FEC) threshold. We model all componentry of the transmitter and receiver, as well as the fiber channel, and apply deep learning to find transmitter and receiver configurations minimizing the symbol error rate. We propose and verify in simulations a training method that yields robust and flexible transceivers that allow-without reconfiguration-reliable transmission over a large range of link dispersions. The results from end-to-end deep learning are successfully verified for the first time in an experiment. In particular, we achieve information rates of 42 Gb/s below the HD-FEC threshold at distances beyond 40 km. We find that our results outperform conventional IM/DD solutions based on two- and four-level pulse amplitude modulation with feedforward equalization at the receiver. Our study is the first step toward end-to-end deep learning based optimization of optical fiber communication systems.
29 June 2015 in Journal of Lightwave Technology
The FEC limit paradigm is the prevalent practice for designing optical communication systems to attain a certain bit error rate (BER) without forward error correction (FEC). This practice assumes that there is an FEC code that will reduce the BER after decoding to the desired level. In this paper, we challenge this practice and show that the concept of a channel-independent FEC limit is invalid for soft-decision bit-wise decoding. It is shown that for low code rates and high-order modulation formats, the use of the soft-decision FEC limit paradigm can underestimate the spectral efficiencies by up to 20%. A better predictor for the BER after decoding is the generalized mutual information, which is shown to give consistent post-FEC BER predictions across different channel conditions and modulation formats. Extensive optical full-field simulations and experiments are carried out in both the linear and nonlinear transmission regimes to confirm the theoretical analysis.
20 July 2020 in IEEE Photonics Technology Letters
The maximum data throughput in a single mode optical fibre is a function of both the signal bandwidth and the wavelength-dependent signal-to-noise ratio (SNR). In this paper, we investigate the use of hybrid discrete Raman & rare-earth doped fibre amplifiers to enable wide-band signal gain, without spectral gaps between amplification bands. We describe the widest continuous coherent transmission bandwidth experimentally demonstrated to date of 16.83 THz, achieved by simultaneously using the S-, C-and L-bands. The variation of fibre parameters over this bandwidth, together with the hybrid amplification method result in a significant SNR wavelength-dependence. To cope with this, the signal was optimised for each SNR, wavelength and transmission band. By using a system-tailored set of geometrically shaped constellations, we demonstrate the transmission of 660×25 GBd channels over 40 km, resulting in a record single mode fibre net throughput of 178.08 Tbit/s.
28 June 2019 in Optics Express
We propose an autoencoding sequence-based transceiver for communication over dispersive channels with intensity modulation and direct detection (IM/DD), designed as a bidirectional deep recurrent neural network (BRNN). The receiver uses a sliding window technique to allow for efficient data stream estimation. We find that this sliding window BRNN (SBRNN), based on end-to-end deep learning of the communication system, achieves a significant bit-error-rate reduction at all examined distances in comparison to previous block-based autoencoders implemented as feed-forward neural networks (FFNNs), leading to an increase of the transmission distance. We also compare the end-to-end SBRNN with a state-of-the-art IM/DD solution based on two level pulse amplitude modulation with an FFNN receiver, simultaneously processing multiple received symbols and approximating nonlinear Volterra equalization. Our results show that the SBRNN outperforms such systems at both 42 and 84 Gb/s, while training fewer parameters. Our novel SBRNN design aims at tailoring the end-to-end deep learning-based systems for communication over nonlinear channels with memory, such as the optical IM/DD fiber channel.
14 February 2013 in Journal of Lightwave Technology
The relative merits of coherent-enabled optical access network architectures are explored, with a focus on achievable capacity, reach and split ratio. We review the progress in implementing the particular case of the ultra dense wavelength division multiplexed (UDWDM) passive optical network (PON), and discuss some challenges and solutions encountered. The applicability of digital signal processing (DSP) to coherent receivers in PONs is shown through the design and implementation of parallelized, low-complexity application-specific digital filters. In this work, we focus on mitigating the impact of local oscillator laser (LO) relative intensity noise (RIN) on receiver sensitivity, and propose an algorithm which compensates for this impairment. This phenomenon is investigated theoretically and then experimentally by evaluating the sensitivity of a coherent receiver incorporating different tunable light sources; a low-RIN external cavity laser (ECL) and a monolithically integrated digital supermode distributed Bragg reflector (DS-DBR) laser. It is shown that the RIN of the signal laser does not significantly contribute to the degradation of the receiver sensitivity. Finally, a 10 Gbit/s coherent PON is demonstrated using a DS-DBR laser as the LO laser. It is found that a receiver sensitivity of -38.8 dBm is achievable assuming the use of hard-decision forward error correction.
28 November 2018 in Journal of Lightwave Technology