A Spatial Filtering Technique in FDTD for Extending the Stability Limit and Its Applications in Subgrids

Chun Chang

The Edward S. Rogers Sr. Department of Electrical and Computer Engineering (Electromagnetics)

Abstract:

The Finite-Difference Time-Domain (FDTD) method has been popular to study the propagation
of electromagnetic waves for many years. Despite being simple and versatile, FDTD is sometimes computationally inefficient due to its uniform grid size and strictly-bounded time step. These 2 factors can dramatically increase the execution time and memory required by an FDTD simulation even if only one tiny object exists in the simulation domain. If so, the grid size has to be reduced accordingly, then more grids (memory) are needed to represent the same structure. Smaller grids also lead to smaller time step because the stability limit requires the maximum of a time step be proportional to the grid size. Hence more time steps (execution time) are needed to reach the same time instant.

This talk will introduce a new spatial filtering technique that enables the time step to surpass the stability limit without causing instability, and therefore accelerates FDTD simulations with a fewer number of time steps required. The price of surpassing the stability limit is to perform Discrete Fourier Transform (DFT) in every FDTD cycle. Hence this filtering technique couples well with subgrids by only being performed locally. The drawbacks and potential future research topics will also be covered.

 

Biography:

Chun Chang is an M.A.Sc. student supervised by Prof. Costas Sarris in the electromagnetics group. He studies on the propagation of electromagnetic waves with FDTD and how to optimize FDTD itself.