Study On Signal Processing Technology Based On Optical Reservoir Computing

With the rapid development of fiber optic communication technology in the 21^stcentury,verification sets have become an important cornerstone for supporting various emerging network applications by providing fiber optic transmission systems with faster access speeds and higher channel capacity.However,there is a more complex problem of signal crosstalk that requires powerful data processing techniques to suppress various types of noise interference,including nonlinear noise.In recent years,as an intelligent processing technology,neural networks have the advantages of low processing-complexity and high compensation-efficiency,and gradually become an important data compensation tool in the field of optical fiber communication.This thesis focuses on the research of signal processing technology based on optical reservoir computing（RC）.Firstly,a specific reservoir structure is constructed through nonlinear optical device models,and then signal compensation and channel equalization are conducted for the noise problems of single mode and few mode transmission systems,respectively.The research work mainly includes the following two aspects:Firstly,a RC scheme based on nonlinear optical loop mirrors（NOLMs）is proposed.By mathematical modeling the nonlinear power transfer characteristics of NOLM,the optical activation unit in the RC system is constructed,and then the dispersion compensation in single mode communication systems and the mode-coupling mitigation in few-mode links are compensated.In dispersion compensation,the compensation performance of NOLM-RC is discussed from the perspectives of the transmission distance and the working wavelength.Through joint optimization of NOLM unit and reserve cell parameters,the 140km dispersion compensation is achieved,as well as the simultaneous compensation of C+L band data.In the compensation of the mode-coupling induced crosstalk,a multiple input multiple output（MIMO）processing method based on NOLM-RC is constructed,and the Q² value of compensated signals reaches 11.95d B in short-reach communication,verifying the application of NOLM-RC in the few-mode system.Secondly,a SOA-RC scheme based on semiconductor optical amplifiers（SOAs）is proposed.By establishing a nonlinear power mapping model under steady-state SOA,a SOA-RC signal processing framework is formed.On this basis,the equalization of wavelength division multiplexing（WDM）channels and module division multiplexing（MDM）channels are carried out.In WDM channel equalization,the SOA-RC compensation method is used to perform multi-channel data equalization for on-off keying（OOK）signals and differential quadrature-phase-shift keying（DQPSK）signals.After compensation,the signal quality reachs the error-rate threshold requirements,and the improvement is observed within the injection power range of-10d Bm or more.In the MDM channel equalization,the data compensation is carried out for a 12 km of 3-mode(LP₀₁,LP_11a,and LP_11b)parallel communication.The maximum Q² value of the compensated signal can reach up to 14.88d B,and meets the error-rate threshold requirements in the range of-6d Bm to-1d Bm,confirming the multi-channel equalization of the proposed SOA-RC scheme on the MDM system.