Is 32-Bit Float broken ?

[EdgarRothermich]

Here is some weird thing about the 32-bit float setting in ProTools.

Here is a great YouTube demonstration about the benefit of using 32-bit float in Pro Tools
https://youtu.be/_pBdcYiEuqY?t=542

That is easily reproducible and quite impressive to see. However, when I tried to do it today it didn't work and the 32-bit float clip still stayed distorted after I reduced the gain. I was 100% it worked when I tried it last year, so I opened an older version of Pro Tools from last year 2018-12 and surprise, it worked.

I switched back and forth with different versions and it was constant, any Pro Tools version after 2018-12 didn't work. This is very frightening unless I overlook something here.

Is it possible that the 32bit float processing is broken in Pro Tools or does it work differently?

Can anybody confirm that or have an explanation for that?

Thanks

[mandal]

mac or windows ?

[XJENSEN]

Yes, I was on the phone with support yesterday for about 5 hours (!) about this and other issues. Although, to a degree, I felt good intention in them trying to help me, I'm now also pretty sure first level support lacks a lot of knowledge to judge whether it really concerns a bug or not. Only if you can convince them, which takes a lot of effort because they don't know what they're dealing with, only then they will forward it to engineering 'for evaluation'.

But I can confirm that the latest versions of Pro Tools are having various, and serious, problems with 32 bit float files. For example, the AudioSuite 'Normalize' function is also broken on 32 bit float files. Iteratively applying it to a region will keep shrinking the audio level to zero. Make sure to read my other thread: http://duc.avid.com/showthread.php?t=407185.

I don't care about any new feature anymore, until they fix the absolute core of the audio program, properly dealing with audio files. After that, they should fix the numerous ADC and LLM bugs, the MIDI timing, and then we can talk again.

[zelenov]

Sadly, it is. I moved to the PT2018 version. PT11 is also OK.

[Carl Lie]

This not good. Even worse if they don't know about it and can't fix it.

Sad.

[mandal]

Hi

I did a test here - using my Zoom F6 to record a 32-Bit float file...

https://soundcloud.com/4sound_mandal...-float-zoom-f6

I shouted "avid" as loud as I could in the microphone

Pro Tools Ultimate 2019.6 & Win 10 => 32-bit float & 48 kHz session:

I imported the file:



The waveform look maybe a bit strange - after reducing 36 dB - but the sound is not clipping - the sound is good !!

[XJENSEN]

Originally Posted by XJENSEN

For example, the AudioSuite 'Normalize' function is also broken on 32 bit float files. Iteratively applying it to a region will keep shrinking the audio level to zero. Make sure to read my other thread: http://duc.avid.com/showthread.php?t=407185.

In all fairness and for completeness: After thorough troubleshooting with Avid, what I reported above was not really a bug after all (but there are others).

It is caused by the default AudioSuite settings for rendering mode and handle lengths. When you use 'Create continuous file' in combination with 'Entire selection', the audio handle length gets added to the playlist selection and the process looks to adjacent clips for determining normalisation ceiling, and not simply within the parent audio file. Quite confusing, but I get it now.

[mandal]

Thanks for good info.

Is my waveform correct after just a reduction of 36 db - or is this a bug ?

[XJENSEN]

Originally Posted by mandal

Thanks for good info.

Is my waveform correct after just a reduction of 36 db - or is this a bug ?

Well, hard to judge, but it might be. Can you use the Normalize or Gain AudioSuite plugin to turn it down even more? Eventually, you should see the 'real' waveform step out of the shadow, to be sure it's not clipping anywhere.

The problem is PT can only render the waveform in the playlist view up to the 24 bit level. Even if you zoom out the waveform vertically, it will keep this level as a 'ceiling' and not reveal the part of the true waveform above the 24 bit level.