Waveform unable to display above 0dbfs 32bit clips correctly?

[Ivan Bergamo]

The thing just confuses me.

If I consolidate some drastically gained Clip (in 32bit session), it sounds the same and the meters are showing the same peaks, but the waveform seems compressed in comparison with the original - Just like the waveform display is not able to show info above 0dbfs correctly, and it just cuts the peaks (visually, but not audibly)

Am I understanding it correctly?

[Darryl Ramm]

What Pro Tools does with > 0dB signal levels potentially varies even if it claims to support 32 bit files and 64-bit mixing. Pro Tools will display a "clipped" signal in the UI in many cases without actually clipping the data, but it can actually clip the signal for some things. This includes that Pro Tools clip gain (unlike fader gain) does not work correctly at extreme gain, which is very disappointing. I've played enough with this and say inspecting downstream audio clip files in python and comparing the raw data in the file with what the Pro Tools UI shows. Again surprisingly disappointing stuff when it unnecessarily actually clips floating point audio.

And the moment you process that signal through any plugin... what it will do is anybody's guess. So test/experiment and work out for your particular case.

[Darryl Ramm]

For example here is a clip created from the signal generator of a sine wave at 0dBFS output. Then recorded to a track/clip and then clip gain at maximum and recorded to another track/clip. It's just clipped to hell ********.

[Ivan Bergamo]

Thanks Darryl for the confirm.
As far as I understand, PT Waveform display is limited to show signal up to 0dbfs. Anything above will be represented flat to 0.

And according to your research, Clipgain is also somehow limited and not working accordingly with the 64bit mixing engine, while faders are working.
Maybe the recording part is not working in 32bit... I will not speculate, but definitely I'll need to run couple of tests just to make sure I know the tools I am using.

Yesterday I was quite surprised when simple gain consolidation cut all my peaks on the waveform, and still done nothing on the meters.
Today I am not surprised. :)

[Ivan Bergamo]

One more thing Darryl, not sure what version are you using, but there is that feature Pro Tools Noise Burst Peak Protection, in the 2019.5+ versions, which also behaves like some kind of hard clip.

This now may be 3 different topics we are talking about. :)

[Darryl Ramm]

I'm using 2022.10. But all versions I've seen recently behave the same and may have done so for a long time.

No idea about Noise Burst Peak Protection, never heard of it or played with it.

[Farhoof]

I've addressed this before here: https://duc.avid.com/showthread.php?t=415064
but it turns out the bug was actually a feature that was seemingly introduced quietly and 3 years later we got the option to turn it off. It is indeed the 'noise burst peak protection', anything over 20dB above zero is muted (not clipped but silenced). Not just files, but faders and inserts as well (maybe this changed in the latest version though).
The waveform display has always been a bit messy though.

[nednednerb]

'noise burst peak protection'

was...

'introduced quietly'??

But will this be "the last we hear of it"?

[Darryl Ramm]

Farhoof - thanks for letting us know about this. Great info, I was unaware of the noise burst stupidity but had noticed the Pro Tools screwing up signals at what turns out to be ~ +20dBFS.

Ivan - and yes my issues with damaged/incorrect float files go away when I disable "Noise Burst Protection".

But Pro Tools waveform UI is still broken, it can't display signals above 20 dB even when it is correctly mixing those signals and writing them out to 32 bit float files.

The screenshots attached show a quick test where applying +36dB of clip gain to a clip/file recorded from the output of signal gen at 0 dBFS.

The waveform display in the clip with the gain applied to it disappears off outside the track view, which is correct UI behavior but when I record that into another track it shows a waveform that appears actually clipped. Which is just broken. Yet if I open that clip's file it's a perfectly good sine wave at 0dBFS + 36dB = 63 amplitude in linear units as shown on my plot of that waveform. That plot I made of the waveform is that file in the highlighted bottom track showing apparent clipping.