Does Replay gain work differtly in Media monkey, Foobar and Media Monkey given 2 differnt Results Options

[jimmanningjr]

Hey Anyone,
For the past 2 years I have been using Media Monkey to Organize and play my music collection which is now about 42,000 tracks about 75% of my collection has Replay gain Values....I am making the switch over to Foobar2000 and was about to replay gain the other 25% with it (Foobar) when I discovered that for the same album I was getting 2 different values...and it is making me wonder does Media Monkey use a different algorithm or is there something I am overlooking...both are set to tag by ALBUM with the track value put in there also. Here are my results:

Miles Davis /Jack Johnson Soundtrack/
Trk1 Right Off -FOOBAR--Album Gain -1.97db Trk Gain -2.49db
Trk 2 Yesternow-FOOBAR Album Gain -1.97db Trk Gain -0.87db

Trk 1 Right Off Media monkey--Album Vol=-7.9db Trk Vol= -8.4db
Trk 2 Yesternow Media Monkey--Album Vol= -7.9db Trk Vol= - 6.8db

Could someone shed some light on this for me.
Thank You

[2Bdecided]

I've downloaded Media Monkey for the first time ever this morning, and tried to figure out what it's doing.

Firstly, AFAICT, I can't make it write a REPLAYGAIN_REFERENCE_LEVEL tag. Has anyone actually seen this program generate these tags? If so, how/when/why?

Secondly, it doesn't call its implementation "ReplayGain" or use any of the conventional names for ReplayGain-related funcationality. However, it clearly credits Glen Sawyer for mp3gain and myself for ReplayGain. The documentation states that MM's volume levelling is ReplayGain...

MediaMonkey supports Replay Gain analysis and playback (via 'Analyze Volume' and 'Level Playback Volume').

from http://www.mediamonkey.com/support/index.p...ownloaditemid=2

In MM, the right-click context menu includes "Analyze volume", which seems to perform a standard ReplayGain scan and store the result in standard tags. (Note: there are settings elsewhere that will affect the whether the Album value is calculated)

The right-click context menu also includes "Level Track Volume", which seems to perform a standard mp3gain-style change to the audio data itself.

In MM, if you go to Tools, Options, Player, Volume Levelling, you will find the following settings:

Target volume level for Playback: This is a pre-amp, but instead of 0dB for "no change from ReplayGain default", it says explicitly 89dB (which is the ReplayGain default). If you change it to 90dB, that's the same as +1dB in fb2k. If you change it to 87dB, that's the same as -1dB in fb2k. etc. Just add or subtract 89dB to transform to/from the more common choice of number. The same is true for all the values discussed below.

checkbox: Level playback volume: enable or disable ReplayGain during playback
per Track / per Album: use Track or Album ReplayGain values

Target volume for leveling(sic) tracks: The target volume when using the mp3gain-style function to change the audio data itself.

Target Volume Level for Ripping Audio CDs: This is the level used if you ask the ripper to volume level the tracks. (the tooltip is completely wrong!)
Level Ripped Tracks: per track / per album: use Track or Album ReplayGain values

checkbox: Clipping Prevention: This enables clipping prevention during playback, and when ripping CDs with volume levelling, but not when applying a gain change to a file AFAICT.

There's one issue that could cause unexpected problems - the mp3 encoding and decoding pipelines don't allow for values greater than 0dB. This means...
1. If you choose settings that clip the audio when ripping with volume levelling enabled, that's it - the clipping is hard coded into audio data of the mp3s. You can't remove it later by lowering the level. You'll have to re-rip from the original CD.
(If you enable clipping prevention, then tracks that would have been clipped will just be peak normalised instead)
2. If there are sample values over 0dB "within" the mp3 file (e.g. due to normal lossy encoding of loud CDs, or CDs that you make loud by enabling volume levelling, or due to using mp3gain-like procesisng to apply a positive gain change to an mp3 file), these will get clipped during playback. It doesn't matter if you turn clipping prevention on. It doesn't matter if the ReplayGain values cause the files to be turned down anyway. The audio will still get clipped - it'll then get turned down after it's been clipped!
3. The ReplayGain_track_peak and _album_peak values written by MM for mp3 files will be incorrect if the audio data "within" the mp3 peaks above 0dB FS. It'll stored any values over 1.0 as 1.0, so no software will know these files are clipped.

If you rip with volume levelling enabled, it'll write a ReplayGain_track_gain tag (only - no peak tag). If Clipping Prevention was enabled when ripping, this tag could be completely wrong - MM seems to store a value that might have been right if clipping prevention hadn't been enabled. MM seems to ignore these tags, but some other players read them, giving unwanted results (e.g. desperately quiet audio).


So, lots of things to nit-pick over, but in terms of its implementation of a pre-amp and the possibility that it may use non-standard tags, MM is actually pretty standard, even if on the surface it uses different names for everything.

Cheers,
David.

This post has been edited by 2Bdecided: Sep 6 2010, 11:07

[chrisjj]

MM is actually pretty standard

Surely the fact it defaults to 89dB reference makes it non-compliant with the ReplayGain standard, specifying 83dB.

[2Bdecided]

MM is actually pretty standard

Surely the fact it defaults to 89dB reference makes it non-compliant with the ReplayGain standard, specifying 83dB.

No, the ReplayGain website is out of date - or more accurately it's just an archive of the initial proposal. It hasn't been updated since 2001. However...

Almost everyone is using a reference level of 89dB, rather than the 83dB in the original ReplayGain proposal. Unless there are any objections, I'll change the official reference level to 89dB.

from this 7 year old thread: http://www.hydrogenaudio.org/forums/index....st&p=154605

If you want complete and total accuracy, the ReplayGain tags stored in the files are 6dB above the gain adjustments required to make the files "sound as loud" as a -20dB RMS pink noise signal when replayed in an SMPTE RP 200 calibrated system. The -20dB RMS pink noise signal will measure 83dB SPL at the listener's seat in such a system. The extra 6dB included in the stored ReplayGain values has been standard for over 7 years - just because I never updated the RG website, please don't suggest it's more "correct" to write tags referenced to 83dB - that would be wrong.

I was going to point you to the HA wiki as being more up to date, but I see you've edited it...!

Cheers,
David.