Mitigating Nonlinear Interference by Limiting Energy Variations in Sphere Shaping

Abstract

Band-trellis enumerative sphere shaping is proposed to decrease the energy variations in channel input sequences. Against sphere shaping, 0.74 dB SNR gain and up to 9% increase in data rates are demonstrated for single-span systems.

Towards FPGA Implementation of Neural Network-Based Nonlinearity Mitigation Equalizers in Coherent Optical Transmission Systems

“…The deep CNN performed worse, improving the optimal power from −1 dBm to 1dBm and the Q-factor by 0.8 dB. Moving to the experimental data 3 , we observe that biLSTM and deep CNN provided 1.7 dB and 1 dB in Q-factor improvement, respectively, and the optimal power shifts from 0 dBm to 1 dBm in both cases. For the DBP case, the Q-factor improvement in the experiment was 1.5 dB and in the simulation was 1.3 dB.…”

“…will push optical networks to meet even more stringent traffic requirements. However, the Kerr effect limits the information rates that optical fiber communication systems can achieve [1]- [3] . Machine learningbased solutions have lately been touted as a possible approach to mitigate the impact of this fiber transmission effect [4]- [8] .…”

“…For applying the trained model to other launch powers, we used transfer learning[13] , thus requiring less than 5 epochs to adjust the NN weights to the other launch powers 3. The Q-factor produced by the Python model and the FPGA implementation were virtually identical because we did not consider quantization at this moment.…”

Band-ESS: Streaming Enumerative Coding with Applications to Probabilistic Shaping

“…6.1]. In this paper, as we hinted in [16], we propose to diverge from this block-based high-storage-complexity shaping paradigm. We introduce a streaming B-ESS implementation, similar to convolutional encoding.…”

“…Such distributions provide higher achievable rates than MB distributions for short-haul optical links [14], [15] and thus, K-ESS has been shown to provide SNR gains and reach increases over ESS. We have also recently proposed band-trellis ESS (B-ESS) [16]. B-ESS modifies the ESS algorithm such that shaped channel inputs with limited energy variations are generated.…”