Left to right: John Strong, Scott Pfeiffer, Ken Platz (section chair), and Nick Dulworth. Chicago AES Section, March 28, 2018.
Room auralization begins by creating a model of the room using CAD software. Inevitably, this model is significantly simpler than the actual room, in order to save both labor in creating the model and hours of computation time needed to extract data from it. As an example of this, individual stair steps are not modeled, rather stairs are modeled as a ramp that is given a diffusion coefficient that mimics the behavior of real stairs.
Traditionally, room simulation has been accomplished by computing the impulse response between an omnidirectional point source on stage, and the equivalent of an omnidirectional microphone in the audience. While this approach has its merits, it isn't really faithful to real sources and listeners. An instrument on stage emits sounds differently in each direction, and a listener in the audience will perceive that sound differently whether it arrives from the front, sides, or above.
To overcome these challenges, the team at Threshold has worked to simulate directivity for both sources and listeners. They created custom source material by recording musicians in an anechoic chamber using five microphones: four around the musician and one above. In the simulation software, each of these were treated as having a cone-shaped directivity emanating from the source location. On the listener side of things, the impulse responses are computed as a 2nd order ambisonic signal, which requires nine channels. All this adds up quickly. Five microphones times nine ambisonic channels means 45 impulse responses need to be generated and then convolved with the source material for each sound source.
The actual computation of the impulse responses is done via ray-tracing. Think of shooting thousands of virtual ping pong balls out from the source location and following their path as they bound off of different surfaces.
Finally, to listen to all of it, Threshold has a listening room with 12 speakers arranged around a group of seats. They have custom software that allows the user to listen to different source signals (such as instrument, singer, or dramatic monologue) and different acoustic treatments of the virtual room. While it is doubtful that any trained listener would mistake the auralized sound for that of a real concert hall, the level of realism achieved is remarkable. It's easy to see how this is a powerful tool for designing an auditorium that only exists as a computer model.
