In urban acoustic environments, sound propagation is heavily influenced by obstacles that block the direct line-of-sight, making diffraction a crucial mechanism for plausible acoustic simulations. While the Uniform Theory of Diffraction (UTD) resolves the unnatural discontinuities present in standard Geometric Acoustics (GA), the computational cost increases exponentially with each additional diffraction order. This thesis investigates the perceptual relevance of diffraction and reflection orders to determine an optimal threshold that can improve computational efficiency without compromising perceptual realism.

Using the Pynamic framework and the MOSQITO sound quality library, a workflow was developed to generate time-dependent psychoacoustic metrics, specifically loudness and sharpness, integrated with Just Noticeable Difference (JND) thresholds. The study analyzed various geometric scenarios, from simple 2D convex and concave edges to 3D building corners and complex urban intersections.

The results demonstrate that the optimal simulation order is highly geometry-dependent. For models with purely convex edges, a simulation of third-order diffraction is required to smooth out discontinuities at shadow boundaries. However, in scenarios with concave edges or in complex urban environments with ground reflections, the masking effect caused by strong reflected energy often renders higher-order diffractions inaudible, allowing for reduced simulation orders. Furthermore, for an urban canyon model characterized by infinite parallel reflections, it is recommended to neglect reflections beyond the third order. Finally, the analysis reveals that stochastic signals with increased low-frequency energy, such as pink noise, and motion trajectories extending deep into acoustic shadow zones represent the stricter conditions for defining perceptual thresholds. These findings provide a set of quantitative guidelines for optimizing dynamic urban acoustic simulations.

Melden Sie sich hier an um Einladungen zu den Kolloquium-Vorträgen per E-Mail zu erhalten.
Register here to receive the invitations to colloquium talks via e-mail.

Zoom-Meeting-ID: 954 4073 3814
Passwort: 450783