replaygain (possible) annoyance, need confirmation, please Options

[guruboolez]

I like very much replaygain, but I suspect some problems with some recordings, especially solo harpsichord. For some tracks, replaygained harpsichord music seems to sound louder than other replaygained tracks. Form taste at least, heavy metal tracks are quiter than a Bach Prelude after reaplyagain calculation.

I've a short sample suit (~10 seconds, mixing three different harpsichord recordings with three different metal/rock samples), encoded with mpc 5, and replaygained.
I'd like to read the experience of some members here [with track gain enabled of course]: do you find the 6 samples equally loud, or not ?

I know that RG calculations are not perfect - but for the moment, I've no tool for batch correction of replaygained values (maybe an idea for RG scanner in foobar2000). And 2Bdecided want to modify RG structure (see here. Maybe should I request another RG field (MANUAL GAIN CORRECTION; INDIVIDUAL TASTE CORRECTION... I don't know).

Thanks

[Linkin]

I like very much replaygain, but I suspect some problems with some recordings, especially solo harpsichord. For some tracks, replaygained harpsichord music seems to sound louder than other replaygained tracks. Form taste at least, heavy metal tracks are quiter than a Bach Prelude after reaplyagain calculation.

i am not a listening expert and my ears are far from perfect but from your samples i can only share your experience. i also think that the metal tracks are quiter than the others.

[guruboolez]

Thank you very much
I don't need expert, audiophile, batman or similar mythological beast. Simple, immediate (even fragile or unsure) judgments are just what I request. It would be nice to hear others members too, sharing or not our impressions.
Thanks again.

This post has been edited by guruboolez: Nov 22 2003, 16:02

[evereux]

I found the difference quite substantial. As big a difference in the two types without replaygain turned on.

[[proxima]]

Form taste at least, heavy metal tracks are quiter than a Bach Prelude after reaplyagain calculation.

I have your same impression, the three classical samples are louder.
Sorry, i can't provide a personal contribute with samples because i use RG only to keep peak levels and use clipping prevention with winamp plugin, with Foobar2000 and the advanced limiter this is even useless and i use RG less and less.

And 2Bdecided want to modify RG structure

What about the proposal of changing 83dB to 89dB ?? I read that 2Bdecided and Gabriel both agrees to raise the 83db old standard....why ? I have seen (but this is quite obvious) that with classical samples or high dynamic tracks the correction is positive and sometimes the track peak is near the clipping level. I'm worried because this proposed 6dB boost can make some tracks clipping because of excessive positive correction. I've read that the 83 dB value represent a compromise between an accettable loudness and the clipping prevention (in case of positive correction). I know that the replaygain standard includes a limiter by design but someone might prefer to not reduce the dinamic range to obtain a equal perceived loudness.
You've a large classical collection.... have you ever found some samples that clip with track gain ? I heard about these (isolated) cases with the actual 83 dB standard... maybe these cases can grow in number with the adoption of 89 dB.....

IMHO mainaining the actual standard is better, the user has always the possibility to use a preamp (or K-meters in case of MPC implementation) to make the track louder than 83 dB. Maybe someone more competent can give me some clarifications: with my current state of knowledge i'm convinced that raising the standard of 83 dB is not a good idea

Maybe should I request another RG field (MANUAL GAIN CORRECTION; INDIVIDUAL TASTE CORRECTION... I don't know).

Another field for storing MANUAL GAIN CORRECTION seems not to be a bad idea

[guruboolez]

,Nov 22 2003, 05:36 PM]You've a large classical collection.... have you ever found some samples that clip with track gain ? I heard about these (isolated) cases with the actual 83 dB standard... maybe these cases can grow in number with the adoption of 89 dB.....

First, thanks for reply (evereux & [proxima]).

Yes, with classical, track gain clipping may appear more often. I've wrote (in french) an article about replaygain & fb2k advanced limiter, based on a "killer" track :

http://forum.hardware.fr/forum2.php3?post=51453&cat=3



Original track dynamic:



Album Gain dynamic [+ 4.51dB]:



Track Gain dynamic [+13.11dB]:




Zoom on clipped part:

http://membres.lycos.fr/guruboolez/AUDIO/l...ectre_title.png
http://membres.lycos.fr/guruboolez/AUDIO/l...ectre_title.jpg

Such files are rare of course, but clipping will appear more often with 89 dB basis than with 83 dB.

[2Bdecided]

Re: 83vs89dB

This is irrelevant when it comes to discrepancies between samples.

As for your comments about 89dB causing clipping with classical samples - yes guruboolez, I know - that's why I set it at 83dB originally. What can I say? You just have to use the appropriate slider in each player/plug-in.

(The original ReplayGain proposal suggests letting the player set this, or permanently reduce this, as appropriate.)


Re: Loudness discrepancies.

On these tracks, the Harpsichord is ReplayGained to be slightly louder than the metal tracks. I totally disagree with evereux though: the volume difference without ReplayGain is much greater.

Hopefully a future version of ReplayGain (specifically: a new calculation) can solve this problem. I have tracks with bigger discrepancies!

There will remain the problem that you expect the Harpsichord to be quiet, and you expect the rock band to be loud, and ReplayGain will (should) match them. The solutions...


Re: manual or individual taste correction

I proposed both a user value (to override the others), and a "Real" value, which could be set by a mastering engineer. If an improved ReplayGain calculation didn't solve the problem, then you could use either of these. Hopefully future players would make the "user" value easy to set. One day, you might find the "Real" value printed on the backs of CDs. Well, I can dream! It's in the DVD-A MLP metadata, but I don't know if it's set correctly.

Cheers,
David.
P.S. relevant replies to this thread please:
http://www.hydrogenaudio.org/forums/index....topic=15445&hl=