Abstract
High-symbol-rate coherentoptical transceivers suffer more from the critical responses of transceiver components at high frequency, especially when applying a higher order modulation format. We recently proposed a neural network (NN)-based digital pre-distortion (DPD) technique trained to mitigate the transceiver response of a 128 GBaud optical coherent transmission system. In this paper, we further detail this work and assess the NN-based DPD by training it using either a direct learning architecture (DLA) or an indirect learning architecture (ILA), and compare performance against a Volterra series-based ILA DPD and a linear DPD. Furthermore, we deliberately increase the transmitter nonlinearity and compare the performance of the three DPDs schemes. The proposed NN-based DPD trained using DLA performs the best among the three contenders. In comparison to a linear DPD, it provides more than 1 dB signal-to-noise ratio (SNR) gains at the output of a conventional coherent receiver DSP for uniform 64-quadrature amplitude modulation (QAM) and PCS-256-QAM signals. Finally, the NN-based DPD enables achieving a record 1.61 Tb/s net rate transmission on a single channel after 80 km of standard single mode fiber (SSMF).
Highlights
T HE exponential increase in the internet traffic due to the emergence of bandwidth-hungry services such as cloud-based applications and video on demand is pushing the existing optical transport network to its limit
We demonstrate that the considered neural network (NN)-based digital pre-distortion” (DPD) leads to a record 1.61 Tb/s data rate over a 80 km fiber link, detail the proposed NN-based DPD and further investigate its performance by training it using the two well-known learning architectures, namely the indirect learning architecture (ILA) [7], [18] and the direct learning architecture (DLA) [8], [9]
Our results show that NN-based DPD trained using DLA performs the best among the considered candidates and obtains gains of 1.6 dB and 1.2 dB in received signal-to-noise ratio (SNR) with respect to the linear DPD for uniform 64-quadrature amplitude modulation (QAM) and probabilistic constellation shaped (PCS)-256QAM, respectively
Summary
T HE exponential increase in the internet traffic due to the emergence of bandwidth-hungry services such as cloud-based applications and video on demand is pushing the existing optical transport network to its limit. It is a common practice to limit the amplitude of the signals applied to transmitter components exhibiting a nonlinear response (e.g. driver amplifier (DA), electro-optic modulator, etc.) when using a linear DPD, which in turn limits the SNR because of the small signal power. To increase the information rate, transmitters will require DPDs that can compensate for both the linear and nonlinear responses. The most common nonlinear DPDs are based on Volterra series which have been investigated for both radio frequency (RF) amplifiers [7]–[11] and coherent optical transmitters [12]–[17]. We proposed an NN-based DPD designed using simple FFNNs and CNNs for a high-baud rate (128 GBaud) coherent optical transmitter [31]. The code with the smallest overhead capable of decoding the bits error-free is chosen
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have