lowen at pari.edu
Mon Jan 19 11:54:34 EST 2015
On 01/19/2015 11:39 AM, Jay Ashworth wrote:
> ----- Original Message -----
>> From: sascha at radio42.de
>> 2015-01-19 16:07, Cowboy wrote:
>>> Rivendell, being a *professional* automation system by and for
>>> *professional* broadcasters, I can see where the production rat
>>> has spent hours getting a piece "just so" to have it destroyed by
>>> a machine's idea of what "sounds even" to a human.
>> "Rivendell, being a *professional* automation system" -> That's one
>> reason more why Rivendell might need Loudness Normalization.
>> Big disagree. In times of the loudness war it is even more necessary
>> to have a proper (loudness) normalization.
> "Normalization", of whatever type, is something which happens inter-track,
> across a library, not intra-track, inside a single song.
This is a rather interesting discussion. I remember back in the 90's,
when the CoolEdit 'normalize' was peak only in nature, of going through
tracks and peak normalizing everything during initial production. Yes,
I know that means different tracks had (and have) different 'loudness'
since loudness has almost nothing to do with peak amplitude. But I also
knew that the Optimod at the transmitter was going to completely blow
away any and all loudness normalization I was doing anyway. I peak
normalized for a completely different reason than loudness levelling
(which peak normalization doesn't do anyway): maximizing signal to
quantization noise ratios. With 16 bit audio, peak normalization can
make things sound marginally better when it is being mashed and munged
by the Optimod or Omnia at the transmitter. In my case, it was for an
AM, where modulation isn't just about loudness, but also about coverage;
if the positive peaks were hitting anything less than 110% the GM wasn't
happy, at all..
With 24 bit production being somewhat the norm these days, the S/N
ratio, at least for quantization noise, isn't nearly as important. But
getting the loudness 'just so' still runs afoul of the processor at the
transmitter; that is, a machine is making loudness decisions for you
anyway in the air chain. And the AM Optimods do some of the absolutely
most invasive processing on audio that you can imagine; the old 9100,
for instance, started out with a phase scrambling all-pass filter, which
mangles peaks like nobody's business. Bob Katz' Mastering Audio book
covers this thoroughly, by the way, and should be required reading for
anyone producing cuts for air.
Hi Jay; different audience from NANOG, no?
More information about the Rivendell-dev