[LAU] OT - impulse response recording

fons at kokkinizita.net fons at kokkinizita.net
Mon Jan 31 11:00:59 UTC 2011


On Sun, Jan 30, 2011 at 07:53:18PM -0800, Ken Restivo wrote:

> On Fri, Jan 28, 2011 at 04:44:47PM +0100, fons at kokkinizita.net wrote:

> > A few percent distortion (most speakers will have this) does not
> > matter if the measurement is done in the correct way: after the
> > deconvolution the distortion products will appear *before* the
> > real IR and they can be removed easily.
> 
> Really? So that little bit of mung before the "snap" in an IR is
> distortion, not "predelay" or "early reflections" or any such thing?
> Wow, that's good to know. I have to get busy and clean up some IR's
> then.

No, the little wobbles that can appear before the main impulse (which
is the direct 'dry' sound and should be removed anyway if the IR is
used for reverb) are not the distortion, but the result of

* the limited frequency range of the sweep (it usually doesn't start
  at 0 Hz and stops before Fs/2),
* the imperfect FR of the speaker + mic.

The distorion appears much earlier.

Suppose you have log sweep going up in frequency by 1 octave per
second. If the speaker produces 2nd harmonic distortion, the
second harmonic will appear in its output 1 second before the
sweep reaches this frequency. The result after deconvolution is
a small 'impulse' 1 second before the main one. Usually you won't
even see it as the time range of the deconvolution is programmed
to start later. 
In a similar way, the 3rd harmonic will appear 1.585 seconds early,
the 4th harmonic 2 seconds, etc.

Ciao,

-- 
FA

There are three of them, and Alleline.



More information about the Linux-audio-user mailing list