Loudness compression, loudness wars.. What exactly it is and why is it happening?

[dc655321]

Bah... I should have put my sandwich down and actually paid attention to what I was writing.
Previous plots were only representing one channel.
Not that there is much of a difference...

 

[DonH56]

Interesting. I remember from ages ago the mean peak-to-average ratio in music was measured to be 17 dB (think in an AES paper). Your data is around that for the peak-to-rms values FWIW.

 

[Guermantes]

Wow, nice to knock up some Python code over lunch

All of those tracks seem brickwall limited at 0 dB FS -- look at the histograms for the peaks! The Tragically Hip Scared track seems to be somewhat healthier based on it's RMS and peak distributions. Perhaps it would be interesting to compare DR analysis for some of the tracks.

So, regarding the DR analysis approach, we would take the top 20% of the histograms for the calculation. Is that too little to be truly representative? I know it cuts out a lot of the outliers, but is it also throwing out some variations that are significant?

I keep thinking that a simple waveform plot shows me all the information I need on judging dynamic range and mastering variations at a glance.

 

[danadam]

My definitely-not-random selection below (I wanted something with rms below -20 dB and also 2 masters of the same track).

Also I'd like to point out that, for whatever reason, the formula for "RMS" in DR algorithm is "the square root of the double sum".

 

[dc655321]

My definitely-not-random selection below (I wanted something with rms below -20 dB and also 2 masters of the same track).

So, you're saying I should put a more flexible front end on this?
i.e. pass in a list of one or more files, or choose a number of files randomly, etc?

The Led Zeppelin "deluxe" edition is interesting.
Why do they almost always produce a more compressed version?
I have a similar example with a Massive Attack song - the remastered version has half the DR of the original and it totally sounds like it!

Your other entries are not something I would normally listen to, but I'll bet they sound fantastic.

Also I'd like to point out that, for whatever reason, the formula for "RMS" in DR algorithm is "the square root of the double sum".

Well, I just wrapped off a quick 'n dirty toy to get some degree of insight: strict adherence to the DR algorithm was not on my agenda.
I'm not sure what the author(s) could be double summing over or why? I took that line to mean summing over both channels.
Do you have a different understanding?

 

[Guermantes]

For your reading pleasure, the algorithms for the EBU R-128 recommendation used in the LKFS meters are outlined in detail here:
https://www.itu.int/rec/R-REC-BS.1770-4-201510-I/

 

[Guermantes]

So, you're saying I should put a more flexible front end on this?
i.e. pass in a list of one or more files, or choose a number of files randomly, etc?

I suppose if you simply put 5 flac files in the target directory you are guaranteed to get all of them, but in a different order each time you run the analysis.

 

[Guermantes]

I suppose if you simply put 5 flac files in the target directory you are guaranteed to get all of them, but in a different order each time you run the analysis.

Ah, no, that doesn't work . . .

Anyway, here is a plot of YELLO The Race (uncompressed) which I analysed with DR and WLM above.

 

[danadam]

So, you're saying I should put a more flexible front end on this?
i.e. pass in a list of one or more files, or choose a number of files randomly, etc?

The first step would be to read artist/album/title from tags instead of file path

I'm not sure what the author(s) could be double summing over or why? I took that line to mean summing over both channels.
Do you have a different understanding?

I can only speculate that maybe he wanted DR value of a sine wave of constant amplitude to be DR0. If not for that doubling, the computed DR value would be DR1 DR3, indicating that constant sine wave has some dynamic range

 

[Guermantes]

I can only speculate that maybe he wanted DR value of a sine wave of constant amplitude to be DR0. If not for that doubling, the computed DR value would be DR1, indicating that constant sine wave has some dynamic range

Good point. I guess that a sine wave does technically have a crest factor (whereas a square wave has none).

 

[Blumlein 88]

Good point. I guess that a sine wave does technically have a crest factor (whereas a square wave has none).

A sine wave has a crest factor of 3 db.

 

[danadam]

A sine wave has a crest factor of 3 db.

Ah, yes, so it would be DR3. I'll fix my post.

 

[DonH56]

Actually, the crest factor of a square wave is 1, same as for a DC signal. https://en.wikipedia.org/wiki/Crest_factor

 

[danadam]

Your other entries are not something I would normally listen to, but I'll bet they sound fantastic.

That is true. Peppino D'Agostino, Acoustic Guitar, Desert Flower:

 

[dc655321]

The first step would be to read artist/album/title from tags instead of file path

I would not consider that a "first step", but sure, that's a good idea for improvement(s). Thanks.

 

[TBone]

The Led Zeppelin "deluxe" edition is interesting.
Why do they almost always produce a more compressed version?

Perhaps because the public (and so many audiophiles) have long accepted such re-mastering as the norm.

The Barry Diament mastered LZ catalog, along with some early Japanese pressings, offer much less compression.

 

[bennetng]

That is true. Peppino D'Agostino, Acoustic Guitar, Desert Flower:

The audio of this video is within Youtube's loudness standard so that playback volume in browsers is not being altered.

 

[dc655321]

The Tragically Hip Scared track seems to be somewhat healthier based on it's RMS and peak distributions.

The best albums I have, in terms of DR, are all from The Tragically Hip: Fully Completely, Day For Night, Trouble at the Henhouse.
These are mid-'90s albums, so their reasonable DR rating were a little surprising.
RIP, Gord Downie :-( You are sorely missed.

 

[TBone]

The best albums I have, in terms of DR, are all from The Tragically Hip: Fully Completely, Day For Night, Trouble at the Henhouse.
These are mid-'90s albums, so their reasonable DR rating were a little surprising.

Nice to see that G.Downie is available with limited compression.

What is the ave. DR rating of those CD's.

I'd have to check again, but if memory serves ... my Hip cd's range from good to crap, in terms of DR values, none close to the more dynamically capable (measured) music - within in my inventory.

 

[dc655321]

I'd have to check again, but if memory serves ... my Hip cd's range from good to crap, in terms of DR values, none close to the more dynamically capable (measured) music - within in my inventory.

On closer inspection, only Fully Completely (1992) stands out as reasonable.
I thought the others were better than what they're rated at...

What is the ave. DR rating of those CD's.

Average DR from here:
Fully Completely - 13
Day For Night - 9
Trouble at the Henhouse - 9