We propose a neural network equalization delta-sigma modulation(DSM)technique.After performing DSM on the multiorder quadrature amplitude modulation(QAM)orthogonal frequency division multiplexing(OFDM)signal at the tr...We propose a neural network equalization delta-sigma modulation(DSM)technique.After performing DSM on the multiorder quadrature amplitude modulation(QAM)orthogonal frequency division multiplexing(OFDM)signal at the transmitting end,neural network equalizer technology is used in the digital signal processing at receiving end.Applying this technology to a 4.6 km W-band millimeter wave system,it is possible to achieve a 1 Gbaud 8192-QAM OFDM signal transmission.The data rate reached 23.4 Gbit/s with the bit error rate at 3.8×10^(-2),lower than soft-decision forward-error correction threshold(4×10^(-2)).展开更多
We demonstrate a 200 m outdoor 2×2 multiple-input multiple-output(MIMO)terahertz(THz)communication system operating at 300 GHz with 200 Gb/s polarization-division multiplexed quadrature phase-shift keying(PDM-QPS...We demonstrate a 200 m outdoor 2×2 multiple-input multiple-output(MIMO)terahertz(THz)communication system operating at 300 GHz with 200 Gb/s polarization-division multiplexed quadrature phase-shift keying(PDM-QPSK)transmission.We propose an iteratively pruned two-dimensional convolutional neural network(2D CNN)equalizer that adaptively captures polarization crosstalk and temporal nonlinearities through 2D convolution kernels.The system achieves a bit error rate(BER)below the hard-decision forward error correction(HD-FEC)threshold at a lower power of 6 d Bm,while reducing the computational complexity by 30.2%compared to the iteratively pruned one-dimensional(1D)CNN approach.This enables high-capacity and energy-efficient operation in long-distance THz links.展开更多
We experimentally demonstrate an 80-channel wavelength division multiplexing(WDM)transmission system over a 400 km fiber link.Raman amplification results in a non-flat WDM signal spectrum.Therefore,bit allocation opti...We experimentally demonstrate an 80-channel wavelength division multiplexing(WDM)transmission system over a 400 km fiber link.Raman amplification results in a non-flat WDM signal spectrum.Therefore,bit allocation optimization is used to enable different channels to carry different order quadrature amplitude modulation signals according to their optical signal-noise-ratios.A neural network equalizer based on a convolutional neural network(CNN),long shortterm memory(LSTM)network,and fully connected(FC)layer structure is adopted in Rx digital signal processing,in which CNN is used for characteristic extraction,LSTM is used for equalization and demodulation,and FC layers are used for output.After transmission,the bit error rate of all channels is below the 25%soft-decision forward error correction threshold,and the line rate reaches 53.76 Tbit/s.展开更多
基金supported by the National Natural Science Foundation of China(Nos.62225503,61835005,and 62205151)。
文摘We propose a neural network equalization delta-sigma modulation(DSM)technique.After performing DSM on the multiorder quadrature amplitude modulation(QAM)orthogonal frequency division multiplexing(OFDM)signal at the transmitting end,neural network equalizer technology is used in the digital signal processing at receiving end.Applying this technology to a 4.6 km W-band millimeter wave system,it is possible to achieve a 1 Gbaud 8192-QAM OFDM signal transmission.The data rate reached 23.4 Gbit/s with the bit error rate at 3.8×10^(-2),lower than soft-decision forward-error correction threshold(4×10^(-2)).
基金supported by the National Key R&D Program of China(No.2023YFB2905600)the National Natural Science Foundation of China(Nos.62127802,62331004,62305067,U24B20142,U24B20168,and 62427815)the Key Project of Jiangsu Province of China(No.BE2023001-4)。
文摘We demonstrate a 200 m outdoor 2×2 multiple-input multiple-output(MIMO)terahertz(THz)communication system operating at 300 GHz with 200 Gb/s polarization-division multiplexed quadrature phase-shift keying(PDM-QPSK)transmission.We propose an iteratively pruned two-dimensional convolutional neural network(2D CNN)equalizer that adaptively captures polarization crosstalk and temporal nonlinearities through 2D convolution kernels.The system achieves a bit error rate(BER)below the hard-decision forward error correction(HD-FEC)threshold at a lower power of 6 d Bm,while reducing the computational complexity by 30.2%compared to the iteratively pruned one-dimensional(1D)CNN approach.This enables high-capacity and energy-efficient operation in long-distance THz links.
文摘We experimentally demonstrate an 80-channel wavelength division multiplexing(WDM)transmission system over a 400 km fiber link.Raman amplification results in a non-flat WDM signal spectrum.Therefore,bit allocation optimization is used to enable different channels to carry different order quadrature amplitude modulation signals according to their optical signal-noise-ratios.A neural network equalizer based on a convolutional neural network(CNN),long shortterm memory(LSTM)network,and fully connected(FC)layer structure is adopted in Rx digital signal processing,in which CNN is used for characteristic extraction,LSTM is used for equalization and demodulation,and FC layers are used for output.After transmission,the bit error rate of all channels is below the 25%soft-decision forward error correction threshold,and the line rate reaches 53.76 Tbit/s.
