Research On Efficient Energy Equalization Schemes In Short Range Optical Communication Systems Based On IMDD

With the emergence of various new growing businesses,network users have increasingly higher requirements for information transmission rate and quality.This will pose great challenges to the further development of future networks.Therefore,fiber-optic networks that can carry large-capacity information urgently need to develop towards high bandwidth,high speed,and high quality.As the transmission system closest to the user,short-reach optical communication systems also need to support high-bandwidth and high-speed transmission.Short-reach optical communication systems are mainly used in optical access networks,data center optical interconnection,and other scenarios where system upgrade costs are relatively sensitive.Therefore,how to efficiently and cost-effectively realize short-reach optical communication transmission systems has become a current research hotspot.Among various candidate schemes,intensity modulation direct detection(IMDD)transmission technology based on large bandwidth and high performance has received widespread attention in the industry due to its low cost and simple deployment advantages in short-reach optical communication systems.However,in this system,the interaction between linear/non-linear noise of photonic devices,chirping of directly modulated lasers(DML),fiber dispersion,and various types of damage seriously affects the transmission performance of the system.Therefore,this paper focuses on the equalization scheme in the IMDD short-reach optical communication system,and studies the feasible equalization scheme with good compensation effect and low complexity.The specific work is as follows.1)Firstly,the research significance and background of the equalization scheme based on IMDD in short-reach optical communication systems were outlined.By summarizing and analyzing the research status at home and abroad,it was pointed out that there are problems such as limited performance improvement or excessive computational complexity in existing equalization schemes.Then,the damage in the short-reach optical communication system based on IMDD was analyzed,including the nonlinear damage of DML,the bandwidth-limited effect of low-modulation-bandwidth photonic devices,fiber dispersion,and their interactions.The experimental and simulation platform was established for the IMDD-based short-reach optical communication system transmission of 80Gb/s quaternary pulse amplitude modulation(PAM-4)signals and 80 Gb/s non-return-to-zero(NRZ)signals,and the performance of traditional equalization algorithms was tested on this platform to provide comparison data for the designed equalization scheme in this paper.2)In order to overcome the problems of high computational complexity and limited performance improvement of traditional nonlinear equalizers,this paper designed the feature simplification Volterra equalization scheme based on linear discriminant analysis(LDA)and multiple-symbol output,and analyzed the principles of the corresponding schemes.The feasibility of the two designed schemes was verified on the simulation and experimental platforms of the IMDD-based short-reach optical communication system.Among them,the equalization scheme based on LDA achieves a reduction in the complexity by choosing the features that contributes most to the performance improvement.The simulation and experimental results show that the equalization effect similar to that of the traditional Volterra equalizer can be achieved with only 30% of the number of features.The equalization scheme based on multiple-symbol output optimizes the input-output structure of the Volterra,which reduces the repeated calculation of the input high-order features and reduces the computational complexity of the equalization scheme.Under the condition of obtaining the same order of magnitude of bit error rate,the designed scheme requires only about 33.4% of the multiplication and addition operations of the traditional Volterra equalizer.3)To address the problem that there is a complex structure for the neural network equalization scheme that is difficult to apply in practical systems,this paper proposes to use the idea of transfer learning to reduce the training complexity of the neural network equalization scheme.Based on this,simplified equalization schemes based on deep neural networks,recurrent neural networks,and gated recurrent unit neural networks were designed using the Adam adaptive optimization algorithm to adjust the learning rate of the parameters in the neural network reasonably,and dropout layers were introduced to reduce the risk of overfitting,and three fine-tune methods are proposed.The three designed schemes were validated through an IMDD-based short-reach optical communication system experimental platform;The results show that the three simplified equalization schemes based on neural network transfer only require 30%,35%,and 50% training sequence length and 23.5%,28.2%,and 26.7% iteration times to achieve stable performance,and can achieve almost the same bit error rate performance as traditional neural network-based equalization schemes.The design of these schemes will provide certain technical support for the application of neural network equalizers in short-reach optical communication systems in the future.