Efficient acoustic modelling of large acoustic spaces using finite difference methods.

Abstract

Time domain methods for solving wave based acoustic models have been of continued interest and development since early work by key figures such as Bottledooren, as these methods can provide a simple and flexible approach for simulating a wide range of acoustic phenomena such as room modes. The nature of many time domain difference methods present significant computational resource requirements, as the size, sampling rate and inherent stability of the simulation has a distinct impact on the memory and execution time required for the simulation to give a satisfactory result. In this study the execution speed is analysed, for variations of the finite difference time domain method that may provide some increase in computation speed for large domains. It is suggested that leveraging a dynamic windowing method may reduce total computation time for some simulations, by decreasing the number of computations per time-step in the early stage of a simulation.