Fixed vs floating point processing

[Mark Ziebarth]

Hi,
I have two setups in my studio: My main room is still PT HD 10 with HD 3 and the second room is PT 12 HD with a Fireface 802. Sometimes I start a session on 12 and mix it in 10. McDSP support told me that the general difference in maths (32 bit fixed point in 10 and 64 floating point in 12) may cause problems. The second room is bigger for more clients but not a real mixing room. My question: What should I take care of to avoid to run into trouble while using a Session started in 12 in 10? Normally I'm using a little Channelstrip here and there in the recording session and never remarked a difference when opening the session with the older software.
What I know is I have to look for the gain staging.
But does anyone have any practical examples when those maths differences are important for session compatibility between 10 and 12?

Any help appreciated, best regards

Mark

[reichman]

Yup. If you send a bunch of hot tracks to a bus in fixed point and it clips, when you turn down the bus, it just makes the clipping quieter. In floating point, turning down the bus actually gets rid of the clipping. First time I heard this, my jaw dropped. Floating point mixing is very different. If you properly gain stage the old-fashioned way in PT12, you shouldn't run into trouble, but keep your ears open.

[Mark Ziebarth]

That's good news, thank you. I'm old school so my gain staging is appropriate, normally... But this support message confused me...

[Mark Ziebarth]

If you started in the early days of digital audio like me the floating point maths is a weird thing. I would never lower the volume of the buss if the signal input is too hot. But I understand it's possible with actual software and that's a good thing.
That's why I like the DUC so much. Just this one answer of you Nathaniel helped me a lot. Good to have you here! Thanks again.

[reichman]

You're welcome. There's a lot more to it than that, but the bussing is a place where people get into trouble.

[Mark Ziebarth]

Yeah, the bussing issue is a practical example to understand the difference. Simply a good advice! I understood and am happy now! Go on with the good work! Next year I'm in here for 20 years, huh...

[Darryl Ramm]

Originally Posted by Mark Ziebarth

32 bit fixed point in 10 and 64 floating point in 12

There is nothing that is 32 bit in "10", and by that I assume you mean 32 bit fixed point in legacy HD/TDM? Pro Tools legacy HD/TDM processing is 24 bit fixed point plugin processing and 48 bit fixed point mixing. Current Pro Tools (and Pro Tools HD Native) and Pro Tools HDX are 32 bit float plugin and 64 bit float mixer. In both case the plugins can also operate internally at the higher (48 bit boxed/64 bit float mode if needed).

[Mark Ziebarth]

Thanks Darryl, you're completely right. But I was just searching for situations where this maths matters. And my result is: If you're gain staging old school nothing obvious can happen. Am I wrong?

[Darryl Ramm]

Originally Posted by Mark Ziebarth

Thanks Darryl, you're completely right. But I was just searching for situations where this maths matters. And my result is: If you're gain staging old school nothing obvious can happen. Am I wrong?

If by "gain staging old school" you mean you never make a mistake in gain staging and you trust the plugin developers did a good job then sure nothing can go wrong. But assuming you never make a mistake is uh optimistic. It's always nice to have more headroom.

But even with 32/64 bit float processing you still need to gain stage stuff/watch levels. The input level matters greatly to what the plugin is modeling/going to do with that signal.

[Bob Olhsson]

Only very simple 32 float plug-ins that aren't modeling analog gear don't require gain staging. Even then some go crazy when confronting over the top signals. It's best to turn the signal down before applying signal processing and then turn it back up at the end of the signal path.