Implementing PTD efficiently for electrically large objects

Abstract

Modeling electromagnetic characteristics of electrically large and complex objects is challenging. Electromagnetic interaction can be complicated. Every modification to part of a structure requires re-analyzing the structure. This is tedious and time-consuming. Therefore, mathematical tools that enable fast modeling of electromagnetic characteristics is crucial in understanding the scattering behavior of electrically large and complex object. Full wave methods like method of moments (MOM) and multilevel fast multipole method (MLFMM) have been improved tremendously over the years and are invaluable. However, they are still far from ideal to fit into the workflow at design phase due to huge computational cost. A better tradeoff to meet aggressive timeline is to perform rapid prototyping using hybrid high frequency methods. In practice, physical optics (PO), shooting and bouncing rays (SBR) and physical theory of diffraction (PTD) are reasonably accurate at predicting most significant scattering signature of electrically large object. With such rapid prototyping tools, engineers are able to quickly identify electromagnetic issue in early stage. Fast and accurate prediction of scattering behavior also enables engineers to quickly iterate between solutions to explore design tradeoffs. In addition to high performance computing facility, this even opens up the capability to perform parametric studies. With PO and SBR implemented efficiently on graphic processing unit (GPU), we question the underutilized central processing units (CPUs) in our application. We have decided to explore the potential of modern CPUs and this paper discusses an efficient implementation of PTD on modern CPUs. In particular, we stress the combination of efficient algorithms and appropriate data structure in order to achieve great performance. Numerical result shows good accuracy and efficiency tradeoff to those obtained from MOM.