[sdiy] Music

[Eric Brombaugh]

On 07/20/2010 09:45 AM, Thomas Strathmann wrote:
> Am 7/20/10 18:32 , schrieb Eric Brombaugh:
>> On 07/20/2010 09:14 AM, Thomas Strathmann wrote:
>>> Raytracing for graphics is possible in realtime even on a CPU and has
>>> been done successfully for some 10 years now. I imagine the only objects
>>> needed for a reverb "scene" would be quads that bound the space in which
>>> the reverb is to be sampled. Kind of like they did with actual rooms, a
>>> loudspeaker, and a microphone in studios. Do I miss something that would
>>> make raytracing impractical for this use case?
>>
>> Real-time for video > 24fps. For audio the sample rate must be a lot
>> higher (2000x). Don't know of any consumer-grade equipment that can
>> compute a 3D model in 21us.
>
> Hehe, you're right. The heat must be getting to me.

Actually, raytracing isn't a horrible idea, but the implementation 
should be different. The key is to assume nothing is moving. Once you've 
established where the sound sources and listener are you can use 
raytracing to create an impulse response. From that you can compute the 
reverb using well-established FIR techniques.

The downside of this is that you won't get that sense of motion/shimmer 
that is prized in the best reverbs. Those use much simpler models of the 
sound environment and introduce subtle modulation to emulate the 
variable density of air, etc.

Eric