On 11/09/2016 01:23 AM, al davis wrote:
It seems to me that peak normalization of -13 doesn't make sense,
because it is just throwing away 13 db of level with no real benefit.
There is a real loss when playing on a cheap sound card that has output
level too low to begin with, and maybe also has a noise level higher
than it should be.
Digital is not magnetic tape, the days of recording to the maximum
available level to get the best signal-to-noise ratio are gone.
I dare hope that no one sends a signal to air with an D/A converter so
old and cheap that its noise is a problem. Even 100$ USB interfaces do
not have that problem.
To me peak normalization at -13dBFS makes sense in the way that a
program or song with correct-to-small dynamic range will average close
to -23LUFS when peak normalized at -13dBFS.
Your entire library and broadcast chain must be calibrated to a
reference level that leaves headroom, lots of it.
If your standard overcompressed pop song sits at -2dBFS in your system
then how can you deal with very dynamic programs ?
The answer is in the initial question, you can't.
The correct way is to shift the average level of everything to a
reference low enough to accomodate every type of program.
The EBU R128 (or ITU-R BS.1770-2 for rest of the world) provides a norm
to implement this.
This way, an overcompressed song will average output at REF_LEVEL, with
peaks at (say) REF_LEVEL + 3dBFS.
A dynamic program will average at REF_LEVEL, with peaks at REF_LEVEL +
20dBFS if need be.
It's not the dynamic program that's "lower than it should be".
It's the vast majority of music production and "amateur" made programs
that are _way_ louder than they should.
. e
_______________________________________________
Rivendell-dev mailing list
Rivendell-dev@lists.rivendellaudio.org
http://caspian.paravelsystems.com/mailman/listinfo/rivendell-dev