A parallel computing architecture based on NVIDIA's compute unified device architecture (CUDA) for the modeling of electromagnetic wave propagation by means of the high-frequency approximation method, shooting and bouncing rays (SBRs), is introduced and evaluated. The algorithm provides a reliable treatment of problems involving bulky and nonperfectly conducting materials. The general case of wave propagation through arbitrary dielectric materials also considering evanescent waves is presented in a concise way and a simplified treatment of the practically important case of thin dielectric layers over perfectly electrically conducting (PEC) bodies is considered. The resulting hybrid SBR approach is capable of dealing with PEC and the various dielectric object cases in a unified manner. A wide variety of scattering problems is considered, and scattered field predictions are compared to exact method of moments (MoM) and finite-element method (FEM) results. In particular, the benefits of the SBR within bulky dielectrics are highlighted.Index Terms-Compute unified device architecture (CUDA), dielectrics, ray tracing, scattering, shooting and bouncing rays (SBRs). 0018-926X

No abstract

No abstract

No abstract

Abstract-An iterative solution scheme based on the magnetic field integral equation (MFIE) to compute electromagnetic scattering for arbitrary, perfect electrically conducting (PEC) objects is topic of this contribution. The method uses simple and efficient approaches for the computation of surface current interactions which are typically found in the well-known iterative physical optics (IPO) technique. However, the proposed method is not asymptotic, since no physical optics (PO) concepts are utilized. Furthermore, a least squares correction method is introduced, which is applied not on the complete current vector, but on individual groups of currents. This helps to quickly reduce the residual error and to improve convergence. The result is a simple method which is capable to improve the simulation results obtained by pure asymptotic methods such as PO or shooting and bouncing rays (SBR). The method can be regarded as a simplified iterative method of moments (MoM) technique. Numerical examples show that the proposed approach is advantageous e.g., in problem cases where the neglect of diffraction effects or currents in shadow regions would cause large errors. It also provides an improved prediction of the peak scattering contributions.