PT mixer headroom severely reduced to 20dB over 0dBFS. Bug?

[dankin]

Originally Posted by Wolfgang Eller

But still, why is there any need to gain a file 20db above 0dbfs?
Any 32bit float file I get from a client I convert it by import into a 24bit fixed.
For me there is no reason for using 32 float.
I have learned to use gain structure.

Same

[daeron80]

Originally Posted by Wolfgang Eller

But still, why is there any need to gain a file 20db above 0dbfs?
Any 32bit float file I get from a client I convert it by import into a 24bit fixed.
For me there is no reason for using 32 float.
I have learned to use gain structure.

It's not about need, nor is it about gaining up an individual file. Look, I've been a pro audio guy since 1993. I started out on 2" tape and analog consoles. I know gain structure very well. But stuff happens, especially under tight deadlines. It's rare. But it happens.

[Wolfgang Eller]

Originally Posted by daeron80

But stuff happens, especially under tight deadlines. It's rare. But it happens.

I just can't imagine how this kind of mistake will be made by a "pro audio guy"?

I'm earning my money in studios and live since 84 but that says nothing.

[Farhoof]

Let's agree nobody is affected by this since nobody noticed it in the last 1.5 years, but we would all like to have a faulty PT mixer be checked because what else might be going on...

[daeron80]

Originally Posted by Wolfgang Eller

I just can't imagine how this kind of mistake will be made by a "pro audio guy"?

I'm earning my money in studios and live since 84 but that says nothing.

I'm happy for you that you never had to face the kinds of situations in which huge surprises happen and there's no time to deal with it the way it needs. But some of us do.

[JFreak]

Originally Posted by daeron80

I'm happy for you that you never had to face the kinds of situations in which huge surprises happen and there's no time to deal with it the way it needs. But some of us do.

If it is recorded wrong, record it again and do it right.

[daeron80]

Originally Posted by JFreak

If it is recorded wrong, record it again and do it right.

Doesn't apply. In certain situations, stuff gets sent to you from people who know nothing about audio, and you have to turn around immediately. You can't just go to a conference in the past and re-record a message, for instance.

[Wolfgang Eller]

Originally Posted by daeron80

I'm happy for you that you never had to face the kinds of situations in which huge surprises happen and there's no time to deal with it the way it needs.

I got tons of shi##y tracks but I never ever need to gain it 20db above 0dbfs to adapt it to the rest of the tracks.
If this happens something else must be completely gone wrong.

[daeron80]

Originally Posted by Wolfgang Eller

I got tons of shi##y tracks but I never ever need to gain it 20db above 0dbfs to adapt it to the rest of the tracks.
If this happens something else must be completely gone wrong.

Quite right. Sometimes, the things that are completely wrong are out of the engineer's control. But I'm not thinking of gaining a single file to that point. You're right that it's pretty hard to imagine a situation where that would happen.

[Rich Breen]

Originally Posted by JFreak

If it is recorded wrong, record it again and do it right.

I'm surprised a moderator would be so dismissive of another professional's concerns.

And...the issue here is that there seems to be something wrong with the math in clip-gain and audiosuite processes - it may be an edge-case that it has to be 20dB above full scale to flag it (and it's good that it's not in the actual floating point mixer dsp - that would make it *not* an edge case), but clips should be recoverable to literally hundreds of dB above "full scale". If they're not, it indicates a problem with the math, and that should be of concern to the coders, period. If one doesn't have faith in the math of basic operations then it calls into question what *else* might be lurking under the hood to bite you when you're not looking?

If someone who actually is involved in the coding said: "we decided to shuffle those kinds of operations off to a different engine that only allows recovery to 20dB over FS", I'd say fine, but this apparently used to work fine, and AFAICT in a fully floating point environment should still work fine, but appears to have broken in the last couple versions. Again, this will not affect anything I do, but it's very odd.

Please stop dismissing competent engineers' justified concerns, and please stop pointing fingers at the user - this should work correctly...full stop.