Re: [Algorithms] Spherical harmonics for room acoustic modelling

[Jon W. <jw...@gm...>]

>  I'd have doubts you'd get good enough response time on CPU alone



Actually, the CPU can give you excellent response times if you're using a
convolution based solution. DJBFFT is a very high performance, unencumbered
convolution implementation, and it takes a minimal amount of CPU for a
single sound stream. The problem comes when you want to combine delay lines
to do overlap-add long convolution (or use the variable-window-size that's
allegedly still patented by Lake DSP). It starts adding up! The GPU might be
good at larger convolutions, but it has higher latency than the CPU.

I'm not sure Spherical Harmonics is the right solution. Sound reflection is
a function of frequency mapping to absorption and directional spread. And
whereas the eye is only sensitive to three separate frequencies, the ear is
sensitive to thousands... And in SH based PRT, you often collapse the
direction-dependent response into a single term, and filter the three
frequencies using a separate RGB "diffuse reflectance" texture. If you
applied the same to sound, you'd probably get a better than nothing, but you
might only get good enough that the problems start to become obvious :-)
(Bass, Mid and Treble, with a fixed reflectance function)

Sincerely,

jw


--
Americans might object: there is no way we would sacrifice our living
standards for the benefit of people in the rest of the world. Nevertheless,
whether we get there willingly or not, we shall soon have lower consumption
rates, because our present rates are unsustainable.



On Sat, Jul 17, 2010 at 2:23 PM, Oscar Forth <
os...@tr...> wrote:

> OK maybe not spherical harmonic based but gives the same sorta results and,
> therefore, just as interesting :)
>
>