Simulation And Calculation Of Nonlinear Signal Propagation In Multimode Fiber

The modern interconnected world is built around a single-mode fiber network.Today,single-mode fiber bandwidth is close to the limit of Shannon's theorem,that today's single-mode fiber systems will not be able to meet the growing demand,and an effective solution to this bandwidth crisis is to adopt multimode fiber.First,for multimode fiber,the numerical simulation of the mode is very challenging because of the complex nonlinearity between modes.Based on this,this thesis makes a simulation experiment on the propagation effect and numerical accuracy of multi-mode fiber nonlinear signals.Secondly,the traditional fiber optic system is simulated on the CPU,and the efficiency is very low.In this paper,the GPU-based calculation is designed to accelerate the simulation process of the traditional multimode fiber system.The main contents of the paper are as follows.(1)Applying the fourth-order Runge-Kutta method(RK4IP)in mutual mapping to the numerical analysis of the nonlinear Schrodinger equation of multimode fiber,and based on this algorithm,the transmission of optical pulse in the fiber is simulated.(2)In terms of computational performance,in order to better express the ef'fect of RK4IP,combined with the general SSFM-RK4 numerical algorithm,the effect diagram of the two steps in step size and precision is simulated.The advantage of RK4IP in accuracy calculation is determined.In the analysis of error accuracy,three error estimation criteria-max,sum,ave criteria for solving nonlinear transmission equations of multimode fibers are proposed.The results show that the graphs under the three criteria all meet the accuracy requirements of the algorithm.(3)Adopting CUDA programming model and accelerating calculation based on GPU.The simulation experiments on the memory and time consumption of SSFM-RK4 algorithm and RK4IP algorithm are respectively carried out,and the acceleration effect of GPU on single and double precision data is explored.As the number of pulses increases,the average pulse time of CPU and GPU is time-consuming.stability.The results show that the GPU-based calculation can be tens of times more efficient than the CPU.In addition,the simulation of each module in the RK4IP algorithm is carried out,and the acceleration effect of the GPU in each module is obtained.(4)Using the Pytorch programming language,the Pytorch language is used to support the tensor calculation,and the GPU operation is very simple,and breaks the disadvantage of not supporting the complex tensor calculation.