International Symposium on Antennas and Propagation
A Study on Accuracy of FDTD Method Using Fixed-Point Arithmetic
Hidetoshi Suzuki, Ryo Yamaguchi, Shinji Uebayashi,
PDF download (187.3KB)
Although the Finite Difference Time Domain (FDTD) method is extensively used as an electromagnetic field analysis technique, it spends considerable time in calculating electromagnetic fields one after the other for each cell. Supercomputers and PC clusters are put to practical use to solve this problem. Furthermore, implementing the FDTD method into a Field Programmable Gate Array (FPGA) is examined as an alternative technique to solve this problem. Although this technique is put to practical use in a simple model such as 2D-FDTD and Mur’s absorption boundary conditions, for complicated models such as 3D-FDTD and the perfectly matched layer (PML) absorption boundary conditions it is not practical. Usually, the FDTD method is calculated using a floating-point arithmetic. The arithmetic in the FPGA, however, is a fixed-point arithmetic in many cases. Therefore, it is important to clarify the accuracy of the fixed-point arithmetic when implementing it into an FPGA for the FDTD method. In this paper, analysis based on a half-wavelength dipole antenna and lossy material with a half-wavelength dipole antenna is performed and the operation accuracy is clarified. The results show that the operation accuracy of 0.01 or less is achieved when the bit length is 28 or more.