Wayverb: A Graphical Tool for Hybrid Room Acoustics Simulation

Acoustic simulation aims to predict the behaviour of sound in a particular space. These simulations can be carried out on commodity computing hardware, and are faster, cheaper, and more convenient than building and recording a physical space. This is useful to architects, who need to know that their designs will meet exacting specifications before building starts. It is also useful to musicians and sound designers, who can use the simulation results to create pleasing, precise, and immersive audio effects.

Currently, users are limited in their choice of simulation software, as existing solutions focus either on speed or on overall accuracy. Fast simulation techniques are often inaccurate at low frequencies, while more accurate techniques become prohibitively slow at higher frequencies. There is a clear need for a program which is fast, accurate at all frequencies, and easy to use without specialist training.

This project identifies a hybrid acoustic simulation technique which combines the efficiency of geometric simulation with the accuracy of wave-based modelling. This fusion of simulation techniques, not available in any existing piece of simulation software, gives the user the flexibility to balance accuracy against efficiency as they require. This hybrid method is implemented in the Wayverb program. The program is made free and open-source, with a simple graphical interface, differentiating it from hybrid simulators found in the literature which are all private and closed-source. In this way, Wayverb is uniquely accurate, efficient, and accessible.

Rather than presenting an entirely new simulation method, the Wayverb project surveys algorithms from the literature, employing those which are deemed most appropriate. It focuses on issues of practical implementation. In particular, for increased performance, Wayverb uses graphic hardware to accelerate calculations via parallelisation. Test data is presented to demonstrate accuracy.

The Wayverb program demonstrates that the hybrid method is efficient enough to be viable in consumer software, but testing reveals that simulation results are not directly ready for production usage. The differing properties of the wave-based and geometric methods can result in different onset and decay times in the upper and lower regions of the output spectrum, which reduces the perceived quality of the output despite increased low-frequency accuracy. Avenues for further research are suggested in order to improve the quality and usability of the software.