A Hardware-Oriented Finite-Difference Time-Domain Algorithm for Sound Field Rendering

Abstract

Sound field renderings are data-intensive and computation-intensive applications. An alternative solution is to directly implement sound field rendering algorithms by using hardware. In this paper, a hardware-oriented finite-difference time-domain (FDTD) algorithm named HO-FDTD is proposed for sound field rendering, which has no complex operations involved, and consumes small hardware resources. In a sound space with 32,768 elements surrounded by rigid walls, the hardware simulation results are in good agreement with the software simulation results except for the one-cycle delay. In the software simulation, when the element scale is 32×32×32 and the time steps are 20,000, the HO-FDTD speeds up computations by 19% against the updated digital Huygen's model (DHM) and Yee-FDTD, and by 132% against the original DHM. Compared with the software simulation, the hardware systems with the parallel architecture and the time-sharing architecture enhance their calculation performance significantly in the case of different element scales, and provide a higher data throughput.