Insights on True Peak Exceedance in Pre-Mastered Tracks Within DAW

[drissa]

Hello Everyone,

I'm looking for some help regarding an observation I've made with true peak levels in my DAW. I've been analyzing tracks that I downloaded from YouTube, running them through a Youlean loudness meter without any additional processing or gain changes. Surprisingly, these tracks consistently show a true peak value of +0.6 dB, despite them all being professionally mastered and presumably optimized for intersample peaks before being uploaded.

This consistent exceedance of 0 dB in true peak readings has me wondering if this could be related to the compression methods used during the upload process to platforms, or if there's another technical factor at play.
Is this a known issue that others have experienced, and what could possibly be causing these higher than expected true peak readings?

Thanks!

[BertKoor]

Why do you find this surprising? I'd say it's to be expected.

Engineers push their levels as loud as they can without youtube lowering the gain. Limiters are essentially clipping.

Imagine these sample values: 0.0, 1.0, 1.0, 0.0.
Draw a smooth curve through these values, and between these 1.0 values you have the peak excessing 1.0.

[drissa]

Why do you find this surprising? I'd say it's to be expected.

Engineers push their levels as loud as they can without youtube lowering the gain. Limiters are essentially clipping.

Imagine these sample values: 0.0, 1.0, 1.0, 0.0.
Draw a smooth curve through these values, and between these 1.0 values you have the peak excessing 1.0.

Thanks for your perspective.
I get that loudness is a target in mastering, but I was under the impression that true peak limiting, which accounts for inter-sample peaks, would cap the levels to avoid any exceedance of 0 dBFS. The recurring true peak readings above 0 dBFS for YouTube-downloaded tracks suggest to me that either YouTube's encoding is introducing these peaks, or there’s a standard practice among mastering engineers to leave a margin that allows for such exceedances. Since true peak limiters are designed to catch these inter-sample peaks, their consistent presence is unexpected. Could it be that YouTube's compression algorithms are responsible, or is it an accepted margin in the mastering process that I'm not aware of?

[BertKoor]

Intersample peaks are not as toxic as they used to be. Modern DACs can cope with them pretty well. And to put it in perspective, music with single-digit RMS / LUFS levels is guaranteed to be distorted anyway.

There are different "schools" of mastering engineerse. One school tries to avoid intersample peaks above 0.0 dB, another school says screw that: it's gotta be LOUDER! Because if you want to avoid them, you gotta TURN IT DOWN.

And then ofcourse you have YouTube, and only god knows what they'll do with it...

[No_Use]

Afaik it can happen when decoding compressed formats.
edit: Or was it when encoding already? I forgot.

[BertKoor]

True that. Going through compression & decompression results in an approximation of what it used to be.

Looking in detail at what was compressed audio (of a streaming platform imposing their own rules) is no use if you want to see what the mastering engineer really did. Better grab losssless FLAC files or the compact disk for that purpose.

[soundmodel]

There's no true peak limiter that acts on all true peaks. It reacts on some true peaks.

I have not seen a limiter that would have attack and release times of < 1 samples.

In other words, no limiter can limit the signal to a certain value without some peaks passing through.

A tool that does this is a clipper.

Or then you can normalize to 0.0 and it's certainly going to have the max peak at that.

You would not see intersample peaks in an "ITB" meter.

My guess is that the track is pretty hot and the Youtube (or whatever) compression or decompression algorithm does not know how to handle it so it can result in erroneous values in some rare cases.

Some support:

Lossy encoding and peak levels
https://izotope-rx.livejournal.com/5760.html

[plexuss]

A true peak limiter will limit peaks and prevent clipping. The issue of ISPs is not about the production/mixing/mastering stage as much as it is the post processing stage including playback. ISPs can happen during any kind of data conversion including data compression, digital to analogue conversion etc. The ISP model used in meters like Youlean is only meant to simulate "downstream" processing to try and catch if ISPs will be generated. ISP metering is not about what the audio is doing when metered but about what it might do during playback. It's an estimation and prediction and it's not fool proof. Generally anything mastered louder than -16 I-LUFS has the risk of generating ISPs on playback or conversion. ISPs wont be much of an audible problem unless the density of ISPs is high, meaning many are generated in a short period of time.