Loudness Normalization Module and Batch Analysis questions

WaveLab
1

Maybe the developers can shed some light on this, as the manual is definitely confusing me.



I want to use the Loudness Normalization Module either in the main editor, or in the batch processor (here it’s among “Multi-Pass Plugins”).

I want to normalize to a SLk value. (read: short term loudness)

But I do get two options, if I don’t want to use “Loudness of Entire File” (aka: integrated loudness)
Top of Loudness Range (Höchstwert des Lautheits Bereichs)
Maximum Short Term Loudness (Maximale kurzzeitig gemittelte Lautheit)


So I took a dive into the manual, and can’t seem to understand the difference here?
Am I missing something?

An example:
I have a track that was mixed with a reference level of -18dBFS / 0 VU.

Wavelab analysis shows on RAW Loudness (setup to a 300ms window) an average signal level of -19,45dB and a maximum of -15,09dB - which seems to be correct if I look at the readouts I got in the Cubase project

The analysis with the EBU R-128 meter shows: LRA (Loudness Range Average) 8,3LU (-26,7 LUFS to -18,4 LUFS) and a short term maximum of -18,3 LUFS

So both the maximum LRA and the maximum SLk value is roughtly -18,4 LUFS.


Then may I ask what the essential difference is with the SLk presets in the Normalizer Module?
Why should I use one over the other?

\



Then I just found out that I can do batch analysis of files (I’m late to the game, barely needed batch analysis so far so I did everything by hand). Very handy. But…!

Is there a way to do that without processing audio files?
And where does Wavelab save the files to?

I couldn’t select them destination for them, and they are not dropped in the folder where the WAV/MP3 files are dropped.


Furthermore - what is meant by “Loudness Balance”?
Nothing in the manual, and it’s the first time I head about it as well.


Thanks in advance.

2

In the batch processor, you can set “No audio output” in the Output tab. This means the file are analysed, but not written. Only the analysis is produced.

“Loudness Balance” is the difference of loudness between the left and right channels.

3

Thanks, that clears two questions.

I don’t really see why the “Loudness Balance” is important in any statistics. But if it can be generated, so be it.

Where does Wavelab save the files though?



Leaves the question with the difference between the Loudness Normalization options :
Top of Loudness Range (Höchstwert des Lautheits Bereichs)
Maximum Short Term Loudness (Maximale kurzzeitig gemittelte Lautheit)


Personally I’d love to see a normalization mode for “average short term”, but this needs a setting of specific rules for analysis. Still… I don’t get the main difference of these two. Aren’t they essentially the same since they both(!) focus on the max values of the determined loudness based upon the ballistics of SLk?

Or did that change with WL8.5 and only WL8.0 has three options?

4

Where does Wavelab save the files though?

The analysis file is saved near the source file (if no audio file is written), or near the output audio file.

Leaves the question with the difference between the Loudness Normalization options :
Top of Loudness Range (Höchstwert des Lautheits Bereichs)
Maximum Short Term Loudness (Maximale kurzzeitig gemittelte Lautheit)

To answer this question, you have to read the litterature about what is the “Loudness Range” and the “Maximum Short Term Loudness”. This is explained in papers related to the R-128 standard.

Or did that change with WL8.5 and only WL8.0 has three options?

AFAIR, this is the same in 8 and 8.5

Philippe

5

Have to check that.
But where is it saved if audio files are processed? So far only the programs popped up (Open Office, Text or Browser, depending on what format I selected), but I didn’t find the actual file.


I understand the difference.

Loudness Range is the range of the measured SLk from lowest measured point to highest measured point.
Max short terms is pretty much self explanatory.

What I don’t get… of both(!) measurement readouts essentially offer the same value - why offer two different LN algos?

You (and your team) programmed it, you should be able to tell me (the user) what the difference is, no?
Especially if it’s not further documented in the manual.



FR (sadly, only for 8.5 from this point forward):

  • The “Analysis” in the batch processor should also drop Loudness Range Min/max and not just the value of the range.
  • loudness (RMS) min/max/average with setup of the time frame (NOT AVAILABLE at present)
  • maybe a custom reference level for the EBU Loudness Analysis (not just -23LUFS)

So I still have to do manual analysis for RMS values… not really time saving, sadly.

6

But where is it saved if audio files are processed?

There are many options for this. Please refer to the Output tab and to the manual.

Loudness Range is the range of the measured SLk from lowest measured point to highest measured point.

No, this is too much simplified view. Loudness Range is a more clever measure. I quote the spec:

The computation of Loudness Range is based on the statistical distribution of measured loudness.
Thus, a short but very loud event would not affect the Loudness Range of a longer segment.
Similarly the fade-out at the end of a music track, for example, would not increase Loudness Range
noticeably. Specifically, the range of the distribution of loudness levels is determined by estimating
the difference between a low and a high percentile of the distribution. This method is analogous
to the Interquartile Range (IQR), used in the field of descriptive statistics to obtain a robust
estimate of the spread of a data sample.

To make it more simple to grasp, the loudness range is not the measurement between the extremes, but between an average of the extremes.

7

Okay, I just took a dive at the EBU R-128 tech paper tech3342 to brush up my knowledge.

The tech paper clearly states on page 5, I quote:

Loudness Range (abbreviated ‘LRA’) quantifies the variation in a time-varying loudness measurement. Loudness Range is supplementary to the main audio measure, Programme Loudness, of EBU R 128. Loudness Range measures the variation of loudness on a macroscopic time-scale, in units of LU (Loudness Units). The computation of Loudness Range is based on a measurement of loudness level as specified in ITU-R BS.1770 [2], albeit with a different gating threshold (see § 3.1). Loudness Range should not be confused with other measures of dynamic range or crest factor, etc.

It further states that the gating settings are -20LU below the absolute gated integrated value (Q10 IIRC), and that the measurement is from 10% (10% above the relative threshold for LRA) to 95% (of the maximum SLk value) of the actual measured range.

Now… and this is the kicker… it’s still measured in SLk.

And sometimes the max values coincide with each other - depending on how loud the material to measure was to begin with. If it is content, that was mastered at -7dB RMS or something along those lines, there is a definite shift in the values. If it’s more hovering around -20dB RMS to -15dB RMS, the values could be similar.


Those that do not educate themselves with the fine details will definitely be confused with the tagging of the algorithm. Maybe add the “10% to 95% range” between ( and ) in the description or extend the manual on that end? For non metering tech heads, this might definitely be essential knowledge.



Still, it doesn’t clear what(!) the ideal route is for loudness normalization for music. In my opinion, for overcompressed content, the Max LRA algo looks more appealing. For non-mastered content, then it’s Max SLk Algo. For video content, then the “Loudness of Entire File” (which could be tagged “integrated” as well btw).

However - most of the loudness normalization tools on the market also only offer either ILk or SLk, not max LRA. So what’s your opinion on this, PG?




Still, my FR’s from the last post remain regarding the batch analysis tool:

  • The “Analysis” in the batch processor should also drop Loudness Range Min/max and not just the value of the range.
  • loudness (RMS) min/max/average with setup of the time frame (NOT AVAILABLE at present)
  • maybe a custom reference level for the EBU Loudness Analysis (not just -23LUFS) to show the offset

And I also tested the function to “drop the analysis file”.

I must be doing something wrong, because the analysis file is NOT saved in the folder where I created the WAV files. Neither is it saved if I don’t process audio files. It’s thrown on C:\ in the temp folder and opens up the corresponding program.

I tried “use subfolder” also “No output”, I tried “also save data list”, the behavior is the same.
This is slightly annoying. Unless I have to save each patch process in an own folder and it will be dropped there. Then again, I mostly don’t need that as the batch process is rarely recalled at a later state (currently only for testing purposes).


Thanks once more.

8

However - most of the loudness normalization tools on the market also only offer either ILk or SLk, not max LRA. So what’s your opinion on this, PG?

Indeed, I added a normalizing option based on LRA top value, because that value is more weighted than the Short Term Loudness. Hence more related to a global perception of the music. Now, you are the judge regarding the results.

It’s thrown on C:\ in the temp folder and opens up the corresponding program.

True. I was wrong in my previous message. If no audio files are written, the analysis files are written in the temporary folder. Opening of the file is optional.

9

So you’re actually offering a different “concept” of the Loudness Normalization process.
That is indeed interesting and might need further tests on my end. And it definitely needs an update of the manual (at least IMO).


Any chance for further options for the batch analysis module, and choosing a location for the file drop?

What about a “save results as TXT” in the regular Analysis (key Y) module? All within the 8.0 or 8.5 cycle of course. If you’re still going to maintain 8.0.x that is.

10

Any chance for further options for the batch analysis module, and choosing a location for the file drop?

What further options? You can choose the temporary folders in the WaveLab Folder preferences.

What about a “save results as TXT” in the regular Analysis (key Y) module? All within the 8.0 or 8.5 cycle of course. If you’re still going to maintain 8.0.x that is.

This could happen in the future (>= 9).

Philippe

11

A third time:

  • The “Analysis” in the batch processor should also drop Loudness Range Min/max and not just the value of the range.
  • loudness (RMS) min/max/average with setup of the time frame (NOT AVAILABLE at present)
  • maybe a custom reference level for the EBU Loudness Analysis (not just -23LUFS) to show the offset

Regarding the regular Analysis tool (main editor):

This could happen in the future (>= 9).

Why not within the 8.x cycle?

Unless v9 is due in Spring 2015 already. :unamused: