IPB

Welcome Guest ( Log In | Register )

23 Pages V  « < 5 6 7 8 9 > »   
Closed TopicStart new topic
R128GAIN: An EBU R128 compliant loudness scanner
benski
post Jan 26 2011, 18:30
Post #151


Winamp Developer


Group: Developer
Posts: 670
Joined: 17-July 05
From: Brooklyn, NY
Member No.: 23375



QUOTE (spies @ Jan 26 2011, 12:14) *
QUOTE (benski @ Jan 26 2011, 07:32) *
This was done on a 45,000 track test library.

QUOTE (lvqcl @ Jan 26 2011, 08:20) *
My small (~2000 tracks only) comparison of foo_r128scan vs. foo_rgscan:

I assume you both did the comparison in track mode and I am curious if the results are the same in album mode.

Duh, I see now looking at lvqcl graphic is that it contains both track and album mode so I assume the result is the averaging of the two. I am still curious what the results are when just comparing track or album mode.


I compared only track mode.
Go to the top of the page
+Quote Post
lvqcl
post Jan 26 2011, 19:21
Post #152





Group: Developer
Posts: 3468
Joined: 2-December 07
Member No.: 49183



Track mode (2124 tracks): foo_R128 = foo_RG - 0.62
Album mode (175 albums): foo_R128 = 0.93*foo_RG - 1.01
Go to the top of the page
+Quote Post
C.R.Helmrich
post Jan 27 2011, 00:15
Post #153





Group: Developer
Posts: 694
Joined: 6-December 08
From: Erlangen Germany
Member No.: 64012



QUOTE (lvqcl @ Jan 26 2011, 18:20) *
My small (~2000 tracks only) comparison of foo_r128scan vs. foo_rgscan:

http://img816.imageshack.us/img816/5396/rgr128.png

Nice plot! My first question would be: could you give us the names of the songs on which the two algorithms strongly diverge, esp. the ones at [R128=-2, RG=6], [R128=-6ish, RG=1ish], [R128=1.5, RG=-2.5]?

The same question of course also goes to benski, if he has such data!

Edit: Guess I should have read this first.

Chris

This post has been edited by C.R.Helmrich: Jan 27 2011, 00:22


--------------------
If I don't reply to your reply, it means I agree with you.
Go to the top of the page
+Quote Post
lvqcl
post Jan 27 2011, 00:25
Post #154





Group: Developer
Posts: 3468
Joined: 2-December 07
Member No.: 49183



Yes, that thread.

I also uploaded 3 short (~20 sec) samples here: http://www.hydrogenaudio.org/forums/index....showtopic=86429

This post has been edited by lvqcl: Jan 27 2011, 00:27
Go to the top of the page
+Quote Post
benski
post Jan 27 2011, 00:39
Post #155


Winamp Developer


Group: Developer
Posts: 670
Joined: 17-July 05
From: Brooklyn, NY
Member No.: 23375



I got snowed in today and don't have access to the data at the lab at the office. I'll find some outlier tracks tomorrow (there were some major ones!) and post details.
Go to the top of the page
+Quote Post
jangk
post Jan 28 2011, 20:20
Post #156





Group: Members
Posts: 36
Joined: 31-December 10
Member No.: 86953



This might be of interest:

Grimm Audio has published an EBU R128 compliant Loudness scanner and Loudness Normalizer called "Level One".

There is a Demo ready for download in the web shop.
Here is the link to the manual: Manual Download

Jean
Go to the top of the page
+Quote Post
spies
post Jan 28 2011, 21:19
Post #157





Group: Members (Donating)
Posts: 73
Joined: 20-July 02
From: Foster City, CA
Member No.: 2685



QUOTE (jangk @ Jan 28 2011, 11:20) *
url="http://www.grimmaudio.com/index.html"]Grimm Audio[/url] has published an EBU R128 compliant Loudness scanner and Loudness Normalizer called "Level One".

Looks nice and perhaps I will download the demo but 450.00 excl. VAT! ohmy.gif
Go to the top of the page
+Quote Post
punkrockdude
post Jan 28 2011, 21:32
Post #158





Group: Members
Posts: 267
Joined: 21-February 05
Member No.: 20022



Hi guys. Is it the oversampling which makes the scanning procedur really slow and is the volume supposed to be quite alot quieter than replay gain? Regards.
Go to the top of the page
+Quote Post
pbelkner
post Jan 28 2011, 22:34
Post #159





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (punkrockdude @ Jan 28 2011, 21:32) *
Is it the oversampling which makes the scanning procedur really slow

Yes. You may switch it off by disabling "True Peak".

The next version will have a bit faster "True Peak" mode.

QUOTE (punkrockdude @ Jan 28 2011, 21:32) *
is the volume supposed to be quite alot quieter than replay gain? Regards.

It is about 5 dB quieter compared to ReplayGain. You may switch on ReplayGain compatible mode, it makes it about 5 dB louder.
Go to the top of the page
+Quote Post
bryant
post Jan 28 2011, 22:58
Post #160


WavPack Developer


Group: Developer (Donating)
Posts: 1297
Joined: 3-January 02
From: San Francisco CA
Member No.: 900



QUOTE (pbelkner @ Jan 28 2011, 13:34) *
QUOTE (punkrockdude @ Jan 28 2011, 21:32) *
is the volume supposed to be quite alot quieter than replay gain? Regards.

It is about 5 dB quieter compared to ReplayGain. You may switch on ReplayGain compatible mode, it makes it about 5 dB louder.

This is the confusion I was referring to earlier.

Have you considered the suggestion of using different tag names when writing incompatible gain information?

Regards,
David
Go to the top of the page
+Quote Post
pbelkner
post Jan 28 2011, 23:30
Post #161





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (bryant @ Jan 28 2011, 22:58) *
This is the confusion I was referring to earlier.

As it seems there is a lot of confusion out there:
  • If I go to the EBU R128 page the first thing I see is -23 LUFS. Written in boldface and highlighted by blue color. If you like to blame someone go there.
  • If I see something labeled EBU R128 conformant than that's exactly what I expect.
  • What seems to be completely mixed-up in this forum is the difference between BS.1770 and EBU R128. EBU R128 is essentially -23 LUFS, about 5 dB below RG.
QUOTE (bryant @ Jan 28 2011, 22:58) *
Have you considered the suggestion of using different tag names when writing incompatible gain information?

Why should I? The ones who want ReplayGain's about -18 LUFS & BS.1770 can easily customize the tool accordingly. The next version will offer even more options.

This tool is about EBU R128 in it's entirety, not only about BS.1770.

EDIT: The following are some out of the reasons why I don't like to have any ad-hoc assumptions regarding the correlation between R128 and RG in the tool:

QUOTE (benski @ Jan 26 2011, 16:32) *
More specifically, regression analysis indicated that the gain adjustment should actually be -17.5 - 1.05*R128 (e.g. an R128 value of -10LU would correspond to -8.55dB gain adjustment)

Especially this one:

QUOTE (benski @ Jan 26 2011, 16:32) *
but I don't like the idea of having the beta coefficient not be 1.0.


This post has been edited by pbelkner: Jan 28 2011, 23:48
Go to the top of the page
+Quote Post
punkrockdude
post Jan 29 2011, 00:12
Post #162





Group: Members
Posts: 267
Joined: 21-February 05
Member No.: 20022



Scanned lossless files with replaygain showed 1.00000 peaks at highest now shows over 1.00000. Does this have something to do with the True Peak option and is it checking if the file clips when being oversampled? Regards
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 00:18
Post #163





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (punkrockdude @ Jan 29 2011, 00:12) *
Scanned lossless files with replaygain showed 1.00000 peaks at highest now shows over 1.00000. Does this have something to do with the True Peak option and is it checking if the file clips when being oversampled? Regards

Yep. That's the whole reason for "true" peak. Exactly the same you can expect to happen during playback, i.e. even if the record itself technically doesn't clip, during playback it may happen. I observed clipping according to "true peak" up to 1 dB for brickwall limited records (not restricted to them).

This post has been edited by pbelkner: Jan 29 2011, 00:19
Go to the top of the page
+Quote Post
jangk
post Jan 29 2011, 10:41
Post #164





Group: Members
Posts: 36
Joined: 31-December 10
Member No.: 86953



QUOTE
Does this have something to do with the True Peak option and is it checking if the file clips when being oversampled?


Yes exactly, please google for "Intersample Peaks".

Recommended comprehensive reading: 0dBFS+ Levels in Digital Mastering

Jean
Go to the top of the page
+Quote Post
Notat
post Jan 29 2011, 18:07
Post #165





Group: Members
Posts: 581
Joined: 17-August 09
Member No.: 72373



QUOTE (pbelkner @ Jan 28 2011, 15:30) *
As it seems there is a lot of confusion out there:
  • If I go to the EBU R128 page the first thing I see is -23 LUFS. Written in boldface and highlighted by blue color. If you like to blame someone go there.
  • If I see something labeled EBU R128 conformant than that's exactly what I expect.
  • What seems to be completely mixed-up in this forum is the difference between BS.1770 and EBU R128. EBU R128 is essentially -23 LUFS, about 5 dB below RG.
QUOTE (bryant @ Jan 28 2011, 22:58) *
Have you considered the suggestion of using different tag names when writing incompatible gain information?

Why should I? The ones who want ReplayGain's about -18 LUFS & BS.1770 can easily customize the tool accordingly. The next version will offer even more options.

If you want to do loudness normalization in strict accordance with R128 specifications you should create and use R128 tags. You should not be reusing ReplayGain tags. You will do as you will, I suppose; there is no tag police to enforce this and documentation for ReplayGain tags is just now coming into existence.
Go to the top of the page
+Quote Post
Notat
post Jan 29 2011, 18:15
Post #166





Group: Members
Posts: 581
Joined: 17-August 09
Member No.: 72373



QUOTE (punkrockdude @ Jan 28 2011, 16:12) *
Scanned lossless files with replaygain showed 1.00000 peaks at highest now shows over 1.00000. Does this have something to do with the True Peak option and is it checking if the file clips when being oversampled? Regards

Same goes for the peak measurement. ReplayGain specifies storing sample peak. Storing true peak in ReplayGain tags is not in accordance with the proposal. Admittedly, the interoperability problems created by such a non-standard implementation are fairly minor and unobtrusive.
Go to the top of the page
+Quote Post
[JAZ]
post Jan 29 2011, 18:15
Post #167





Group: Members
Posts: 1798
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



QUOTE (pbelkner @ Jan 28 2011, 23:30) *
QUOTE (bryant @ Jan 28 2011, 22:58) *
Have you considered the suggestion of using different tag names when writing incompatible gain information?

Why should I? The ones who want ReplayGain's about -18 LUFS & BS.1770 can easily customize the tool accordingly. The next version will offer even more options.


If i get a file with ReplayGain information, as in the file has the tags used by a Replaygain-compatible scanner to indicate gain i would expect that tool to conform to the replaygain standard meaning for those tags. It is not important what is used to calculate the tags, or the amount of difference from the reference implementation. What is important is that the tags mean exactly what Replaygain says it means.


If you write replaygain tags, you ought to write them with the meaning that Replaygain applies.
If you want to conform to R128, the tags need to be different, and those can be in conformance with what R128 says.

So you got the option wrong. Specifying replaygain in your tool should specify "write the scanning data as replaygain-compatible tags", while r128 should mean "write the scanning data as a new tag format which is not the replaygain tags".


Else, you're just making a double standard, and this is what you are being warned about repetitively in this thread.
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 18:50
Post #168





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (Notat @ Jan 29 2011, 18:07) *
If you want to do loudness normalization in strict accordance with R128 specifications you should create and use R128 tags. You should not be reusing ReplayGain tags. You will do as you will, I suppose; there is no tag police to enforce this and documentation for ReplayGain tags is just now coming into existence.

QUOTE (Notat @ Jan 29 2011, 18:15) *
Same goes for the peak measurement. ReplayGain specifies storing sample peak. Storing true peak in ReplayGain tags is not in accordance with the proposal. Admittedly, the interoperability problems created by such a non-standard implementation are fairly minor and unobtrusive.

The tool makes it obvious to any user when it is in accordance with RG and when not. If you don't like the idea to have "True Peak" for RG I change the tool accordingly.
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 18:52
Post #169





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE ([JAZ] @ Jan 29 2011, 18:15) *

If you write replaygain tags, you ought to write them with the meaning that Replaygain applies.
If you want to conform to R128, the tags need to be different, and those can be in conformance with what R128 says.

It's not me writing the tags. It's possibly you by hitting the button. If you don't like it just leave it.

QUOTE ([JAZ] @ Jan 29 2011, 18:15) *

this is what you are being warned about repetitively in this thread.

WTF?
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 19:14
Post #170





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (Notat @ Jan 29 2011, 18:07) *
If you want to do loudness normalization in strict accordance with R128 specifications you should create and use R128 tags.

QUOTE ([JAZ] @ Jan 29 2011, 18:15) *
If i get a file with ReplayGain information, as in the file has the tags used by a Replaygain-compatible scanner to indicate gain i would expect that tool to conform to the replaygain standard meaning for those tags. It is not important what is used to calculate the tags, or the amount of difference from the reference implementation. What is important is that the tags mean exactly what Replaygain says it means.

Just to make it clear for another time: the tool follows benski's proposal to indicate the difference:

QUOTE (benski @ Jan 13 2011, 19:52) *
I think it might be worth doing something like
REPLAYGAIN_ALGORITHM=ReplayGain 1.0

or

REPLAYGAIN_ALGORITHM=EBU R128

so that if you are converting your library that's already been processed with ReplayGain, you would be able to keep track of what still needs to be done
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 19:23
Post #171





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (Notat @ Jan 29 2011, 18:15) *
QUOTE (punkrockdude @ Jan 28 2011, 16:12) *
Scanned lossless files with replaygain showed 1.00000 peaks at highest now shows over 1.00000. Does this have something to do with the True Peak option and is it checking if the file clips when being oversampled? Regards

Same goes for the peak measurement. ReplayGain specifies storing sample peak. Storing true peak in ReplayGain tags is not in accordance with the proposal. Admittedly, the interoperability problems created by such a non-standard implementation are fairly minor and unobtrusive.

Done.

The next release will flag checked "True Peak" as non RG compliant.
Go to the top of the page
+Quote Post
Nick.C
post Jan 29 2011, 20:10
Post #172


lossyWAV Developer


Group: Developer
Posts: 1815
Joined: 11-April 07
From: Wherever here is
Member No.: 42400



.... but surely if you are using existing ReplayGain tag names then a legacy program will not be able to differentiate using the new REPLAYGAIN_ALROGITHM tag as it will ignore it. Is there therefore a risk that the program will apply the wrong gain?


--------------------
lossyWAV -q X -a 4 --feedback 4| FLAC -8 ~= 320kbps
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 20:44
Post #173





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE (Nick.C @ Jan 29 2011, 20:10) *
.... but surely if you are using existing ReplayGain tag names then a legacy program will not be able to differentiate using the new REPLAYGAIN_ALROGITHM tag as it will ignore it. Is there therefore a risk that the program will apply the wrong gain?

The algorithm is applied during scan time, not at playback time. Both, RG and R128, scan the files, determine a gain, and write it to a tag. At playback time the player reads the tag and applies the gain accordingly. That's all.

The discussion is only about the loudness against which the gain is measured. RG says (about) -18.0 LUFS, R128 says -23 LUFS, hence R128 appears quieter during playback (by about 5 dB).

QUOTE (punkrockdude @ Jan 28 2011, 21:32) *
Hi guys. Is it the oversampling which makes the scanning procedur really slow and is the volume supposed to be quite alot quieter than replay gain? Regards.

That's the only consequence. It's just like discussing Fahrenheit vs. Celsius. If you don't like the one, chose the other. It's just one click on the GUI. The player will apply exactly the gain value according to the standard you've chosen.

It's your choice. Nothing is hard coded. The tools does nothing behind your back. It does exactly what you tell it. It's all visible on the GUI.
Go to the top of the page
+Quote Post
[JAZ]
post Jan 29 2011, 22:26
Post #174





Group: Members
Posts: 1798
Joined: 24-June 02
From: Catalunya(Spain)
Member No.: 2383



QUOTE (pbelkner @ Jan 29 2011, 19:14) *
Just to make it clear for another time: the tool follows benski's proposal to indicate the difference:

QUOTE (benski @ Jan 13 2011, 19:52) *
I think it might be worth doing something like
REPLAYGAIN_ALGORITHM=ReplayGain 1.0
or
REPLAYGAIN_ALGORITHM=EBU R128

so that if you are converting your library that's already been processed with ReplayGain, you would be able to keep track of what still needs to be done



That tells which algorithm was used to calculate the tags, not to which reference level the gains are based on. Replaygain expects it to be compared to -89dB (K-14 if you preffer). You can make the calculation with R128 at -23LUs, but when saving a replaygain tag, the gain needs to be relative to -89dB.


So no, this is not "A click on the GUI" neither about the unit used in the value. LUs vs dBs would be like Fahrenheit vs Celsius. A different base for the gain information is not.

This post has been edited by [JAZ]: Jan 29 2011, 22:33
Go to the top of the page
+Quote Post
pbelkner
post Jan 29 2011, 22:34
Post #175





Group: Members
Posts: 412
Joined: 13-June 10
Member No.: 81467



QUOTE ([JAZ] @ Jan 29 2011, 22:26) *
You can make the calculation with R128 at -23LUs, but when saving a replaygain tag, the gain needs to be relative to -89dB.

Are you really know what you are talking about? Sure?

Ever heard about the difference between absolute and relative values? What of the two you guess is a gain?
Go to the top of the page
+Quote Post

23 Pages V  « < 5 6 7 8 9 > » 
Closed 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: 26th December 2014 - 04:18