Room acoustic simulation have been a practical tool in the design of acoustic spaces for several decades. More recently common prediction methods have been optimized and increasingly applied to real-time systems such as virtual reality and augmented reality applications. Modern computing hardware has made approaches that were previously too computationally intensive applicable for wide bandwidth simulation. Wave-based methods, and especially time-domain solvers for scalar wave equation have gained increasing attention for the use in room acoustic simulation. One of the methods, the finite-difference time-domain method has been proposed as a potential candidate due to convenient scalability properties and the parallel nature of the explicit forms of the method. However, the finite-difference time-domain method has several problematic properties that limit its usage. Due to its direct discretization approach, the dispersion relation in the numerical domain is anisotropic and nonlinear, thus resulting in dispersive wave propagation. Although the properties of the method have been well studied, the implications for its use in auralization have received limited attention. This thesis studies the numerical error involved in the finite-difference time-domain method within the context of auralization. The absolute thresholds of dispersion error are measured for several different explicit finite-difference schemes as well as under different acoustic conditions. Additionally the method is studied in terms of its application for real-time auralization and analysis of modal time evolution in room acoustics. The results indicate that the audibility of the dispersion error is dependent on simulation distance, which determines the amount of the accumulated group delay error in the simulation result. The group delay error is masked inaudible by the absorption of air when the phase velocity error is less than 0.28%. Early reflections are shown to further mask the dispersion error. Usage of the method for real-time virtual reality applications with a distributed approach is found to be limited by the synchronization latencies of the computing hardware. The proposed modal time-evolution analysis offers a new approach for evaluating the time-domain characteristics of local interference patterns in general room geometries that can be considered very difficult with existing approaches.
|Publication status||Published - 2019|
|MoE publication type||G5 Doctoral dissertation (article)|
- finite-difference time-domain method
- room acoustics