An efficient initialization method for FDTD computation of plane wave scattering

Target identification processing requires accurate diffraction modelling. When the data have to be processed in the time domain (TD), for example with short pulse excitation, a high dynamic range is necessary to be able to take advantage of both the early and late times parts of the signal. Also there is a need of being able to compute very low observable targets scattering. It is then necessary, in any TD computation method, to get rid of round-off errors which may become significant in the case of sharp, rapidly varying incident pulse excitation. Usually the finite difference time domain method (FDTD) computations are done on the total field by using Yee´s algorithm. As the scattered field has to be calculated in the far zone and also because one wishes to avoid spurious reflections of the incident field on the absorbing boundaries, it must exist near the boundary of the volume calculation a sub-volume free of incident field. For this purpose, Mur (l981) and Taflove (1990) have proposed an initialization of the incident field which takes advantage of the fact that the incident field is known analytically everywhere. What is proposed in this paper is a variant of the previous method. The incident field is updated by the Yee´s algorithm in exactly the same way as the total field, which eliminates round-off errors and prevents any leakage of the incident pulse out of the inner volume. Furthermore, the incident field is calculated analytically at only one node of the grid which alleviates the computational effort.