[linux-audio-user] Fwd: [Jamin] Re: soft clip: Achieving Gain, inconsequential overloads

[Anahata]

On Tue, Apr 13, 2004 at 06:34:20AM -0700, R Parker wrote:
> The best sense I can make of Normalizing is I might
> want to apply it to all songs after they are mastered.

That's what I understand to be its most appropriate use.

> Then again I probably won't do that either because
> mastering is the process of achieving appropriate
> levels amongst groups of songs.

Correct. The normalize program will do a good job of quickly levelling
out the differences between tracks when they are badly matched, but it
can't determine that you actually *want* a particular track a few dB
quieter than the others for artistic reasons.
 
> Is "rms" another method for calculating peaks? What
> does the acronym stand for?

Literally root-mean-square. Its significance in this context is that it
relates to power, and hence arguably to audible loudness, better than
peak measurements do, and better than taking a simple mean of absolute
values of samples or voltage.

Normalizing by the criterion of RMS level is therefore the best option
for bringing different tracks to something like the same perceived loudness 
level, at least better than peak normalization will do.

On a compressor, detecting peak or average (RMS) levels is one of those
things that you tweak for best results. Occasionally, with certain types
of material, it makes a difference; much of the time it doesn't matter
very much. Peak detection tends to apply to limiting to protect following
stages from clipping, and as with normalization RMS is more for compression 
consistent with audible loudness.

-- 
Anahata
anahata at treewind.co.uk       Tel: 01638 720444
http://www.treewind.co.uk    Mob: 07976 263827