Research On Dispersion Compensation Algorithm In Digital Coherent Optical Communication System

With the application of new technologies such as autonomous driving,virtual reality,augmented reality,mixed reality and naked-eye 3D in the 5G era,higher demands have been placed on existing optical communication networks.Coherent optical communication has become the mainstream of the current optical communication field due to its advantages of high sensitivity,long relay distance,good selectivity,large communication capacity and multiple modulation methods.However,dispersion is a key factor restricting the development of coherent optical communication towards long distance,large capacity and high rate.Therefore,it is significant to study the dispersion compensation technology of coherent optical communication.This paper focuses on the research of electric domain dispersion compensation technology for coherent optical communication.The main contents of the paper include:1.The principle of fiber dispersion and its effect on signal transmission are analyzed and the classic dispersion compensation method is introduced.A new type of optimization function is designed to optimize the weight of the classic time-domain equalizer.The filter is simulated and verified in the single-polarization coherent optical QPSK communication system.The simulation results show that the frequency spectrum of the filter has the characteristics of wider passband,narrower transition band and smoother spectrum curve,which reduces the spectrum leakage caused by truncation effect and has better equalization effect than the time domain equalizer.2.An adaptively trained dispersion compensation filter based on gradient descent is designed.The filter can use the principle of pulse shaping to limit the effective bandwidth of the signal and be designed in a narrow frequency band to reduce the number of taps.The filter is simulated and verified in the coherent QPSK/M-QAM system.The simulation results show that the filter has better equalization effect under the same number of taps than the time-domain equalizer,and the increase of the tap number will not cause the deterioration of the compensation effect.The filter can reduce the number of taps by 37.5% through the narrow band design,which effectively reduces the hardware implementation complexity and filter delay.3.According to the relationship between the dispersion transfer function and the frequency domain function of the finite-length impulse response(FIR)filter,an overdetermined linear equation set is constructed.By iteratively solving the least squares solution of the equation set,the optimal dispersion compensation filter in the sense of least squares is obtained.The filter is simulated and verified in the coherent QPSK/M-QAM system.The simulation results show that the filter has better equalization effect under the same number of taps than the time domain equalizer and can also be designed in a narrow frequency band,reducing the tap number by 37.5% without the deterioration of the equalization effect.In particular,the filter requires fewer frequency sample points to obtain a faster iteration update rate of the tap coefficients.