[LAU] Batch normalising wav files

Folderol folderol at ukfsn.org
Sat Dec 25 12:01:49 UTC 2010


On Sat, 25 Dec 2010 12:40:33 +0100
fons at kokkinizita.net wrote:

> On Fri, Dec 24, 2010 at 06:51:18PM +0100, Jörn Nettingsmeier wrote:
> 
> > On 12/24/2010 06:07 PM, Q wrote:
> 
> > > Out of interest, how high can intersample peaks get above the highest
> > > peaks in a file?
> > 
> > imagine the positive half of a sine wave, so that two consecutive
> > samples representing the very top of the wave are at 0dBFS. it is clear
> > that between those samples, the real peak value of the sound must be
> > higher than 0dBFS, and the reconstruction filter will see this higher
> > value. that means unless your analog stage has headroom for this, it
> > will clip. the higher the frequency, the higher above full scale those
> > inter-sample peaks can be.
> 
> In theory it could be any value. In practice it's limited by the lenght
> of the antialiasing filter. You get the maximum output for a series of
> samples at max amplitude (+ or - 1) and matching the sign of the IR of
> the filter at half a sample delay.
> 
> For example, if '+' means a sample of +1, and '-' a -1, then the
> sequence of 20 samples:
> 
>  -+-+-+-+-++-+-+-+-+-  
> 
> will produce of peak of more than 8 dB above FS.
> 
> It's very unlikely to get such a sequence with natural sounds,
> but quite possible with synthetic signals or when using agressive
> mastering techniques.
> 
> Ciao,

Interesting.

When I first started to use a digital recorder I used -6dBFS as my target
value simply on a 'gut feeling' basis. Seems I wasn't so far out then. 


-- 
Will J Godfrey
http://www.musically.me.uk
Say you have a poem and I have a tune.
Exchange them and we can both have a poem, a tune, and a song.


More information about the Linux-audio-user mailing list