[LAU] Jconvolver, IRs, and gain

[Fons Adriaensen]

On Thu, Mar 22, 2012 at 09:35:48AM -0400, S. Massy wrote:
> Hello,
> 
> I'm attempting to test some new IRs using jconvolver. I read the
> instruction on the config file syntax and it seems straightforward
> enough, and I am also using the hybridreverb.conf as a base template. My
> problem is that the IRs I want to test are *much* louder than the
> hybrid_reverb ones and are causing distortion when convolved. So far, I
> went by ear and adjusted the gain to about 0.01-0.03, to a point where I
> don't hear any distortion, but that got me wondering. Is there a better
> way to estimate what gain setting to use from, say, the mean and peak
> levels of the file? Is there such a thing as a preferred amplitude
> range/peak for an IR? 

The 'normal' gain for a reverb IR would be such that the direct sound,
if it were part of the IR, would have an RMS level of 1.0. On condition
that the IR was taken with a source to mic distance that produces a good
direct/reverb ratio. 

Another way to get reasonably good levels is to measure the gain
using pink noise. Measure input and output with a RMS meter such
as jkmeter. Note the difference, and adjust the gain in the config
accordingly. 

Not that any distorition you hear is not produced by jconvolver,
since it uses floating point. It just means that you are driving
your DA converters to high.

Ciao,

-- 
FA

A world of exhaustive, reliable metadata would be an utopia.
It's also a pipe-dream, founded on self-delusion, nerd hubris
and hysterically inflated market opportunities. (Cory Doctorow)