IPB

Welcome Guest ( Log In | Register )

ReplayGain2, ReplayGain2 proposal
jukkap
post Jul 23 2011, 16:33
Post #1





Group: Members
Posts: 26
Joined: 14-July 11
Member No.: 92296



I suggest we discuss about possibility of creating ReplayGain2 specification that would replace the current ReplayGain.

ReplayGain2 would be playback compatible with the current ReplayGain (all current ReplayGain compatible players would work), but it should use ITU-R BS.1770/EBU R128 to calculate gain levels.

Current ReplayGain algorithm is odd and difficult and not properly specified. It also does not work for multichannel audio. ITU-R BS.1770/EBU R128 would overcome these issues.







Go to the top of the page
+Quote Post
 
Start new topic
Replies
Pat_
post Aug 6 2013, 11:01
Post #2





Group: Members
Posts: 10
Joined: 2-August 13
Member No.: 109428



QUOTE (2Bdecided @ Aug 5 2013, 11:35) *
It is fine to use a new calculation method with a new reference loudness, but the values must be converted to the original reference loudness before being written to ReplayGain tags. That way, loudness matching will work on all ReplayGain compatible players across all ReplayGain tagged files.


Hi. I agree.
I watched this short clarification:
Florian Camerer (2012) ITU BS.1770-2 Gating Scheme, http://www.youtube.com/watch?v=Y-zhAEu5nVo

I believe much of the confusion is because loudness metering and loudness normalization scheme are confounded. These are two separate dimensions but are NOT presented in an orthogonal way. A system should clearly indicate which method was used for measuring loudness (loudness metering) and which gain adjustment scheme was used for loudness normalization (leveling).

After a forum exchange with Peter Belkner (author of R128GAIN) and reading the standards, this is my -tentative- understanding of loudness metering and normalization.

(1) LOUDNESS METERING
(1.1) PROGRAMME LOUDNESS = single value to indicate integrated loudness for duration of the programme (track / album)

Two options:
(1.1.1) Classic Replaygain algorithm (2001).
(1.2.1) ITU BS. 1770 algorithm. Including gating function when signal drops below a certain point, specified as of revision ITU-R BS.1770-2 (03/2011). Latest: ITU-R BS.1770-3 (08/2012).

(1.2) LOUDNESS RANGE
(1.2.1) EBU-128 specified as difference between low and high percentiles of distribution of loudness as measured with ITU BS. 1770 algorithm including gating

(2) LOUDNESS NORMALIZATION / LEVELING
Loudness leveling of programmes/music compliant to:

(2.1) EBU-R128 norm, using -23 LUFS as target level.
(2.2) Classic ReplayGain 1.0 spec, reference -89 dBFS. Only applicable if (1.1.1) was used.
(2.3) ReplayGain 2.0 spec ("RG2"), mapping Loudness (L) metered with (1.2.1) onto dB's using empirical formula: RG2gain = -18 - L
(2.4) ATSC-A/85, using -24 LUFS

Is that how you see it also?

This post has been edited by Pat_: Aug 6 2013, 11:09
Go to the top of the page
+Quote Post
2Bdecided
post Aug 6 2013, 15:40
Post #3


ReplayGain developer


Group: Developer
Posts: 5177
Joined: 5-November 01
From: Yorkshire, UK
Member No.: 409



QUOTE (Pat_ @ Aug 6 2013, 11:01) *
I believe much of the confusion is because loudness metering and loudness normalization scheme are confounded. These are two separate dimensions but are NOT presented in an orthogonal way. A system should clearly indicate which method was used for measuring loudness (loudness metering) and which gain adjustment scheme was used for loudness normalization (leveling).
Yes. Both ReplayGain and EBU R128 do both very clearly though.

A choice must be made in tagging: do you store the measured loudness, or the gain change required to hit a reference loudness.

Initially the first option seems simple and attractive. However it's not, for two reasons...
1) Imagine you could create a perfect algorithm that perfectly told you how loud something sounded to a human listener. You might then assume that, to reach a given reference level, you must change the volume by x dB (x being the difference between the measured loudness and the reference level). You might assume that, but you would not be quite right. If you raise the volume by 10dB, the loudness might increase by 11dB. It sounds nonsense, but look at the equal loudness curves - as things get louder, the bass and treble gets disproportionately more audible compared to the midrange. Therefore, if you want to loudness match with complete accuracy, it's no good storing how loud something is - it can only be completely accurate if you store how much gain is required to meet a given reference level.
2) Far more simply, the player just wants to know what gain change to apply, so you might as well just store that.

Hence ReplayGain, not ReplayLevel (which is was called for about a week at the very start!)


Cheers,
David.
Go to the top of the page
+Quote Post

Posts in this topic
- jukkap   ReplayGain2   Jul 23 2011, 16:33
- - mobyduck   http://www.hydrogenaudio.org/forums/index....showt...   Jul 23 2011, 16:40
- - Notat   This has been discussed. I believe there's act...   Jul 23 2011, 17:36
- - C.R.Helmrich   QUOTE (jukkap @ Jul 23 2011, 17:33) Repla...   Jul 23 2011, 18:10
- - jukkap   So that is the way to go. I suggest someone will c...   Jul 23 2011, 18:40
- - nazgulord   If I may ask, how different are gain values betwee...   Jul 23 2011, 23:16
- - db1989   R128 versus ReplayGain   Jul 23 2011, 23:31
- - Notat   I've been contacted by jaynyy and we've di...   Oct 8 2012, 22:55
- - Pat_   Suggestion for ReplayGain specification 2.0 ... I...   Aug 2 2013, 15:17
- - 2Bdecided   Writing ReplayGain track and album gain tags with ...   Aug 5 2013, 11:35
- - Pat_   QUOTE (2Bdecided @ Aug 5 2013, 11:35) It ...   Aug 6 2013, 11:01
|- - 2Bdecided   QUOTE (Pat_ @ Aug 6 2013, 11:01) I believ...   Aug 6 2013, 15:40
- - Case   Pbelkner's decision to write incompatible data...   Aug 6 2013, 13:42
|- - 2Bdecided   QUOTE (Case @ Aug 6 2013, 13:42) Pbelkner...   Aug 6 2013, 15:52
|- - pbelkner   QUOTE (Case @ Aug 6 2013, 14:42) Pbelkner...   Aug 6 2013, 17:17
- - [JAZ]   @pbelkner: When you designed and implemented your ...   Aug 6 2013, 20:09
- - Pat_   Hi Guys, It was not my intention to stir things u...   Aug 7 2013, 11:13
|- - 2Bdecided   QUOTE (Pat_ @ Aug 7 2013, 11:13) I believ...   Aug 8 2013, 11:31
|- - [JAZ]   QUOTE (2Bdecided @ Aug 8 2013, 12:31) No ...   Aug 8 2013, 15:48
- - Pat_   QUOTE (2Bdecided @ Aug 8 2013, 11:31) If ...   Aug 8 2013, 14:59
- - greynol   The reference level is calibrated to pink noise wi...   Aug 8 2013, 15:23
- - Pat_   QUOTE (greynol @ Aug 8 2013, 15:23) The r...   Aug 8 2013, 15:38
- - Pat_   QUOTE ([JAZ] @ Aug 8 2013, 15:48)...   Aug 9 2013, 14:47
- - 2Bdecided   Just to be 100% clear: you can use ReplayGain tags...   Aug 9 2013, 17:15


Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 24th October 2014 - 14:16