Skip to Main Content
A new approach to wave oriented radio propagation modeling based on extended finite difference time domain (FDTD) method has been developed. In order to model radio wave propagation using standard FDTD, the entire propagation path would need to be included in the FDTD grid. This would require prohibitive amount of computer memory and time. The new approach takes advantage of the fact that when a pulsed radio wave propagates over a long distance the significant pulse energy exists only over a small part of the propagation path at any instant of time. This allows the use of a relatively small FDTD computational mesh that exists only over a portion of the propagation path and move along with the pulse. At the leading edge of the FDTD mesh inside the moving window the appropriate terrain, foliage, and atmospheric parameters are added to the mesh. At the trailing edge the terrain and foliage that have been left behind by the pulse are removed. Since this method solves Maxwell's equation directly, all physics relevant to radio wave propagation is included in this model In addition, since it is based on the FDTD method, this model can take advantage of the extensive research that has been done on the FDTD algorithm, such as the higher-order finite differencing and multiresolution time domain (MRTD). The moving window FDTD (MWFDTD) method has previously been applied to propagation over different types of irregular terrain. This paper extends this approach to forest covered terrain by treating the foliage as a lossy dielectric layer. In addition, using MWFDTD, atmospheric refractive effects can also be included simultaneously with irregular terrain effects and foliage. We apply our method to radio wave propagation in atmospheric ducts. Comparisons with path loss measurements show good accuracy and illustrate the advantages of a full wave method. While this method is slower than other propagation models, it is potentially much more accurate. Therefore, MWFDTD can be used to validate other faster propagation models, and can provide predictions in cases where a high degree of accuracy is desired.