24 Dec
2010
24 Dec
'10
2:53 p.m.
Hello
I'm wanting to batch process a bunch of wav files, normalising them all
to -0.3 dBFS.
I thought I could do this with normalize/normalize-audio, but it doesn't
look to be possible. The man page suggests that simply analysing the
file and raising the level so the loudest peak is 0 dBFS isn't
normalising, which is news to me:
" --peak
Adjust using peak levels instead of RMS levels. Each file will be
adjusted so that its maximum sample is at full scale. This just gives a
file the maximum volume possible without clipping; no normalization is
done."
That would be okay, except I want them normalising to something other
than full scale.
Is there a quick and easy way to do this?
I normally avoid command line stuff as I waste so much time trying to
figure out how to get one little one-off task done that I could have
done it manually in the meantime. However, although I've wasted a lot of
time so far, I have so many files to process that I'm willing to give it
a shot.
Thanks in advance
Q
24 Dec
24 Dec
5:32 p.m.
On 12/24/2010 02:53 PM, Q wrote:
Hello
I'm wanting to batch process a bunch of wav files, normalising them all
to -0.3 dBFS.
that's high! you will almost certainly get distortion in the analog
stages of cheaper playback equipment with a signal that hot. think
inter-sample peaks.
I thought I could do this with
normalize/normalize-audio, but it doesn't
look to be possible. The man page suggests that simply analysing the
file and raising the level so the loudest peak is 0 dBFS isn't
normalising, which is news to me:
depends. that's peak normalisation. what you probably want is loudness
normalisation, which adjusts the levels so that the perceived loudness
is constant across tracks.
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...
" --peak
Adjust using peak levels instead of RMS levels. Each file will be
adjusted so that its maximum sample is at full scale. This just gives a
file the maximum volume possible without clipping; no normalization is
done."
that is a funny definition of normalisation.
That would be okay, except I want them normalising to
something other
than full scale.
Is there a quick and easy way to do this?
like julien suggested, use a standard peak normalizer and then attenuate
by 0.3dB )(if you are sure that's what you want to do).
6:08 p.m.
Jörn Nettingsmeier wrote:
No, peak normalisation is what I'm after; this isn't a CD collection I'm
processing.
I get the impression that the term normalisation has taken on another
meaning among non-music producers to mean replay gain normalisation for
whole albums or music collections. Doesn't replay gain either peak
normalise the loudest track and raise all the others by the same amount
(to preserve inter-song dynamics), or in some case do loudness
normalisation? Personally, I don't bother with such things, volume knobs
are what we evolved fingers for!
When you put it as simply as that it's blindingly obvious! I can't
believe it didn't occur to me and one extra step when batch processing
is not a big deal at all. Blame the cold weather and a brain freeze!
Thanks again
Q
On 12/24/2010 02:53 PM, Q wrote:
I got the impression most masters were peak normalised to 0 dBFS these
days and less frequently to slightly under to avoid playback issues. The
original files were all done to 0 and Rezound showed a pink clip line in
one or two of the files, which is why I opted for slightly below for the
ones I processed.
Anyway, playback on cheap equipment will not be an issue for these files
-- they won't be played on music players.
Out of interest, how high can intersample peaks get above the highest
peaks in a file?
Hello
I'm wanting to batch process a bunch of wav files, normalising them all
to -0.3 dBFS.
that's high! you will almost certainly get distortion in the analog
stages of cheaper playback equipment with a signal that hot. think
inter-sample peaks. I thought I could do this with
normalize/normalize-audio, but it doesn't
look to be possible. The man page suggests that simply analysing the
file and raising the level so the loudest peak is 0 dBFS isn't
normalising, which is news to me:
depends. that's peak normalisation. what you probably want is loudness
normalisation, which adjusts the levels so that the perceived loudness
is constant across tracks.
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...
I'd rather listen to pink noise any day!
" --peak
Adjust using peak levels instead of RMS levels. Each file will be
adjusted so that its maximum sample is at full scale. This just gives a
file the maximum volume possible without clipping; no normalization is
done."
that is a funny definition of normalisation. That would be okay, except I want them
normalising to something other
than full scale.
Is there a quick and easy way to do this?
like julien suggested, use a standard peak normalizer and then attenuate
by 0.3dB )(if you are sure that's what you want to do).
7:01 p.m.
On 12/24/2010 06:07 PM, Q wrote:
No, peak normalisation is what I'm after; this isn't a CD collection I'm
processing.
ok. though i have a hard time seeing any use for peak normalisation
unless you want to deliver your material at low bit depths, such as a cd.
I get the impression that the term normalisation has taken on another
meaning among non-music producers to mean replay gain normalisation for
whole albums or music collections. Doesn't replay gain either peak
normalise the loudest track and raise all the others by the same amount
(to preserve inter-song dynamics), or in some case do loudness
normalisation? Personally, I don't bother with such things, volume knobs
are what we evolved fingers for!
correct replay gain should strive for constant perceived loudness, so
that you can use those wonderful evolved fingers for more important
things, such as doing the dishes or ironing, without having to rush to
the volume knob after each track.
i hope that come spring, your freshly thawed brain will even realize
that you don't want peak normalisation at all :)
ok, down from the soapbox, and season's greetings,
jörn
Jörn Nettingsmeier wrote:
I got the impression most masters were peak normalised to 0 dBFS these
days and less frequently to slightly under to avoid playback issues.
eat shit, millions of flies can't be wrong :)
mastering that hot is a major technical blunder proposed by idiots with
either bad hearing or a profound hate for music.
I'm
wanting to batch process a bunch of wav files, normalising them all
to -0.3 dBFS.
that's high! you will almost certainly get distortion in the analog
stages of cheaper playback equipment with a signal that hot. think
inter-sample peaks. 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.
I thought I could do this with
normalize/normalize-audio, but it doesn't
look to be possible. The man page suggests that simply analysing the
file and raising the level so the loudest peak is 0 dBFS isn't
normalising, which is news to me:
depends. that's peak normalisation. what you probably want is loudness
normalisation, which adjusts the levels so that the perceived loudness
is constant across tracks. " --peak
Adjust using peak levels instead of RMS levels. Each file will be
adjusted so that its maximum sample is at full scale. This just gives a
file the maximum volume possible without clipping; no normalization is
done."
that is a funny definition of normalisation. like julien
suggested, use a standard peak normalizer and then attenuate
by 0.3dB )(if you are sure that's what you want to do).
When you put it as simply as that it's blindingly obvious! I can't
believe it didn't occur to me and one extra step when batch processing
is not a big deal at all. Blame the cold weather and a brain freeze! 
25 Dec
25 Dec
12:40 p.m.
On Fri, Dec 24, 2010 at 06:51:18PM +0100, Jörn Nettingsmeier wrote:
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,
--
FA
There are three of them, and Alleline.
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. 
1:01 p.m.
On Sat, 25 Dec 2010 12:40:33 +0100
fons(a)kokkinizita.net wrote:
On Fri, Dec 24, 2010 at 06:51:18PM +0100, Jörn
Nettingsmeier wrote:
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.
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. 
12:47 p.m.
Em Fri, 24 Dec 2010 17:07:48 +0000
Q <lists(a)quirq.ukfsn.org> escreveu:
Out of interest, how high can intersample peaks get
above the highest
peaks in a file?
here mix to max. -1,0 dBFS is recommended:
1. Leave lots of headroom in your mix: This is the
obvious solution,
and it requires no special processing. As long as you keep the peak
levels in your mix below 0dBFS, the DAC will never encounter the
situation described above. The consensus I’ve found is that real-world
inter sample peaks never exceed 1dB, so keep your levels below -1.0
dBFS and you’ll be fine.
from: http://www.hometracked.com/2007/11/08/prevent-intersample-peaks/
2. Use a mastering tool designed to prevent these
peaks: iZotope’s
Ozone is one such tool. The loudness maximizer has a one-click option
to prevent inter-sample clipping.
3. Use a metering tool that highlights the peaks: SSL recently
released the X-ISM Plug-In which adds an inter-sample peak meter to
any VST-capable DAW.
wondering if any of those tools in 2. or 3. are available for linux!?
here it is shown that intersample peaks can eventualy/teoriticly exceed
10db!!!:
If you want to see something really bad on the
oversampled meter -
try a sequence of maximum and minimum values that goes like this:
"1010101101010" - notice that the alternating 1's and 0's suddenly
change direction in the middle. The results depends on the filter
being used in the reconstruction, with the intersample peak easily
exceeding 10dB!
from:
http://www.gearslutz.com/board/tips-techniques/334385-intersample-peaks.html
3:29 p.m.
On Sat, Dec 25, 2010 at 6:47 AM, Fabio <capoeirista(a)arcor.de> wrote:
X-ISM is very compliant and simple VST plugin that will run in any
host on linux that supports win/x86 VST plugins.
> 3. Use a metering tool that highlights the peaks:
SSL recently
interesting definition of "recently" - i got my copy about 2-3 years ago :)
released the
X-ISM Plug-In which adds an inter-sample peak meter to
any VST-capable DAW.
wondering if any of those tools in 2. or 3. are available for linux!?
12:49 p.m.
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.
In practice, at normal sample rates, you get something like 11.5 dB
for a few minutes and around 13 dB after an hour or so. But if you
wait long enough you can get any value.
Currrent mastering techniques easily do better than 11.5 dB, so
no surprise...
Ciao,
--
FA
There are three of them, and Alleline.
2:16 p.m.
On 12/25/2010 12:49 PM, fons(a)kokkinizita.net wrote:
On Fri, Dec 24, 2010 at 05:31:59PM +0100, Jörn
Nettingsmeier wrote:
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.
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. 
12:39 a.m.
Hi,
On Friday 24 December 2010 14:53:44 Q wrote:
I'm wanting to batch process a bunch of wav files,
normalising them all
to -0.3 dBFS.
I thought I could do this with normalize/normalize-audio, but it doesn't
look to be possible. The man page suggests that simply analysing the
file and raising the level so the loudest peak is 0 dBFS isn't
normalising, which is news to me:
" --peak
Adjust using peak levels instead of RMS levels. Each file will be
adjusted so that its maximum sample is at full scale. This just gives a
file the maximum volume possible without clipping; no normalization is
done."
That statement there is bullshit. Scanning a wave-file for maximum absolute
sample and then scaling that to make the file fit into [-1:1] with the abssolute
of the maximum sample at 1 _is_ normalization. That is the very definition of
normalization.
That would be okay, except I want them normalising to
something other
than full scale.
Given the fact that [-1:1] is just an arbitrary setting, there is no problem
with normalizing/scaling to [-0.9:0.9] or within the range of -3dB instead of
0dB...
Is there a quick and easy way to do this?
I seem to recall that normalize-audio can actually do this. Otherwise have a
look at the fine tools contained in libsoundfile.
Have fun and a merry Christmas,
Arnold
1:49 p.m.
On Sat, 25 Dec 2010 00:39:12 +0100
Arnold Krille <arnold(a)arnoldarts.de> wrote:
I seem to recall that normalize-audio can actually do
this. Otherwise have a
look at the fine tools contained in libsoundfile.
Have fun and a merry Christmas,
Arnold
Last time I tried normalise-audio it couldn't handle ogg files but had to
convert them to mp3s :(
--
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.
1:30 p.m.
10-myths-about-normalization:
http://www.hometracked.com/2008/04/20/10-myths-about-normalization/
1:48 p.m.
On Sat, 25 Dec 2010 10:30:26 -0200
Fabio <capoeirista(a)arcor.de> wrote:
10-myths-about-normalization:
http://www.hometracked.com/2008/04/20/10-myths-about-normalization/
Good link
!!! SHOCK NEWS !!!
Ears were invented before computers
:P
--
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.
4890
days inactive
4891
days old
linux-audio-user@lists.linuxaudio.org
14 comments
7 participants
participants (7)
-
Arnold Krille -
Fabio -
Folderol -
fons＠kokkinizita.net -
Jörn Nettingsmeier -
Paul Davis -
Q