Practical example of intersample peak greater than +6 dBFS

[bennetng]

Thanks for checking it. But yes, I have not disabled my 'room correction' which is just xo set for upmixing to MCH.

Here is it with everything set to stereo:
View attachment 71972

The total 'loudness integration is at -6 so that's what the OP was actually pointing at and not the true peak? I'm just guessing...

Duh... of course, you're probably right with what you showed in your tests.

Your Toneboosters screenshot shows +4.8dBTP with -6.3LUFS.

LUFS values are usually quite consistent among different analyzers since they are averaged values (a lot of samples). If you read my previous posts, Get Away has a Track gain of -11.76dB. foobar uses -18LUFS as default target, so it is 18 - 11.76 = -6.24LUFS, pretty close to -6.3 in your screenshot.

dBTP on the other hand can vary greatly among different analyzers since it estimates a single sample peak value. The one using SoX 95% passband shows +4.84dBTP (pretty close to yours), with 99% it shows +6.48dBTP, with SSRC (steeper than 99%, not configurable) it shows +6.87dBTP. So it means Tonebooster could be using a filter similar to SoX's 95% settings.

Ultimately, the "truth" about true peak depends on the actual playback environment, what the analyzers show are just educated guesses.

 

[KSTR]

I'm posting about this because I'm pretty disappointed with DAC digital filter design. It is this forum which has highlighted this problem, Many thanks to @amirm's measurements. Most DACs on the market don't even attenuate properly at the nyquist frequency for 44.1 kHz audio, which comprises the entirety of CD digital audio and many, many more digital music files. On top of that, all of these intersample peaks that exceed 0 dBFS means that a digital filter needs headroom to properly oversample. Lots of headroom, not a wimpy 3 dB, but at least 6 dB or one whole bit, and ideally 8-9 dB to cover these fringe cases too. This isn't rocket science, folks. It's 2020 and the math has been beaten to death. #MakeDACFiltersGreatAgain

While you are fully correct and IS-Overs up to, say 10dB, could easily be handled there is a noise 10dB penalty no-one is willing to make. IHMO it is good enough when IS-overs are handled faithfully up to 2dB or so and above that simple softclipping will do (nobody will ever hear a difference) without sacrificing too much signal-to-noise ratio. With folks here being so much focussed on SINAD, "IS-safe" DACs would measure and rank much worse that they do now. Not worth the trade-off...

 

[KSTR]

Further, demanding "full" attenuation at 22.05kHz (yet with 0dB @ 20.00kHz) is a two-sided sword. It may avoid the imaging frequencies but at the cost of an extremely long ringing of the impulse reponse. I'd prefer the occasional harrmless imaging (not: aliasing) any day over an "oscillator" type stop filter.
The only way to solve this is a wider transition range or higher sample rate. With an 18kHz end-of-passband or 48kHz the problem is greatly reduced, pretty much to insignificance.

 

[ernestcarl]

Ultimately, the "truth" about true peak depends on the actual playback environment, what the analyzers show are just educated guesses.

Yep, I see what you mean.

I tried your extreme test tracks just to see what my meters would show... I attenuated the L&R active speaker amp settings by a lot and turned off the rest...

Desk sitting MCH upmixing setting

Desk sitting stereo only setting -- which I very rarely use -- I guess I'm an MCH upmixing addict.

As KSTR stated, in most cases and with soft clipping protection, it probably wouldn't matter/be audibly obvious. Normally I would not be able to listen at such high volumes anyway with such highly compressed tracks for very long. Already, I pull back the volume much lower from music streaming sites/apps. If I had an 85dB SPL reference volume for all of my listening setups, I would pull the volume back even way more than I already do now...

 

[Gregg Hermetech]

Absolutely. Recording "engineers" don't have a fucking clue what they're doing these days. I wish people would just use 32-bit floating point for recording and mixing instead of whatever god-awful nonsense goes on in studios these days.

We are not all that bad! I don't think I've let an ISP through on anything I've mastered in about a decade now, and everything stays at 32 and 64 bit float throughout the mastering process until my final deliverable (I am not in control of what clients send me though). I'm known and hired for keeping the dynamics intact. There should be more communication and less blame between artists, recording, mix and mastering engineers, hi-fi buffs and regular listeners, IMO, we are all after the same goal I think.

To address the OP, the loudest tracks in my collection (about 24k lossless tracks) are all from Merzbow's 'Pulse Demon' album, they all clock in at just under 6dB TP (check out the positive LUFS Integrated figures too!) I've never seen one over 6dB. That is nuts...

 

[sarumbear]

Why not limit the audio with a look-ahead limiter at -0.1dBFS or thereabouts? Isn't that standard on digitising achieve recordings?

In live/studio recording your reference is -20dBFS, hence the chances of peaks reaching even 0dBFS is nil.

Meanwhile, how can one achieve a +dBFS value on a digital audio file. FS is short for Full Scale. What is fuller than full?

 

[KSTR]

Why not limit the audio with a look-ahead limiter at -0.1dBFS or thereabouts? Isn't that standard on digitising achieve recordings?

Look-ahead not enough. You have to upsample before the limiting and then downsample again to really avoid IS-Overs with brickwall limiting and that's what the better limiters do.

 

[sarumbear]

Look-ahead not enough. You have to upsample before the limiting and then downsample again to really avoid IS-Overs with brickwall limiting and that's what the better limiters do.

I was simplifying as I was making a point that limiting is standard practice. Why is the OP so cross?

Also, how can you have +dBFS value on an audio signal?

 

[KSTR]

Also, how can you have +dBFS value on an audio signal?

You do understand what Intersample-Overs are, do you? You have to look at the output values after reconstruction (basically doing in digital what the DACs digital and analog reconstruction filters would do), not at the the sample values alone.

 

[sarumbear]

You do understand what Intersample-Overs are, do you? You have to look at the output values after reconstruction (basically doing in digital what the DACs digital and analog reconstruction filters would do), not at the the sample values alone.

The OP said:

I have over 12,000 tracks in lossless format in my music collection, and a few of them hit a peak above +6 dBFS, while hundreds have peaks above +3 dBFS.

My question is how can a digital file system that stores sampled audio data can have +dBFS values? Where are those values read?

 

[DonH56]

Conventional limiting will not necessarily solve the intersample problem. If the signal trajectory is "up" on the first sample at FS, and "down" for the next, then the output will be over full-scale (+dBFS) after the output anti-imaging/reconstruction filter. It is not the DAC itself going over full-scale.

Edit: @KSTR is saying the same thing and he got there first.

 

[KSTR]

My question is how can a digital file system that stores sampled audio data can have +dBFS values? Where are those values read?

You look at it after a make-shift upsampling, or just directly at the analog output (if the DAC can handle IS-Overs, that is. Otherwise you reduce the level to the DAC to 1/4th -- -12dB roughly -- and look how much above that 1/4th FS equivalent peak voltage you read on the scope).
The details of the upsampling filter (or the DAC's filter) matter a lot, the closer you are to true sinc reconstruction the larger ar the IS-Overs.

 

[KSTR]

The whole problem is the DAC in the end. Either it has digital and analog headroom to handle IS-Overs (of 6dB or more, preferably) or it doesn't and clips the waveform.
You can avoid the problem totally by reducing the playback level to 1/2 or 1/4 full scale but then you loose 6dB ord 12dB of signal-to-noise ratio. Plus you'd ideally need to dither the signal again, notably when the DAC is 16bit only.

 

[sarumbear]

The whole problem is the DAC in the end. Either it has digital and analog headroom to handle IS-Overs (of 6dB or more, preferably) or it doesn't and clips the waveform.
You can avoid the problem totally by reducing the playback level to 1/2 or 1/4 full scale but then you loose 6dB ord 12dB of signal-to-noise ratio. Plus you'd ideally need to dither the signal again, notably when the DAC is 16bit only.

So those +dBFS values are internal values of the DAC and the letters FS does not mean Full Scale signal at input of the DAC. It is simply the DAC's way of quantifying the error it has generated internally during the D to A conversion.

 

[KSTR]

So those +dBFS values are internal values of the DAC and the letters FS does not mean Full Scale signal at input of the DAC. It is simply the DAC's way of quantifying the error it has generated internally during the D to A conversion.

Yep, that's it, basically.
EDIT: The proper unit is dBTP (dB true peak).

A hen vs egg issue. DAC manufacturers say "well, don't feed me with signals that generate IS-Overs" whereas producers say "well, a DAC should handle ANY sample sequence I'm feeding it."
Originally, when you have a digital signal that is created by (a sum of) ADC signals and "properly processed" you won't get IS-Overs.

 

[DVDdoug]

This track contains a lot of clipping, as is common in the genre and modern music,

I doubt the inter-sample peaks make it sound any worse.

It might even be worse with higher peaks than the actual-original cleanly-clipped waves!

 

[KSTR]

I doubt the inter-sample peaks make it sound any worse.

It might even be worse with higher peaks than the actual-original cleanly-clipped waves!

Yep, that was mentioned already, and what I wrote in post #20

 

[bennetng]

We are not all that bad! I don't think I've let an ISP through on anything I've mastered in about a decade now, and everything stays at 32 and 64 bit float throughout the mastering process until my final deliverable (I am not in control of what clients send me though). I'm known and hired for keeping the dynamics intact. There should be more communication and less blame between artists, recording, mix and mastering engineers, hi-fi buffs and regular listeners, IMO, we are all after the same goal I think.

To address the OP, the loudest tracks in my collection (about 24k lossless tracks) are all from Merzbow's 'Pulse Demon' album, they all clock in at just under 6dB TP (check out the positive LUFS Integrated figures too!) I've never seen one over 6dB. That is nuts...

View attachment 214435

Is this post yours?

Setup Foobar2000/FB2k for True Peak & LUFS Integrated Readouts

Preamble A quick guide to get FB2K displaying LUFS Integrated and True Peak levels per track. All info subject to change as and when FB2K is updated....

forums.stevehoffman.tv

Specifically this part:

True Peak Scan: On/checked Using: auto 4x oversample

auto 4x is using RetroArch low quality 4x and the filter is pretty inaccurate, as shown in the post below:

How DAC reconstruction filters would sound like if we were bats (listening examples)

There is a lot of discussion on how different DAC reconstruction filters sound like, besides the pure technical aspects which are easily explained and visualized with frequency response plots where the imaging -- mirrored frequencies above half the sample rate -- can be seen (if any), as well as...

www.audiosciencereview.com

Try the settings in the post below and scan Merzbow again:

Practical example of intersample peak greater than +6 dBFS

I've mentioned it before here on this forum, but it's entirely possible to have music that has intersample peaks beyond +6 dBFS, or one whole bit. It was mentioned again later on this thread by @LTig. Here is a 30-second excerpt of "Get Away" from the Electronic Soundtrack album Code Lyoko by...

www.audiosciencereview.com

 

[Gregg Hermetech]

Is this post yours?

Setup Foobar2000/FB2k for True Peak & LUFS Integrated Readouts

Preamble A quick guide to get FB2K displaying LUFS Integrated and True Peak levels per track. All info subject to change as and when FB2K is updated....

forums.stevehoffman.tv

Specifically this part:


auto 4x is using RetroArch low quality 4x and the filter is pretty inaccurate, as shown in the post below:

How DAC reconstruction filters would sound like if we were bats (listening examples)

There is a lot of discussion on how different DAC reconstruction filters sound like, besides the pure technical aspects which are easily explained and visualized with frequency response plots where the imaging -- mirrored frequencies above half the sample rate -- can be seen (if any), as well as...

www.audiosciencereview.com

Try the settings in the post below and scan Merzbow again:

Practical example of intersample peak greater than +6 dBFS

I've mentioned it before here on this forum, but it's entirely possible to have music that has intersample peaks beyond +6 dBFS, or one whole bit. It was mentioned again later on this thread by @LTig. Here is a 30-second excerpt of "Get Away" from the Electronic Soundtrack album Code Lyoko by...

www.audiosciencereview.com

Yes, my post, isn't that obvs? No, I don't use Retro Arch, I use SoX in FB2K which corresponds very closely with the other ISP meters I use (iZotope RX9 Advanced, RME Digicheck, Youlean Loudness Meter Pro, REAPER etc.), so moot point... The figures above ARE using the SoX Resampler.

Why, what figures would you expect from the crappy RetroArch?

 

[bennetng]

Yes, my post, isn't that obvs? No, I don't use Retro Arch, I use SoX in FB2K which corresponds very closely with the other ISP meters I use (iZotope RX9 Advanced, RME Digicheck, Youlean Loudness Meter Pro, REAPER etc.), so moot point... The figures above ARE using the SoX Resampler.

Why, what figures would you expect from the crappy RetroArch?

I meant, in this post:

Setup Foobar2000/FB2k for True Peak & LUFS Integrated Readouts

Preamble A quick guide to get FB2K displaying LUFS Integrated and True Peak levels per track. All info subject to change as and when FB2K is updated....

forums.stevehoffman.tv

If you choose auto 4x, then foobar will use RetroArch lowest quality even when you have the SoX plugin enabled. You can try the file linked below:

https://www.audiosciencereview.com/forum/index.php?attachments/40-space-traveler-stage-3-bgm-zip.45777/

auto 4x's true peak level is identical to manually choose RetroArch lowest quality 4x.
I also scanned the same Merzbow album with better resamplers and the dBTP is over 10dB.

Also observed by others:

Pathological example of a intersample peak, 11dB, discussion. - Page 2

Pathological example of a intersample peak, 11dB, discussion. - Page 2

hydrogenaud.io