Effect of model detail level on room acoustic computer simulations

Abstract

Computer modeling of room acoustics is limited due to approximations regarding wavelength to surface size ratios and diffraction. In fact, it is suggested that a model with high geometrical fidelity may actually lead to less accurate predictions. This project seeks to clarify how the level of model detail affects the accuracy of these simulations. Three models of the University of Nebraska’s Strauss Recital Hall have been generated at varying levels of detail; low (a rough model of the essential architectural dimensions), medium (a more accurate representation of the space), and high (a complete rendering of the surfaces within the space). The room acoustics modeling program Odeon is used to calculate objective measures (i.e., reverberation time and clarity) and create auralizations for each model. Meanwhile, the actual acoustical properties of Strauss Hall have been measured. The objective measure results are compared between the models and against measured data. Additionally, subjective tests are run with auralizations from all three models and the original space to determine if the difference in level detail can be perceived psychoacoustically. Quantitative results from these comparisons are presented to help users of room acoustic simulation software understand how the level of model detail will produce maximum accuracy.