ReplayGain - True Peak 2x oversample setting

[krafty]

Hi folks,

I'm trying to understand the role of True Peak setting in ReplayGain. Releases ago, I hadn't seen this setting and now it is here.

My current settings are:

Album ID Tag
EBU R128
-18 LUFS
[ ] Disable Clip-Prevention
True Peak - 1x (no true peak)

This deems:
Soundgarden - Let Me Drown (FLAC)

Track Gain -7.22 dB
Track Peak 0.953857
Album Gain -6.11 dB
Album Peak 0.953857

When True Peak is set to 2x oversample, results are:

Track Gain -7.22 dB
Track Peak 0.994300
Album Gain -6.11 dB
Album Peak 1.017472

• What did exactly go on here?
• How is this information going to be used?
• What would it happen if someone would set to 16x oversample?
• How to get the "True Peak 2x" in already ripped albums without reripping?
• What is the best thing to do with the settings this DSP is presenting now?

Sorry if my question is dumb - it's just that I feel is not very well explained on the help page.

Thank you!

[Spoon]

"On the advanced page is a True peak option, this allows clipping to be calculated if the sound is going to be oversampled on playback (by doing oversampling can increase the peaks)."

[krafty]

Wow, that was a brilliant explanation.

[krafty]

garym care to drop a light over the subject, as you're known by the completeness of your answers and proper English? Thanks

[garym]

I’ll try in the morning.

[garym]

I’m no expert on this, but my understanding is that using 2x oversampling will give you better potential clipping protection on playback because the peaks are “better” calculated. Not all servers/players/DACs oversample on playback but many do in the background. So having this better peak calculation is simply “insurance” against clipping. Most files are not going to potentially clip in any case, so all of this is irrelevant for those files. But for those files that might clip upon playback, having this more precise peak info in the tags will be a good thing. Others may have a better technical explanation for how this all works.

I use these settings personally

Create track and album RG values
EBU R128
-18 LUFS
[ ] Disable Clip-Prevention (not ticked)
True Peak - 2x oversample

You can easily re-do the ReplayGain tags for your library without reripping. Do a batch conversion of your files, “converting” to the utility DSP “ReplayGain” with the settings above. This will simply recalculate the RG information and rewrite those tags. It will not touch or modify the audio in the files. Note, do NOT use “ReplayGain (Apply)”, as this will modify the actual audio data.

i hope this is useful.

[krafty]

Right. Thank you. That brought me to the understanding that, let's say...

Say the material is 16-bit and 44.1kHz.
The player, let's say a stereo, for some matter, will oversample the output to 48kHz (or let's say it's an "audiophile" equipment that oversamples to 192kHz).
In that condition, if the player does not have an understanding of the ReplayGain tag this is useless.
If the player does have the understating of ReplayGain 2.0 specs, the "no true peaks" will end up clipping upon this process (the output).
Not something that someone will actually hear, but technically speaking, the tag will help to keep the true high peaks below 0.0dB.
My understanding is that if the soundcard of the device outputs 48kHz, the 44.1 material will oversample - so that is the condition for it to happen.

I watched a couple of videos about this, but particularly talking about this 2x/4x/8x/16x setting for True Peak, why did you come to the conclusion that 2x will suffice to you? Is it just a default no brainer that you chose or have you got an understanding that the CD material is appropriate? That's what I am trying to understand... will there ever be a reason so one would specifically choose 8x or say 16x?

Thank you for your help concerning the re-scan of the files. I believe foobar2000 also has a component that will do this. But this is good to know.

Thanks again.

[krafty]

Right. Just did an experiment here and I encoded 5 times, one disc album.


Noel Gallagher's High Flying Birds
Back The Way We Came - Vol. 1 (2011 - 2021) Disc 02
Track: Sample for track 01. Black Star Dancing

Results:

No true peak (x1) - FLAC 4
Track Gain -9.20 dB
Track Peak 0.995392
Album Gain -10.23 dB
Album Peak 0.995392
Time to rip this CD (pristine, no frames re-read): 1m42s

True peak (x2) FLAC 4
Track Gain -9.20 dB
Track Peak 1.011167
Album Gain -10.23 dB
Album Peak 1.054810
Time to rip this CD (pristine, no frames re-read): 1m44s

True peak (x4) FLAC 4
Track Gain -9.20 dB
Track Peak 1.016745
Album Gain -10.23 dB
Album Peak 1.057646
Time to rip this CD (pristine, no frames re-read): 1m47s

True peak (x8) FLAC 4
Track Gain -9.20 dB
Track Peak 1.021825
Album Gain -10.23 dB
Album Peak 1.074482
Time to rip this CD (pristine, no frames re-read): 1m59s

True peak (x16) - FLAC 4
Track Gain -9.20 dB
Track Peak 1.021825
Album Gain -10.23 dB
Album Peak 1.074483
Time to rip this CD (pristine, no frames re-read): 2m39s

So it looks like rising the times value, it will increase the accuracy against these peaks. RG levels will stay the same. That means the limiter of the hardware will have to understand not only the Track Gain and Album Gain tags but also, the Track Peak and Album Peak tags, so it can apply a limiter.

I am sorry for the criticism in my previous posts but I think this feature should be more better documented so users can understand this setting as it does have a huge impact on encoding time if one accidentally pushes this to x16. One may have this going unnoticed just because he/she is not knowing what is taking place and find the software "very slow" as the difference between x1 and x16 is almost a minute longer to encode, and x8 not far behind.