[linux-audio-user] Re: Decent reverb

davidrclark at earthlink.net davidrclark at earthlink.net
Thu Feb 3 17:43:02 EST 2005


Hi Marije,

Sorry I did not repost it previously:

http://home.earthlink.net/~davidrclark/linux_audio_users

The IR generator method I'm using is constuction of Green's functions
using eigenfunctions.  So far I've stuck to rectangular geometries, but
do have a circular one (not posted but available).  This method,
in case you're not familiar with it, would be like ray tracing every 
possible ray throughout the entire geometry for a particular pair of
source and solution points.  However, it's usually limited to simpler
geometries and boundary conditions, as you may know.

Ray-tracing is good for more irregular geometries, but it's a lot like
Monte Carlo techniques: You launch something and hope that it produces
a useful result each time --- but usually produces a result that is
similar to previous results.  I'm sure there are optimizations that
are done, just as with MC techniques, but I'm not familiar with
what those might be.  The useful ones probably somehow ensure that
the ray goes very, very close to both source and listener fairly soon.
But it seems that this would be difficult to ensure in advance for
any rays other than the obvious ones.  I would guess that this is
a serious limitation on resolution for large rooms and high frequencies.
It's also bound to be comparatively slow without rather extensive
optimizations.

Regards,
Dave.





More information about the Linux-audio-user mailing list