[LAU] Batch normalising wav files

[Jörn Nettingsmeier]

On 12/25/2010 12:49 PM, fons at kokkinizita.net wrote:
> On Fri, Dec 24, 2010 at 05:31:59PM +0100, Jörn Nettingsmeier wrote:
> 
>> in a presentation i heard recently, the presenter played a metallica
>> single, in direct comparison to pink noise at full scale. the metallica
>> mix was significantly louder. go figure...
> 
> That's cheating. Pink (or any color) noise doesn't have a defined
> peak/RMS ratio. If it's Gaussian its peak value is unlimited.

true. the fun thing was you could see his slide announcing the next demo
snippets. imagine reading this:

1. peak-normalized metallica single
2. pink noise at full scale

now when you frequently use your average run-of-the-mill pink noise
generator at -20dBFS to test speaker arrays, you do have a gut feeling
of how "loud" pink noise is. in fact, for reasonable time windows,
you're pretty likely to get an average loudness - with JAPA's pink noise
generator for instance, the loudness osciallates around a mean value
such that if your meter integrates for longer than, say, a second, the
reading will be pretty much constant.
so the second line was enough to make me cringe and grope for my ear
plugs. so it was quite a laugh when the perceived loudness was a lot
lower than for the first example.