Recording: 24bit vs. 32bit float

[mikeyman]

how would I switch to 32 bits at mixdown.?Do you mean if i don't mixdown within the same session to stereo tracks?I have been mixing down within the same session.One time i did bring the stereo mix into a new session and masterd in the new session.How are you mixing down ?thanks

[Bob Olhsson]

Go into the session setup dialog and switch it to 32 float from 24.

[mikeyman]

thanks bob

[John McDaniel]

I firmly believe that the much criticized harshness attributed to the sound of Pro Tools, going back a decade or more, was largely due to the failure of people to pay strict attention to gain staging within the Pro Tools mixing environment. If you recall, the identification of overs in internal pathways was not as good in past versions as it is now. For those of use who were trained in gain staging before the rise of digital mixing solutions, it seems rather silly that so many people would abandon good practice when entering the digital realm in the late 90's and early 2000's (many of those people having also worked fully analog pathways in their past). Others, entirely new to mixing when engaging with Pro Tools, might be given some leeway for their ignorance, but it still remains true: attention to gain staging is important, valuable, and ultimately makes things sound better.

For the most part, having never not paid attention to gain staging, I'm not convinced that the benefits of 32 bit are so overwhelming as to make it a de facto standard --or a reason to pay less attention to good gain staging practices.

j

[WombatStudio.Org]

Originally Posted by JFreak

Try this:

#1: open up a session that does not clip. want to be sure? set up a limiter that doesn't allow over -6dBFS out. bounce it out for a test.
#2: create a 32bit float session and import/convert that bounced test file into that session
#3: drive the faders up and/or use whatever plugins you need to get it 100% surely clipping and sounding bad. then bounce it out as 32bit float.
#4: create another 32bit float session and add that clipping bounce into the new test session
#5: use clip gain to bring levels down and prepare to be amazed.

This is what is demonstrated in the video I linked from Russ, Pro-Tools-Expert. The plus for 32bit is within the realm of after you record in your LIVE tracks (at 24bit), but again - you'd have to be doing something wrong if you even clip at 24bit - but bouncing out your final mix at 32bit float seems like a decent safeguard.

But the consensus seems to be that recording from your converters at 32bit float is not necessary and won't grant you "unlimited headroom" due to the limitation of going through your 24bit converter. Within the DAW you benefit from the 32bit float setting, but again - not necessary unless you're crazy with your levels.

[mikeyman]

so 32 bit is just a safeguard incase you clipped a track and want to fix it.

I didn't realize I could switch out of 32bit floating once I had setup a session that way but I guess I was thinking it was like the session sample rate.Well I think I'm getting a better sound on pro tools than my trusty roland 2480.But I think a lot has to do with my new Lucid 88192 converter .Still have to get use to recording at not so hot levels.But I sure is a lot less stressful staying away from that clip light . And all these plugins make it a lot more fun.

[Craig F]

In theory the 32 bit file will better capture the sum of the tracks over the 24 bit
And for switching bit depths, you can switch to 16 bit if you wanted to for some kind of overdub if that's you fancy

[WombatStudio.Org]

Originally Posted by Craig F

In theory the 32 bit file will better capture the sum of the tracks over the 24 bit

This is where it gets dicey. Yes, you would certainly think this, BUT - if you don't overextend the 24bit headroom, a resulting 32bf file should be exactly the same. (exactly the same with all the extra bits being 0).

In other words, the internal mix engine uses 32bit float, so everything you're mixing is now 32bit float so by your theory "in theory", you would lose the benefit of the 32bf mix engine by printing a 24bit file.

When in actuality if you look at it blatantly, 24bit vs. 32bit float files are simply volumes. Nothing to do with frequency range, etc. If you don't exceed the 24bit headroom, you'll never touch the capability of the 32bf benefits so printing a 24bit file from a 32bit float mix loses nothing if you're mixing at proper levels.

[Craig F]

Who said anything about frequency? The theory is in quantization steps no frequency.

And I never said I subscribed to the theory, I was just putting it out there so it's a fair discussion

[Pete Gates]

Originally Posted by Top Jimmy

Bob, you are correct in that a 32-bit floating point word only encodes 23 bits of signal. The most significant bit is always implied since it can never be anything other than 1. There's no need to encode it if it never changes.

This isn't correct. A IEEE754 single precision number (32 bit float) can correctly (with no error) represent all unique 24 bit two's complement integer values. I think the source of the confusion is the 23 bit mantissa.

Pete