In RevolVerb, I've noticed that some impulse responses don't sound too loud - the wet control can be much higher with the mix sounding ok. Looking at one of the "good" impulse responses, it has a much lower signal level (circa -20dB), and it's also encoded with 24 bits.
So, I thought I'd try and "fix" one of the bad impulse responses. I converted it to 24 bits first (to retain resolution), and then normalised to -20dB. This did fix the level problem, however when I listened VERY closely, I could hear a difference. The specific impulse response I'm testing with at the moment is this one: http://www.impulseresponse.org/Real%20S...hurch2.zip (the Chicago 2 Micset 1 recording). Using the original, and playing middle G (S Erard, but it probably doesn't matter), I hear a very sharp resonance which occurs about half a second after playing the note. Using the modified version of the impulse response, this resonance doesn't sound anywhere near as sharp - it's sort of "smeared". So, I've decided that "fixing" the impulses in this manner just doesn't work - I don't think RevolVerb uses enough of the bits in the higher resolution impulse response.
I loaded this same I.R into two commercial products, and this sharp resonance was present equally in both. I then loaded my modified I.R into the commercial products, and it sounded identical, so I didn't do anything wrong when I processed it.
I think RevolVerb is probably working ok with 16-bit impulse responses that have a high signal amplitude (except for the level problem). I do NOT think it's working 100% for high resolution recordings where the impulse response has not been amplified/normalised to a high signal level.
Greg.
p.s I think I'll end up buying a convolution reverb.
Last edited by skip (09-02-2009 14:45)