On fre, 2005-02-04 at 22:09 +0100, David Olofson wrote: [...] > > I think this method will not catch the mild doppler effect arising > > from turbolence? (.. nor the *wild* turbolence, if house is on > > fire!) > > I guess not, because that's kind of equivalent to the room changing > shape all the time, I think - and you can't capture that without a > constantly "morphing" impulse response. No idea how you would go > about recording something like that, though.
Neither do I .. and I think it would literally take forever. But since the wigglings and wobblings in the reverb-tail are all slight variations on a common theme, I might like to try some of the following: Use a 'wobble-generator' to crossfade between two (or more) samples taken at different locations. Cheat and 'Wobble' the input and 'wiggle' the tail slightly. > > I *do* have an idea about what kind of problems that would pull in, > though: It would be like looping a sampled waveform, except each > sample is an impulse response... So, you need a nice, click free, non > repetitive sounding loop. :-) Good thinking! Click free loop can be automated. Quantifying 'nice' is harder ... Non repetitive ... Say we morph between two 'nice' loops: Loop A is 2.00 seconds Loop B is 2.71 seconds Crossfade 3.14 seconds (full circle) ... then we would get in the right ballpark for a simulation of the sameness, though everchanging sound of complex sustained notes. This is for the sustained part though, and might not translate verbatim to the decay of a room? [...] > This is a well known way to "cheat" to save cycles. Or perhaps it is: A simplified model of a rooms characteristic decay combined with a simplified model of the turbolence, sounds better than without the simplified turbolence. In a parallel universe it could argued that, a speech-synthesizer singing 'Lieder' is much more convincing if you have the appropiate Bontempi Piano to go along with it. -- ( ) c[] // Jens M Andreasen