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.
