Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Is the K-Weighting in EBU-R128 is really better than the old one? (Read 6283 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Is the K-Weighting in EBU-R128 is really better than the old one?

Dear All,

Foobar uses EBU-R128 instead of old ReplayGain. But I still found some audio files could not keep in the same loudness level.
For example, some classic music, especially the piano sonatas, will be louder than the heavy rock songs after applying the ReplayGain.

For example, the EBU-R128 gives the result as below:
Get Over --> -10.68dB
Diver --> -12.84dB
(Beethoven) Moonlight the 1st movement --> +19.55dB
lettera Amorosa : Kapsberger - Toccata VI --> +19.49dB


Listening to these tracks, I prefer to increase about 3dB for Driver & Get Over, and reduce 4dB for Toccata VI.

So, I am now doubting the K-weighting used in R128 is really respecting to the loudness perceived by human.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #1 –
I also found some crazy things with track gain - even when tracks would be played random. Both RG and r128.  The more loud the album is the worse it can be .  In Queenadreena 'butcher and the butterfly' LP the album is -12db then theres a soft track 'cold light of day'  the almost whispered vocals sound louder than metal music - you can hear the noise floor.  Another one is Opeth 'Coil' track - the acoustic guitar strums are louder than most other music next to it.  German band Flowing Tears , album Jade - you can't listen to some instrumentals even on random as it sounds wrong. Dredg 'el cielo ' LP - 1st track 'brushstrokes' its simply unreal what RG track mode does.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #2 –
Why was David Robinson's algorithm abandoned?

I think it's becoming clear that the rush to the shiny, new, industry-aporoved "professional" R128 was done in haste.

There could be an upside to all this, however.  As this standard becomes more prevalent, the trend may shift towards being dynamic, if only in order to be louder.

(*) I'm beginning to wonder whether it involved any bias-controlled preference testing.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #3 –
I certainly performed my own testing before it was taken into foobar2000 and it produced better results. Why aren't these complainers posting any samples?

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #4 –
AFAIK R128 better supports high samplerates and multichannel audio.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #5 –
Quote from: shadowking on
I also found some crazy things with track gain - even when tracks would be played random. Both RG and r128.  The more loud the album is the worse it can be .  In Queenadreena 'butcher and the butterfly' LP the album is -12db then theres a soft track 'cold light of day'  the almost whispered vocals sound louder than metal music - you can hear the noise floor.
Um, yeah, that's the purpose of track gain. If you want to preserve the relative dynamics use album gain.

Quote from: shadowking on
Another one is Opeth 'Coil' track - the acoustic guitar strums are louder than most other music next to it.
Again, use album gain. If the dynamics are like that on the very CD, blame the producers.

Quote from: greynol on
Why was David Robinson's algorithm abandoned?

I think it's becoming clear that the rush to the shiny, new, industry-aporoved "professional" R128 was done in haste.
You're pretty quick to dismiss something which has been analyzed and discussed at length, based only on two anecdotes.

Quote from: greynol on
There could be an upside to all this, however.  As this standard becomes more prevalent, the trend may shift towards being dynamic, if only in order to be louder.
There is no indication of that happening. Music production is a shit now as it was 5-10 years ago.
It's only audiophile if it's inconvenient.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #6 –
There will be always tracks for which RG computation isn't perfect no matter whether old or new method. Just modify the computed RG value according to your needs.
lame3995o -Q1.7 --lowpass 17

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #7 –
Quote from: Case on
I certainly performed my own testing before it was taken into foobar2000 and it produced better results.

I'm sure you and others did, and I'm in no position to second guess the impressions your test corpus gave you, so I won't.

Quote
Why aren't these complainers posting any samples?

In this discussion I see people have made mention of track titles.  I hope that should suffice.

I'll offer up this...
https://www.youtube.com/watch?v=0EIRF7QW0eA

Anyway, I'm sure there are examples where the original algorithm has issues as well.  Identifying samples shouldn't be that difficult.  You simply need to identify tracks with the largest discrepancies in track gain between the two versions and compare them against tracks which have the smallest discrepancies.

It might be interesting to compare the relative averages between tracks that have a correction gain of smaller than -8dB (using the old algorithm) and greater than -4dB to determine whether R128 isn't further punishing tracks with heavy DRC (for better or for worse).

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #8 –
Quote from: Kohlrabi on
has been analyzed and discussed at length

It has?  How much time transpired between initial coding and adoption?

Quote from: Kohlrabi on
based only on two anecdotes

Yes, based solely on two anecdotes. 

Quote from: Kohlrabi on
Quote from: greynol on
There could be an upside to all this, however.  As this standard becomes more prevalent, the trend may shift towards being dynamic, if only in order to be louder.
There is no indication of that happening. Music production is a shit now as it was 5-10 years ago.

The standard will not become more prevalent?

Here's what Ian Shepherd has to say about Sound Check.  I share the same opinion:
https://youtu.be/yAB_SBioRI4?t=311

Full disclosure: he does explicitly say he hasn't found that the loudest songs being played back the quietest as being a problem with R128.  Then again, he made this video before the Green Day re-releases.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #9 –
Quote from: halb27 on
There will be always tracks for which RG computation isn't perfect no matter whether old or new method. Just modify the computed RG value according to your needs.

Indeed!

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #10 –
I also noticed this. Now I know I'm not alone ) Maybe it depends on frequency response of speakers/headphones?

Quote from: Case on
Why aren't these complainers posting any samples?

Samples: 1 (harp solo), 2 (brutal death). After applying replaygain (mode: track, processing: apply gain and prevent clipping) first sample sounds noticeable louder to me (I'd say too loud in comparison with second sample). My headphones are HD 380 pro.

Quote from: halb27 on
There will be always tracks for which RG computation isn't perfect no matter whether old or new method. Just modify the computed RG value according to your needs.

This is main reason why I not use replaygain at all. Beacause I have many such tracks.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #11 –
Quote from: greynol on
I think it's becoming clear that the rush to the shiny, new, industry-aporoved "professional" R128 was done in haste.

[...]

(*) I'm beginning to wonder whether it involved any bias-controlled preference testing.

I apologize for the snark.  I've been overly-preoccupied with the taste of hypocrisy from a different discussion.
 

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #12 –
Rollin, I checked the spectrogram of those two tracks in Spek: http://i.imgur.com/fFlxMOu.png

I wonder if this could be due to how R128 doesn't account for the equal loudness contours above ~15kHz.

Could be interesting to try applying a low-pass filter starting around 15kHz that's 30dB down by 20kHz or so (just guessing those numbers by looking at https://en.wikipedia.org/wiki/Equal-loudnes...ile:Lindos1.svg ), and then run the low-passed versions through the volume scanner to get the track gains.

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #13 –
That's my main objection also. High frequencies. The weighting should have a lowpass..
"I hear it when I see it."

Is the K-Weighting in EBU-R128 is really better than the old one?

Reply #14 –
Quote from: Rollin on
I also noticed this. Now I know I'm not alone ) Maybe it depends on frequency response of speakers/headphones?

Quote from: Case on
Why aren't these complainers posting any samples?

Samples: 1 (harp solo), 2 (brutal death). After applying replaygain (mode: track, processing: apply gain and prevent clipping) first sample sounds noticeable louder to me (I'd say too loud in comparison with second sample). My headphones are HD 380 pro.

Quote from: halb27 on
There will be always tracks for which RG computation isn't perfect no matter whether old or new method. Just modify the computed RG value according to your needs.

This is main reason why I not use replaygain at all. Beacause I have many such tracks. So JUST having track gain is only the 1st step. Realistically its somewhere between the two or a super intelligent track gain is the ideal.



Album gain will solve it and work ok even when playing random tracks. Sometimes there are slight differences between 'normal' tracks but its usually not annoying. When you have only normal tracks the track gain works well. I noticed this too: Best Of compilations [pat benatar / thompson twins] have album gain set from the factory when in theory it should be TG. Or they apply some sort of manual adjustments to make the subtle tracks not stand out in a weird way.