Research On Transmitter/receiver IQ Imperfection Compensations And Optical Network Monitoring

The rapid growth of high-traffic demand services such as the Internet of Things,smart cities and virtual reality drives the evolution of optical networks towards large capacity.As high-speed and large-capacity transmission enabling technologies,high symbol rate,high-order modulation format and Nyquist shapping have significantly improved the transmission capacity of optical network.However,the optical receiver/transmitter(Rx/Tx)in-phase/quadrature(IQ)imperfection and frequency response insufficiency of transmitter become one of the critical elements restricting the performance of large capacity optical networks.Although the existing schemes can compensate or calibrate the Rx/Tx IQ imperfection,it is difficult to be applied practically due to high complexity or high cost.With the transmission capacity of backbone network,metro optical network and data center Internet growing rapidly,the scale is also expanding and the optical performance monitoring is becoming more important.To verify whether each wavelength channel of high-capacity optical network adopts the optimal parameters set by the control plane,there is an urgent need for a technology at intermediate node to monitor the current wavelength channel parameters and feed them to the control plane for comprehensive analysis.For the intermediate nodes are widely distributed and numerous in the optical network,the monitoring at the intermediate nodes requires low cost.However,the cost of the most existing schemes is too high to deploy.Therefore,this thesis conducts indepth research on the low tolerance of Rx/Tx IQ imperfection in highcapacity optical transmission with high-symbol rate and high-order modulation format,and the high cost of optical performance monitoring for physical layer parameters at the intermediate node in the optical network.The main research works and the innovations are as follows.1.Multiple hardware impairments of optical transmitters will lead to performance deterioration of coherent optical transmission systems with high symbol rate and high-order modulation formats,such as frequency response insufficiency of the transmitter and Tx IQ imperfection.Aiming at these issues,this paper proposes a joint pre-compensation scheme for frequency response insufficiency of transmitter and Tx IQ imperfection.The scheme uses a deep neural network to model the damaged transmitter,and then trains the inverse system of the damaged transmitter based on the symmetric structure of the autoencoder.The autoencoder optimizes the joint data of the entire transmitter to achieve joint pre-compensation of multiple hardware impairments such as Tx IQ imbalance and insufficient frequency response.In addition,the autoencoder after the convergence of offline training can directly pre-compensate the online transmitter without increasing the implementation complexity.The simulation results of 128 GBaud 16 constellation quadrature amplitude modulation(16 Quadrature Amplitude Modulation,16 QAM)system under 10×80 km transmission show that,under 60%of the Nyquist signal bandwidth,the proposed scheme can compensate Tx IQ phase,amplitude imbalance and skew of 15 degree,3 dB and 0.2 symbol period with 1.5 dB Q factor penalty.2.In view of the issue that the current Rx or Rx/Tx IQ imbalance compensation scheme has high complexity or is not suitable for high-order modulation formats,a blind frequency domain 4×2 multiple-input multiple-output(BFD 4×2 MIMO)algorithm based on radius-directed equalization is proposed.By equalizing the input sequence block by block in frequency domain,the complexity can be decreased under long-distance transmission compared with the time-domain(TD)equalization algorithm.In addition,the discrete frequency response of the BFD 4×2 MIMO equalization updated adaptivly based on the radius-directed equalization not only has high spectral efficiency without training sequence,but also is suitable for high-order modulation formats.By studying the connection between the converged discrete frequency response of the BFD 4×2 MIMO equalization and the Rx/Tx IQ imbalance,the Rx/Tx IQ imperfection can be estimated.The 40 GBaud polarization multiplexing(PM)-16QAM Nyquist offline experiment results show that for 3×80 km transmission,when the Rx/Tx IQ skew,amplitude and phase imbalance within 8 and 6 ps,5.5 and 3.5dB,40 and 12 degree,the maximum Q penalty is 1.9 dB.The simulation results under 128 GBaud PM-16QAM Nyquist system show that for 10×80 km transmission,at the cost of 0.1 and 1 dB system Q factor,the compensation ranges for Rx/Tx IQ skew,amplitude and phase imbalance are[-70%,70%]and[-25%,25%]symbol period,[-7.5,7.5]and[-3,3]dB,[-40,40]and[-20,20]degree.The corresponding monitoring accuracy is 0.15 and 0.31ps,0.15 and 0.5 dB,0.5 and 0.5 degree.In addition,under 35×80 km transmission,compared with the blind TD 4×2 MIMO equalizer,the proposed algorithm reduces the computational complexity by 63.3%under the same performance.3.Monitoring the performance of intermediate nodes in an optical network enables the control plane to sense the status of the current wavelength channels.Nevertheless,most of the existing monitoring schemes at intermediate node have the problem of high cost and difficulty in large-scale deployment.To cope with this issue,a low-cost multiparameter monitoring scheme is proposed at intermediate nodes in optical networks,such as optical signal to noise ratio(OSNR)and so on.By conveniently scaning all wavelength channels of a fiber port based on coherent detection with wavelength-tunable local oscillator,using lowbandwidth coherent detection to filter out high-frequency components in the monitoring signal without chromatic dispersion(CD)compensation,and jointing with the random forest with lower computational complexity than the fully connected neural network,the proposed multi-parameter monitoring scheme at intermediate nodes is not only small bulk but also low cost.8/16 GBaud PM-4/16/32/64QAM simulation and PM4/16/32QAM coherent optical transmission offline experimental results show that,125 MHz detection bandwidth and 2560,10240 and 15360 ps/nm CD,the symbol rate-modulation format identification accuracy is 100%,and the mean absolute errors of OSNR estimation are less than 0.37,0.48 and 0.59 dB under 16 GBaud PM-4/16/32QAM.