102.3 Tbit/s C+L-band transmission over 1512 km SSMF enabled by an uncertainty-aware Bayesian CNN-BiLSTM hybrid equalizer

We experimentally demonstrate a high-capacity dense wavelength division multiplexing (DWDM) transmission covering a 12.2 THz bandwidth over a 6×252 km (1512 km) standard single-mode fiber (SSMF) link. To mitigate the severe transmission impairments inherent in wideband systems, such as fiber nonlinearity and inter-symbol interference, we propose an equalizer that cooperatively combines convolutional neural networks (CNN) for local feature extraction, bidirectional long short-term memory networks (BiLSTM) for temporal modeling, and a Bayesian head for uncertainty-aware residual prediction. Experimental results show that with the help of the proposed Bayesian CNN-BiLSTM hybrid equalizer, the generalized mutual information (GMI)-estimated throughput is significantly improved from 97.1 Tbit/s to 102.3 Tbit/s, while the training epochs are reduced by 80% comparing to conventional neural network methods. Consequently, the system achieves a spectral efficiency (SE) of 8.4 bit/s/Hz and an average single-carrier capacity of 838.5 Gbit/s over 1512 km SSMF, validating the effectiveness of the proposed equalizer for wideband long-haul transmission.