Replaygain question: using it in EAC andalready ripped FLAC's

[Mark Lanctot]

--- EnochLight
<EnochLight.22fn8b (AT) no-mx (DOT) forums.slimdevices.com>
wrote:

> 1. I've already ripped 100 CD's into FLAC. What is
> the easiest way to
> apply "Replaygain" information to the Vorbistag's on
> these particular
> FLAC's?

I believe you said in another thread you didn't like
foobar2000. Still, it would be the fastest and
easiest way to apply it to all those files at once.

The metaflac command-line metatag editor would only be
suitable for one file at a time.

>
> 2. Going forward, I would like to change my EAC
> options so that the
> "Replaygain" data is inserted into the tags so the
> SB3 reads it. Below
> is what I currently have entered for my tags in the
> "Additional
> Commandline Options":
>
> -5 -V -T "artist=%a" -T "title=%t" -T "album=%g" -T
> "date=%y" -T
> "tracknumber=%n" -T "genre=%m" %s

Add:

--replay-gain

after the -V and before the -T.

See

http://flac.sourceforge.net/document...coding_options

[Josh Coalson]

the reference for FLAC replaygain is 89dB. recent encoders store
the reference level in a tag REPLAYGAIN_REFERENCE_LOUDNESS

--- Club1820 <Club1820.2t8djz1183609801 (AT) no-mx (DOT) forums.slimdevices.com>
wrote:

>
> Does anyone know to what db level is set when applying replay gain
> through EAC & FLAC External Compression?
>
> I had previously set my Mp3s to replay gain of 92db. Some of my FLAC
> encoded files play louder than this. Is the Replay Gain by default
> set
> to 89db? Is there a way to choose the db level?
>
> Thanks.
>
>
> --
> Club1820
>
------------------------------------------------------------------------
> Club1820's Profile:
> http://forums.slimdevices.com/member.php?userid=10999
> View this thread:
> http://forums.slimdevices.com/showthread.php?t=20575
>
>

[Timothy Stockman]

the reference for FLAC replaygain is 89dB. recent encoders store
the reference level in a tag REPLAYGAIN_REFERENCE_LOUDNESS

I've never been quite sure why the replaygain people chose to express their target gain as SPL, since they do not directly control the SPL of the replay, but only influence it by their adjaustment. Somewhere, buried in the spec. is the REAL target that replaygain is trying to hit, -20 dBFS (as I remember). As I understand it, a recording that the replaygain algorithm scores at an "average" level of -20 dBFS will have a replaygain of 0.0 dB. Whether the resultant playback actually corresponds to 89 dB "average" SPL depends on the amplifier gain, speaker sensitivity, distance to listener (at higher frequencies), volume of the room (at lower frequencies), frequency response of the reproducing system, etc.

Most often, replaygain computes a negative value. This is because -20 dBFS is a rather generous amount of headroom, and many popular recordings can be mastered at a somewhat higher level without clipping even a single peak. I do have a few popular/jazz CDs which generate positive replaygain values, but in general, at no point do their peaks approach 0 dBFS. On one CD, the highest peak is around -6 dBFS.

Actually, hasn't the movie industry tied an 83 dB SPL monitor level (not 89 dB SPL) to -20 dBFS on the digital recording medium? The movie industry wants a little more headroom so their explosions can have more impact. I believe the suggested SPL for radio/popular music is 89 dB SPL tied to -14 dBFS, so they just raised everything 6 dB, apparently feeling that 14 dB of headroom was sufficient for peaks. While movie studios seem reasonably good about following the 83/-20 guideline, radio stations and recording companies for the most part ignore the 89/-14 guideline, hence the need for something like replaygain in the first place.

dBFS = dB referenced to Full Scale, i.e. that largest value that can be expressed by digital audio. So, by definition, all dBFS values are 0 or negative.
db SPL = dB Sound Pressure Level, IIRC is referenced to the average human threshold of hearing. So by definition, all sounds that an average human can hear will have positive SPL values.

http://www.jimprice.com/prosound/db.htm

This guy gets it mostly right, except that digital audio signals are almost always SIGNED, so the highest possible value is not 1111 1111 1111 1111 (the twos complement representation of -1), which is actually the LOWEST possible signal above 0, along with +1. The highest possible signals are 0111 1111 1111 1111 (+32767) and 1000 0000 0000 0000 (-32768). I believe replaygain, as part of its algorithm, sqaures each sample, so the large negative values count equally to large positive sample values.

[Timothy Stockman]

BTW Mark and/or EnochLight et al - I wrote a quick-and-dirty Windows program to generate a BAT file of metaflac replaygain commands for one or more "albums" worth of material. It will navigate a hierachy of directories, assuming all files in a given directory are to be analyzed as a single album. If all files already have replaygain tags and the albumreplaygain tags have the same dB value, it assumes it doesn't have to generate a metaflac command for that directory.

The only problem it currently has is 88.2 and 96 KHz files. It will generate the command, which will fail because replaygain doesn't support these rates. Don't know whether to generate another set of resampled files that replaygain supports and copy the tags. Or I could just wait to see if the replaygain guy eventually supports higher sample rates...

Let me know if you want me to email you with the C source or the EXE of my program.