GPU accelerated fast MoM-PO for modeling complex onboard antenna mounted on large-scale platforms

Abstract

Summary form only given. Efficient modeling of onboard antennas with large-scale platforms is a hot topic in electromagnetic compatibility (EMC) study and antenna system design. The hybrid method of moments (MoM) - physical optics (PO) is a possible attempt to the onboard antenna problem, where the antenna and its small surrounding area are modeled with MoM, and the rest part of the platform is represented with PO. Then the PO contribution is coupled into the MoM impedance matrix, and no unknowns are introduced by PO. So the number of unknowns can be saved. However, this conventional MoM-PO method is exceedingly time-consuming in depicting large PO-region. To overcome the limitation of the conventional MoMPO, an efficient iterative MoM-PO (EI-MoM-PO) technique (Z. L. Liu, and C. F. Wang, IEEE Trans. on AP, vol. 60, no. 7, 3520-3525) has been proposed through an iterative process between the antenna and the platform. This EI-MoM-PO method can significantly reduce the computational time of the PO-region and efficiently achieve results with good accuracy. To further improve the efficiency and reduce the peak memory usage in the application of EI-MoM-PO, the fast MoM-PO is presented by employing adaptive integral method (AIM) to speed up the MoM part, and has been proved very suitable for the problem of complex antenna arrays mounted on electrically large platforms. Although the PO part and the MoM part have been accelerated with the EI-MoMPO and the fast MoM-PO, one more bottleneck of the hybrid MoM-PO method is the ray tracing on large-scale platforms. On the other hand, graphics processing units (GPUs) is capable for the simple but intensive computational work and have been applied to speed up some computational electromagnetic algorithms. To achieve desirable efficiency of the ray tracing in the hybrid method, Octree structures will be set up first by subdividing the MoM-region recursively into non-overlapping sub-regions. Then an advanced ray tracing technique on GPUs is utilized to track rays radiating from each sub-region to determine the lit POregion of the large-scale platform. With this GPU ray tracing technique, the efficiency of the fast MoM-PO can be significantly enhanced. The details of implantation and representative examples will be presented at the conference to show how we can efficiently handle various onboard antennas and arrays installed on large-scale realistic platforms with the fast MoM-PO and the GPU-based ray tracing technique.