Sheaffer, J and Fazenda, BM 2010, 'FDTD/K-DWM simulation of 3D room acoustics on general purpose graphics hardware using compute unified device architecture (CUDA)' , Proc. Institute of Acoustics, 32 (5) .

PDF - Published Version Download (604kB)

Abstract

The growing demand for reliable prediction of sound fields in rooms have resulted in adaptation of various approaches for physical modeling, including the Finite Difference Time Domain (FDTD) and the Digital Waveguide Mesh (DWM). Whilst considered versatile and attractive methods, they suffer from dispersion errors that increase with frequency and vary with direction of propagation, thus imposing a high frequency calculation limit. Attempts have been made to reduce such errors by considering different mesh topologies, by spatial interpolation, or by simply oversampling the grid. As the latter approach is computationally expensive, its application to three-dimensional problems has often been avoided. In this paper, we propose an implementation of the FDTD on general purpose graphics hardware, allowing for high sampling rates whilst maintaining reasonable calculation times. Dispersion errors are consequently reduced and the high frequency limit is increased. A range of graphics processors are evaluated and compared with traditional CPUs in terms of accuracy, calculation time and memory requirements.

Actions (login required)