Author Topic: Replace dB with LUFS, and have a 'loudness' tag.  (Read 15227 times)

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
Once in a while there is still confusion about MusicBee's volume normalization.
In my opinion an important part of the confusion results from the fact that MusicBee displays the normalization setting scale in relation to the old ReplayGain algorithm, which has been replaced by an EBU R128 algorithm for a while now.

Therefore for example the current '0dB' designation is now both technical and conceptual incorrect.
The correct unit is LUFS. (loudness units) This is an important difference.

If the slider is set to 0dB, the actual result is a perfect -18 LUFS. (perceived integrated loudness)
Set to +6dB results in -12 LUFS, and -9dB results in exactly -27 LUFS.

I believe the current -9 , 0 , +6 dB scale indicators should be replaced by the above LUFS values.
Holding on to dB's would be clinging on to an incorrect and obsolete concept.
LUFS is also a standard for this, being used for a long while in broadcasting, professional audio software, etc. etc.
There will probably be people that object to such change, but let's be factual, and name things as they are.
Even if it means that we might have to re-educate ourselves a little bit.

And a following request and suggestion on this matter:
It would be very nice if the resulting (integrated) LUFS value after analysis would also be written in a tag.
That would give a very good and easy to find indication if a track indeed has the 'loudness' that you want it to have.

And an additional third wish on this matter:
If and when this is going to happen, it would also be a good moment to change the occurrences of the wordings such as 'volume analysis', 'volume leveling' etc. to the correct wordings 'loudness analysis', 'loudness leveling'.

The first reason for this is that 'loudness' is just the correct terminology here, which is also used in all papers and publications on R128/bs1770.
Second reason, 'volume normalization' is an old concept with different workings. It made loud passages within a track less loud, and soft sounds louder. So it actually changed the content (the volume of the waveform so to speak), and limited the dynamics to make sure you could hear the softer parts also at a lower volume setting.


The introduction of MB v3 could be a nice moment to do this?
Last Edit: June 10, 2015, 08:21:34 AM by hiccup

psychoadept

  • Global Moderator
  • Sr. Member
  • *****
  • Posts: 10691
+1
MusicBee Wiki
Use & improve MusicBee's documentation!

Latest beta patch (3.5)
(Unzip and overwrite existing program files)

redwing

  • Guest
Therefore for example the current '0dB' designation is now both technical and conceptual incorrect.

Wrong. There's nothing wrong about that. "0 dB" here means the amount of adjustment, compensation, not an absolute measure.

The correct unit is LUFS. (loudness units)

Again wrong. If following EBU standard it should be LU, not LUFS, here as it's a relative measure. But who knows what LU or LUFS means?

If the slider is set to 0dB, the actual result is a perfect -18 LUFS. (perceived integrated loudness)
Set to +6dB results in -12 LUFS, and -9dB results in exactly -27 LUFS.

So you know 1 LU/LUFS equals to 1 dB. Then why change it only to make it incomprehensible to most users?

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
Therefore for example the current '0dB' designation is now both technical and conceptual incorrect.

Wrong. There's nothing wrong about that. "0 dB" here means the amount of adjustment, compensation, not an absolute measure.

The correct unit is LUFS. (loudness units)

Again wrong. If following EBU standard it should be LU, not LUFS, here as it's a relative measure. But who knows what LU or LUFS means?

If the slider is set to 0dB, the actual result is a perfect -18 LUFS. (perceived integrated loudness)
Set to +6dB results in -12 LUFS, and -9dB results in exactly -27 LUFS.

So you know 1 LU/LUFS equals to 1 dB. Then why change it only to make it incomprehensible to most users?

Wow, are you the touchstone of what is right and wrong here?

1. You suggest 0dB means the amount of adjustment. So 0dB is no adjustment then? That would be a confusing and incorrect statement. Almost all tracks run through analysis with setting the slider to 0dB will be adjusted.

2. LUFS is a perfectly good term here: loudness units full scale.

3. You are suggesting that 'I know' 1dB = 1 LUFS. I am not. You are missing the point that dB and LU are of a different concept, and that is one reason for my suggestion.
You might like it or not, but the simple fact is that MusicBee now uses the LU concept in it's processing, but it displays a dB scaling.

The concept of 0 dB is in my opinion (and experience) difficult to explain and very arbitrary.
-18LUFS says the loudness is 18LU (or 18dB if you would like to think of it in that way) below full scale (maximum signal) I think that is quite simple to understand and to explain. It is also a common standard in the professional audio and video industry. Do you believe they are also 'wrong' about using LUFS for this?

4. Do you speak for most users? I don't think so.

I believe I am only stating facts, which are easily verifiable if you put a little effort in it.
If you personally want to hold on to an old (and in my opinion incorrect) habit, that's fine.
But then just say that. Don't try to kill a perfectly sensible suggestion by attacking it with throwing your 'wrongs' in this manner.

I would understand from a certain point that some people might miss a comfortable '0' setting.
But I personally prefer not to underestimate others in learning and understanding. It would also be easy to make an entry in the Wiki explaining this both in laymen's terms, and more in detail for the folks that really want to understand what is what.

redwing

  • Guest
Really disappointed at your attitude. I only stated my opinion that disagrees with you.
I won't waste more time on you.

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
I believe you were not just giving your personal opinion, but attacking my proposal with incorrect statements.
I am very open to contributing and funded objections, so if you would reconsider that would be fine by me.
If not, no problem.
Last Edit: June 06, 2015, 02:51:53 PM by hiccup

Bee-liever

  • Member
  • Sr. Member
  • *****
  • Posts: 3831
  • MB Version: 3.6.8849 P
+1

Since all the newer standards (ie:  ITU-R BS.1770, EBU R128, ATSC A/85 and ReplayGain 2.0) are using LU & LUFS as the preferred measurement, then it would also make sense to change MB.

As MB currently uses R128gain in ReplayGain compatibility mode, the scale should center on -18 to identify that this is the starting point with no manual adjustment.

As the author of R128gain has released an updated version (now as BS1770Gain) MB3.0 also should take advantage of this and update as well.

As the newer version supports switches for R128, ATSC and ReplayGain compatibility, MB3 could have selectable options for which reference loudness level you want to use and would automatically redraw the adjustment scale centered around that selected point;
  •    EBU R128 (Europe): -23.0,
  •    ATSC A/85 (Northern America): -24.0,
  •    ReplayGain 2.0: -18.0.

BS1770Gain also has the advantage of also writing a loudness reference and algorithm tag;
  • REPLAYGAIN_REFERENCE_LOUDNESS: -23.00
  • REPLAYGAIN_ALGORITHM: ITU-R BS.1770
   
MusicBee and my library - Making bee-utiful music together

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
That is some great input Bee-liever.
Hopefully this won't be attacked as it being 'to technical', since it is just factual, and the matter of fact is that we are dealing wit a very technical (background) of the matter.

My suggestion is of course also to make it easier to grasp for new users.
With the slider set to -3dB currently, you would have to explain it by saying: well, there was an old standard, which has never been official acknowledged by anybody, which MusicBee no longer uses, that at a certain moment advised on a certain perceived loudness level (which was calculated through a by now obsolete and inferior model).
Setting the slider to -3dB would result in a loudness that might be perceived as 86dB loudness, which is in itself is an incorrect statement, since if you would turn down your volume knob all the way down, it would be 0dB loudness.

With my proposal of using LUFS you could explain:
The setting of -21 LUFS means the humanly perceived loudness of the whole track has a loudness of 21 LU's below the maximum.


The only thing I hesitate about reading your posting, is if there indeed should be (possibly selectable) indicators as:
  •    EBU R128 (Europe): -23.0,
  •    ATSC A/85 (Northern America): -24.0,
  •    ReplayGain 2.0: -18.0.

This is certainly good practice in most existing professional audio mastering and broadcasting software, but I feel it might not be completely appropriate for an audio player. (even if it is MusicBee ;-)

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
A mock-up of how it could look:



The position (or even the existence) of the '0' could be argued, but considering the history and the way how many users will probably use this feature it's probably wise to leave it at it's familiar position.

Steven

  • Administrator
  • Sr. Member
  • *****
  • Posts: 34313
without having thought it though, my first reaction would be to have a setting so that MB fully uses the new standard instead of replay gain compatability mode - and when that option is enabled then the new scales would be used

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
Ok, great.

For my understanding: I do not understand this 'compatibility mode'.

The currently used algorithm seems to be exactly compliant 'loudness-wise' with the most recent "ITU-R BS.1770-3/EBU R128" standard? (I checked the results with professional tools)

Does that mean the 'compatibility mode' only lies in the naming of dB's in the scale, and the way the loudness tags are read and written?

Xyzzy

  • Sr. Member
  • ****
  • Posts: 396
From my POV it's replacing something I know and (roughly) understand with something I neither know nor understand.
I would prefer an option of leaving things as they are, even if a bit inaccurate.

Bee-liever

  • Member
  • Sr. Member
  • *****
  • Posts: 3831
  • MB Version: 3.6.8849 P
I do not understand this 'compatibility mode'.

R128gain.exe that MB uses for calculation actually calculates to the -23 LUFS  of the R128 standard. 
It has a 'compatibility mode' to make the loudness levels closer to the original ReplayGain standard.
All it does is add 5 LU to the result, to make it more 'compatible' with the 89dB SPL specified.
MusicBee and my library - Making bee-utiful music together

hiccup

  • Sr. Member
  • ****
  • Posts: 7790
Ok, so in other words, it is completely R128 compliant, but with the current default setting of 0 dB in MusicBee it only adds 5 LU to the loudness?
That's why I was a bit confused, the word 'compatibility mode' to me seemed to imply more different workings.
Thnx.

Bee-liever

  • Member
  • Sr. Member
  • *****
  • Posts: 3831
  • MB Version: 3.6.8849 P
From my POV it's replacing something I know and (roughly) understand with something I neither know nor understand.

Xyzzy,

Here is a simplified and hopefully 'plain english' explanation of loudness leveling, to try and help you understand the newer R128 and BS1770 way of doing it.

ReplayGain (RG) loudness leveling was designed for computer audio formats, to help overcome the problem of different albums and tracks being mastered at differing loudness levels, so you weren't continually having to use the volume control.

RG 1.0 standard was published back in 2001.  It was based on research done for playback levels in cinemas that was worked out in decibels (dB).  This was before multichannel sound was commonplace;  before researchers discovered that the human brain actually prefers loud sounds (why most of us turn the volume up on a song we like);  before music, ad and film producers started mastering their product louder and louder in an attempt to make us like their product more;  and before personal digital music devices where everywhere (remember the 1st generation iPod was released in October, 2001).

R128 is based on new research on how "loudness of sound" is perceived by the human brain and is measured in Loudness Units (LU) and designed primarily for broadcast TV.  So viewers weren't being deafened by the ads in between their regular viewing.

BS1770 expands on this to apply to all forms of broadcast or stored audio (eg: radio, television, cinema, CD's, digital files).

So basically it's the same - but better.  Not relying on the old dB SPL (Sound Pressure Level) way as this doesn't always equate to what your brain 'hears' as being loud or soft in volume.

Hope that improves the knowledge and understanding for you, Xyzzy, and anyone else that is having trouble with this.

MusicBee and my library - Making bee-utiful music together