Summary

International Symposium on Antennas and Propagation

2006

Session Number:3POS

Session:

Number:3POS-7

Improvement of Accuracy of Fixed Point Arithmetic FDTD Method

Ryo YAMAGUCHI,  Hidetoshi SUZUKI,  Yuta TAKAGI,  Shinji UEBAYASHI,  

pp.1-4

Publication Date:2006/11/2

Online ISSN:2188-5079

DOI:10.34385/proc.34.3POS-7

PDF download (116.4KB)

Summary:
From the viewpoint of application to antenna analysis, this paper examines the calculation error of a fixed point arithmetic that causes a problem in the implementation of the finite difference time domain (FDTD) method in field programmable gate arrays (FPGAs). We evaluate the relative error while focusing on the difference between the fixed point arithmetic and the floating point arithmetic rather than the calculation accuracy of the FDTD method itself. We achieve improvement in the relative error in the analyzed field by introducing a technique that scales the intrinsic impedance in order to cancel significant digits of the magnetic field component that are generated based on the ratio of the electric field to the magnetic field. As a result, we achieve operation with the relative error improvement of more than 50 dB using a 32-bit fixed point arithmetic in free space field analysis.