Equalising relative volumes in compilation
Robert J. Baker
run into the problem that although my running order seems to hang
together, because the tracks are from widely differing sources (some
soft rock, some classical, two jazz) I get alarming jumps in volume
level between some pairs of tracks. This of course happens especially
when moving from a classical track to a rock one or vice-versa, e.g.
the first two tracks.
I've read Andy McFadden's FAQ [3-41], and the software it recommends
is CoolEdit or Audiograbber. Unfortunately, Audiograbber (as the name
implies) is of no help for files one already has on hard disk, whilst
CoolEdit (when I tried to load all the files into it) insisted on
loading them all as one massive mega-WAV, 72 minutes long -- and in
the wrong order, i.e. the order in which the files are in the actual
DOS directory listing, rather than ASCII order of name (the filenames
begin "01", "02" etc.) which is what I wanted. So it doesn't look as
if CoolEdit will be of any use to me either.
So, how (preferably using GoldWave, as I'm already used to that -- I
may learn CoolEdit, but only if GoldWave definitely can't do what I
have in mind) can I equalise the volumes? I've used GoldWave to read
the average volume (RMS) for each track, and also the peak volume, but
I don't know how to calculate what value to feed into the "Adjust
Volume" dialog for each track.
* Should I base it on the average of the maximum volumes, or of the
RMS values?
* What is "RMS" anyway? (I think it may be the square root of the
arithmetic mean of the squares of the samples, but could be wrong on
this...)
* What is the best "average" of the RMS values -- the arithmetic mean,
the geometric mean or the harmonic mean? I suspect it's the
geometric, but again I'm not sure...
* If I need to adjust each track's RMS value to the "average", how do
I calculate the level of adjustment in each case?
Here endeth the essential part of this post -- what follows is the
gruesome details of the problem, should anyone wish to read them:
--------------
(Track -- Composer or Artiste(s); RMS, maximum volume)
01 Venus, Bringer of Peace (The Planets) -- Holst; 0.0290, 0.5297
02 Oh! Well (part 2) -- Fleetwood Mac; 0.0805, 0.7586
03 Aquarium (Carnival of the Animals) -- Saint-Säens; 0.0297, 0.2755
04 Nucleus -- Alan Parsons Project; 0.0829, 0.6518
05 Day After Day (The Show Must Go On) -- Alan Parsons Project;
0.0954, 0.6932
06 Going Home (Largo from "New World", excerpt) -- Dvorák; 0.0155,
0.3915
07 Albatross -- Fleetwood Mac; 0.1082, 0.6629
08 Flying -- Beatles; 0.0729, 0.6604
09 Horizons -- Genesis; 0.1486, 1.0000
10 Sweet Dreams -- Roy Buchanan; 0.2205, 0.9127
11 Blackbird -- Beatles; 0.0504, 0.5212
12 Extrapolation -- John McLaughlin; 0.1683, 1.0000
13 Total Eclipse -- Alan Parsons Project; 0.0651, 0.6522
14 Genesis 1:32 -- Alan Parsons Project; 0.1125, 0.9313
15 Kathy's Waltz -- Dave Brubeck Quartet; 0.0582, 0.7493
16 Tubular Bells part 2 (excerpt) -- Mike Oldfield; 0.0424, 0.4071
17 Dance of the Spirits of Water (The Perfect Fool) -- Holst; 0.0163,
0.1722
18 Unquiet Slumber for the Sleepers... -- Genesis; 0.0637, 0.3513
19 Neptune, the Mystic (The Planets) -- Holst; 0.0086, 0.084
Arithmetic mean of Max 0.600257895
Arithmetic mean of RMS 0.0773
Geometric mean of RMS 0.057810513
Harmonic mean of RMS 0.039095642
--------------
PS: what *does* Genesis ch. 1 v. 32 say??
PPS: Any suggestions for version 2 of the compilation? Next time
I'll use AudioGrabber instead of WinOnCD to rip the tracks...
--
------------------------------------------------
*** REALITY CHECK ***
Your universe has performed an illegal operation and will be arrested.
Evacuate it while you can...
Mike Richter
balancing the sound) in the primer at my CD-R site. There are two
different kinds of equalization needed for what you want to do - maximum
and mean values. To get both, which you will need, you must be able to
compress the dynamic range as well as to set the volume. GoldWave does
not have dynamic compression; CoolEdit does.
Mike
--
mric...@cpl.net
http://mrichter.simplenet.com
CD-R http://resource.simplenet.com
Robert J. Baker
1999 12:48:03 -0700, Mike Richter <mric...@cpl.net> would say:
>Please read the material on recording analogue (and particularly on
>balancing the sound) in the primer at my CD-R site. There are two
>different kinds of equalization needed for what you want to do - maximum
>and mean values. To get both, which you will need, you must be able to
>compress the dynamic range as well as to set the volume. GoldWave does
>not have dynamic compression; CoolEdit does.
As several of the tracks are classical I don't want to use compression
if I can avoid it. To this end, I think the way to go for me is to
"average" the RMS values; but neither Andy Fadden's site nor yours has
anything (that I can find quickly) which tells me which of the three
ways of "averaging" data is the correct one for RMS values, nor how to
calculate the proper scale factor to adjust any given RMS value to the
calculated "average" value.
Maybe I could find this information if I spent perhaps half an hour on
an in-depth perusal of the site(s), but I am in the UK and have my Net
access at work, so am not allowed to do private web-surfing because of
the phone bill...
And BTW, your site several times mentions a PDF version, but again I
couldn't find any link to download this in a few minutes of
searching...
Jim Gilliland
>
> As several of the tracks are classical I don't want to use compression
> if I can avoid it. To this end, I think the way to go for me is to
> "average" the RMS values; but neither Andy Fadden's site nor yours has
> anything (that I can find quickly) which tells me which of the three
> ways of "averaging" data is the correct one for RMS values, nor how to
> calculate the proper scale factor to adjust any given RMS value to the
> calculated "average" value.
If you're expecting to have the answer handed to you on a silver
platter, I suspect that you'll be disappointed. You'll need to learn a
bit about digital audio to accomplish your goal. And there's going to
be some trial and error involved - depending on the dynamic range and
recording level of the various pieces, you may need to apply some
judicious compression to achieve a reasonable match of levels among the
various pieces. I don't think the RMS values are going to help you with
this, but perhaps you'll formulate a new technique that works well.
Good luck with it.
> And BTW, your site several times mentions a PDF version, but again I
> couldn't find any link to download this in a few minutes of
> searching...
Huh? It's right there in the middle of the front page:
http://resource.simplenet.com/
Where it says "By request: the primer in Acrobat format", click on the
word "primer".
RWG
> > As several of the tracks are classical I don't want to use compression
> > if I can avoid it. To this end, I think the way to go for me is to
> > "average" the RMS values; but neither Andy Fadden's site nor yours has
> > anything (that I can find quickly) which tells me which of the three
> > ways of "averaging" data is the correct one for RMS values, nor how to
> > calculate the proper scale factor to adjust any given RMS value to the
> > calculated "average" value.
>
> If you're expecting to have the answer handed to you on a silver
> platter, I suspect that you'll be disappointed. You'll need to learn a
> bit about digital audio to accomplish your goal. And there's going to
> be some trial and error involved - depending on the dynamic range and
> recording level of the various pieces, you may need to apply some
> judicious compression to achieve a reasonable match of levels among the
> various pieces. I don't think the RMS values are going to help you with
> this, but perhaps you'll formulate a new technique that works well.
> Good luck with it.
IMHO for this kind of thing there is really no substititue for the human
ear. This is particularly true for something like classical music, where the
short-term average volume level varies all over the place.
Mike Richter
parameters to accomplish that end. There is no 'right' way to do it. I
seldom compile classical tracks, but will say that if you want to avoid
compression, you will have to make your own recording. Every commercial
recording is compressed to a greater or lesser extent - even classical
ones. Each engineer does it as he thinks best for the label, so you will
have to jump through hoops to get them all to sound the same.
The link to the PDF version of the primer is on the home page. There are
only about half a dozen links there - you'll find it.
Mike
"Robert J. Baker" wrote:
>
> In his wisdom, Jedi Master Eboreg Onxre foretold that on Fri, 27 Aug
> 1999 12:48:03 -0700, Mike Richter <mric...@cpl.net> would say:
>
> >Please read the material on recording analogue (and particularly on
> >balancing the sound) in the primer at my CD-R site. There are two
> >different kinds of equalization needed for what you want to do - maximum
> >and mean values. To get both, which you will need, you must be able to
> >compress the dynamic range as well as to set the volume. GoldWave does
> >not have dynamic compression; CoolEdit does.
>
> As several of the tracks are classical I don't want to use compression
> if I can avoid it. To this end, I think the way to go for me is to
> "average" the RMS values; but neither Andy Fadden's site nor yours has
> anything (that I can find quickly) which tells me which of the three
> ways of "averaging" data is the correct one for RMS values, nor how to
> calculate the proper scale factor to adjust any given RMS value to the
> calculated "average" value.
>
> Maybe I could find this information if I spent perhaps half an hour on
> an in-depth perusal of the site(s), but I am in the UK and have my Net
> access at work, so am not allowed to do private web-surfing because of
> the phone bill...
>
> And BTW, your site several times mentions a PDF version, but again I
> couldn't find any link to download this in a few minutes of
> searching...
>
> --
> ------------------------------------------------
> *** REALITY CHECK ***
> Your universe has performed an illegal operation and will be arrested.
> Evacuate it while you can...
--
dubious
>
>
> As several of the tracks are classical I don't want to use compression
> if I can avoid it.
ok, I may be missing something here, but here's my two cents -- this is
not a difficult problem (at least if I understand what you're asking).
If you just want to bring all tracks to the same peak volume, without
further compressing the classical selections, all you have to do is
_normalize_ each track individually. The pop tracks will be far more
compressed than the classical tracks, so you may find after normalizing
all tracks that the pop tracks still seem too loud. At this point just
reduce the volume of the tracks that seem too loud.
I think almost any wave editing software has a normalize function, and a
volume adjustment function where you can just specify the new volume as
a percentage of current volume -- probably no need to master the arcana
of compression algorithms for this task.
Robert J. Baker
1999 15:38:31 -0400, Jim Gilliland <usemyl...@altavista.net> would
say:
>If you're expecting to have the answer handed to you on a silver
>platter, I suspect that you'll be disappointed. You'll need to learn a
>bit about digital audio to accomplish your goal. And there's going to
>be some trial and error involved - depending on the dynamic range and
>recording level of the various pieces, you may need to apply some
>judicious compression to achieve a reasonable match of levels among the
>various pieces. I don't think the RMS values are going to help you with
>this, but perhaps you'll formulate a new technique that works well.
>Good luck with it.
Surely I'm not the only madmagician on this group? *Someone* out
there must have the answer to my "What is RMS?" question, and maybe
even the "How do I 'average' RMS values?" question, even if the
answers don't help...
>> And BTW, your site several times mentions a PDF version, but again I
>> couldn't find any link to download this in a few minutes of
>> searching...
>
>Huh? It's right there in the middle of the front page:
>
> http://resource.simplenet.com/
>
>Where it says "By request: the primer in Acrobat format", click on the
>word "primer".
Strange -- it wasn't there when I checked yesterday; maybe it got lost
on its 7,000-mile journey. I'll try checking again...
Jim Gilliland
>
> In his wisdom, Jedi Master Eboreg Onxre foretold that on Tue, 31 Aug
> 1999 15:38:31 -0400, Jim Gilliland <usemyl...@altavista.net> would
> say:
>
> >If you're expecting to have the answer handed to you on a silver
> >platter, I suspect that you'll be disappointed. You'll need to learn a
> >bit about digital audio to accomplish your goal. And there's going to
> >be some trial and error involved - depending on the dynamic range and
> >recording level of the various pieces, you may need to apply some
> >judicious compression to achieve a reasonable match of levels among the
> >various pieces. I don't think the RMS values are going to help you with
> >this, but perhaps you'll formulate a new technique that works well.
> >Good luck with it.
>
> Surely I'm not the only madmagician on this group? *Someone* out
> there must have the answer to my "What is RMS?" question, and maybe
> even the "How do I 'average' RMS values?" question, even if the
> answers don't help...
I revisited your questions, and I still think that you'll be better off
adjusting the levels by ear than by some arithmetic method. If you have
software that will allow you to manipulate the samples according to the
RMS values, that would be fine, but then you wouldn't be asking these
questions if you did, so I assume you don't <g>. Neither do I.
I don't use Goldwave, so I don't know what mean it uses for its RMS
calculation, but it does seem likely (as you surmise) that it is the
root of the arithmetic mean of the sum of the squares of the sample
values. Whether you'd be better off with a different mean is a question
that I can't answer, but I think it's academic anyway unless you have
some software that gives you that choice.
Adjusting the levels to match a particular mean is yet another problem.
There, too, I have seen no software with such a capability. And even
then, you still have a potential problem, since a routine that adjusted
samples on that basis could easily push some of them beyond the range of
a 16-bit signed number, thereby clipping your digital signal. Not good.
I suppose a clever programmer could write a routine that would calculate
the RMS value for each of a set of wav files, adjust the levels to
match, then normalize the entire batch to the largest resulting sample.
Anyone want an interesting programming project for their spare time? <g>
TomC
Mary had a little lamb, its' fleece as white as snow.
When I record my Jimi Hendrix album using MP3, it sounds as good as the .Wav
file I also recorded does, playing it back through the same set of ACS-290
speakers . Of course, the one song I'm using as a comparison is 7 minutes
long (Machine Gun), and the MP3 file is about 7mb when complete, compared to
79mb for the .wav file. (mb = Megabytes which are actually 1024bytes/per,
not a total 'mega') how's that for techno-speak?)
But then, (leading preposition with dangling participle complement), one
must still convert the MP3 to .wav to master to CD-R audio. It would confer
that this would be a 'lossy' process, since there is no more inherent data
in the MP3 file than the original 7mb, and excepting what software used for
the conversion to .wav, the supposed algorithm cannot add anymore usable
data than what is encoded in the 7mb file.
Therefore, post hoc ergo propter hoc would prove to be untrue (not a
fallacy, for those of you unread) when you surmise that:
Using MP3 to master an audio CD from vinyl, although initially keeping file
size small, would not give any benefit over recording directly to .wav file
format.
And anywhere that Mary went, the lamb was sure to go.
TomC
Jim Gilliland <usemyl...@altavista.net> wrote in message
news:37CD7FBF...@altavista.net...
Random Tox
just that.
The following is a cut and paste from the help file:-
"The Meta Normalizer
The Meta Normalizer plug-in is unique to WaveLab. It is special in that it
work in two passes: during the first pass, each file is analysed one after
the other, then, in the second pass, each file is processed one after the
other. It works by globally analyzing peak or RMS levels on all files in a
Batch. It then processes the files so that they will all play back at equal
peak or RMS levels. The highest possible level found amongst all the files
will be used as a reference. Note that the level of a file can either be
amplified or reduced. The purpose of the plugin is to apply the same and
maximum possible level/loudness in all files, without ever clipping.
This can be extremely useful for matching differing levels between files
that are being compiled for mastering to CDs for example, ensuring that
levels are equal from track to track. Depending on how much the levels
differ between files and the desired result, there are a number of options
available to ensure that the applied processing optimizes the levels without
clipping."
I've been using the program a while, but only ever used the manual when
looking for specific information. A couple of weeks ago, while looking for
something else, I came across the "Meta normalizer". The reason I'd never
stumbled upon it within the program is that it is (well) hidden in the batch
processing options, which I rarely use. I haven't had a chance to try it out
yet, but if it performs as well as it claims it could be useful for mixed
source material.
Hope this helps.
Random Tox.
Robert J. Baker
Spectrums, they would eventually reproduce the following, which was
posted on Wed, 01 Sep 1999 15:36:51 GMT by rober...@bigfoot.com
(Robert J. Baker):
[answer may not be forthcoming?]
>Surely I'm not the only madmagician on this group? *Someone* out
>there must have the answer to my "What is RMS?" question, and maybe
>even the "How do I 'average' RMS values?" question, even if the
>answers don't help...
I've now done a bit of experimenting with Microsoft Excel (a sine-wave
model consisting of one cycle broken into 3,601 samples), and it would
seem that the answers to the above two questions are:
1) The RMS value is the square root of the arithmetic mean of the
squares of the samples.
2) The correct way to "average" RMS values is to treat them as
meta-samples and calculate a meta-RMS from them.
Surely these should be in one or both of the FAQs? And I *still*
haven't a clue as to how to calculate the scale factor required to
convert a .WAV file with a certain RMS to have a given RMS... that
should be there also...
[couldn't find .PDF on the Simplenet FAQ site]
I've checked again, and this time succeeded in downloading it. Must
have been one of those hiccups that plague Net usage sometimes...
Robert J. Baker
Spectrums, they would eventually reproduce the following, which was
>[...] (mb = Megabytes which are actually 1024bytes/per,
>not a total 'mega') how's that for techno-speak?)
Actually, it's Mb, and it's actually 1,048,576 bytes... <g>
Robert J. Baker
Spectrums, they would eventually reproduce the following, which was
<no_e...@null.com>:
>Steinbergs Wavelab comes with a "Meta normalizer" plug-in that claims to do
>just that.
>
>The following is a cut and paste from the help file:-
[...and has been snipped...]
>Hope this helps.
It may, except for one thing: Where can I get Wavelab? It helps to
give a download URL if possible -- I try to always do that...
Random Tox
can find more information regarding Wavelab at http://www.steinberg.net
Regards.
Random Tox.
Robert J. Baker <rober...@bigfoot.com> wrote in message
news:37ce5f02...@news.freeserve.net...
Robert J. Baker
rober...@bigfoot.com (Robert J. Baker) spake thus about "Re:
Equalising relative volumes in compilation":
>I've now done a bit of experimenting with Microsoft Excel (a sine-wave
>model consisting of one cycle broken into 3,601 samples), and it would
>seem that the answers to the above two questions are:
>
>1) The RMS value is the square root of the arithmetic mean of the
>squares of the samples.
>
>2) The correct way to "average" RMS values is to treat them as
>meta-samples and calculate a meta-RMS from them.
>
>Surely these should be in one or both of the FAQs? And I *still*
>haven't a clue as to how to calculate the scale factor required to
>convert a .WAV file with a certain RMS to have a given RMS... that
>should be there also...
Further experimenting with my sine-wave model has revealed that
adjusting a .WAV to a given RMS value is done in exactly the same way
as adjusting it to a given peak value, i.e. calculate the scale factor
by dividing the desired level by the actual level. Multiplying the
already-calculated RMS by this scale factor gives exactly the same
value as multiplying each sample by the factor and calculating a new
RMS from the new samples. *Exactly* the same value -- I subtracted
one from the other, and the difference came up as zero every time,
regardless of what scale factor was used, and even when I replaced the
sine wave with random noise in the range -1 to 1...
I think this information ought to be in the FAQs as well...
Since the scale factor affects RMS and peak values equally, I was able
to calculate a scale factor and see whether it causes clipping. The
meta-RMS does so for three of the tracks, but judicious tweaking has
given me a value which adjusts all the tracks to the same RMS
*without* clipping any of them (highest peak is 0.94)...
Of course, whether the new balance *sounds* right remains to be heard,
but at least it's a start...
P.S. Any suggestions for Relax Vol. 2?
Terry King
multiple cuts....
--
Terry King ...In The Woods In Vermont
Robert J. Baker
rober...@bigfoot.com (Robert J. Baker) spake thus about "Re:
Equalising relative volumes in compilation":
>Since the scale factor affects RMS and peak values equally, I was able
>to calculate a scale factor and see whether it causes clipping. The
>meta-RMS does so for three of the tracks, but judicious tweaking has
>given me a value which adjusts all the tracks to the same RMS
>*without* clipping any of them (highest peak is 0.94)...
>
>Of course, whether the new balance *sounds* right remains to be heard,
>but at least it's a start...
I've just tried running the collection from beginning to end, and the
new balance sounds superb! No compression, just volume scaling.
All that remains now is to burn it to CD, this afternoon...
>P.S. Any suggestions for Relax Vol. 2?
I think the closing track should be Genesis' "Afterglow", and that "Us
and Them" (Pink Floyd playing jazz) and something from the "Pastoral"
should also be there. Anyone got any other ideas?
Tim Kroesen
Stevie Wonder's Songs ITKOL "Easy goin Evening - Mamma's Call"
Eric Essix from Small Talk "Hang Time"
Clannad's newest Grammy winner Landmarks "Loch Na Cailli"
...Guess I should be making this CD...
; - ) Tim K
Robert J. Baker <rober...@bigfoot.com> wrote in message
news:37d7a135...@news.freeserve.net...