D-type Optical Fiber Sensor Research Based On FDTD Method

Light is transmitted in fiber by total reflection. In the event of a total reflection,light is not absolute reflection on the interface between the optical dense mediumand the optical sparse medium. Light will permeate and spread some distance inoptical sparse medium before returning to the optical denser medium, which is calledevanescent wave. Therefore, when light propagates in the fiber,at the interfacebetween the fiber core and cladding are formed evanescent field. If the refractiveindex of fiber cladding is changed, the light propagating here will not fully return todense medium. There is a part of the light leaking from the optical fiber. Bymeasuring the emitted light energy, the situation of light leakage can be knowd.Thereby the cladding refractive index changes can be derived. D-fiber sensor utilizesthis principle, generate light leakage by the partial removal of the cladding, and thusthe purpose of sensing arrival. In order to understand the performance of D-fiber,analysis the effect of source, ablation depth, resection length and the refractive indexof the external environment by numerical simulation. Before launch evanescent fieldsensing experiments, numerical simulation results provide a reference for the choiceof experimental parameters, also to predict the experimental results.The study use the FDTD method for D-fiber simulation, analysis thepropagating light and the light leakage. First, use the FDTD professional computingsoftware EASTFDTD for D-fiber simulation, the result shows the phenomenon ofevanescent field and light leakage. Then, the result of concrete D-fiber struct is given,includes the the cladding refractive index of the D-surface, the depth of the removalcladding and so on. Due to the need for simulation in three-dimensional space, thecomputation time is always long. In order to increase computing speed, the parallelFDTD method is studied.First, in order to verify the correctness of the parallel result,a serial result is given. Next, OpenMp, MPI, windows API and Cuda altogether fourdifferent methods are used for parallel calculation, the result is same with the serial result. At last, a comparison of the parallel and serial speed is given.