switching to 32 float from 24 for working?

[sayatnova]

Hello Everyone,

I have been at 24/96 for all of my work for a while. I fairly recently switched to a new Mac Studio. I have the horse power, I am wondering if I should start working at 32 float, now? I see that a few folks here have stated they are working at 32 float (Mike Halloran, for one), and I am wondering if this would be best? I believe someone also stated that Matt from Motu also said this was a good idea? Is this only for clipping issues, or something else?

I always keep levels moderate, and am very careful with gain staging at all times. Is there any other benefit other than protection from clipping I would get from the 32 float? Is it more taxing on the CPU, I know there will be larger file sizes...

Also, any benefit to switching existing 24 bit session to 32 float mid-work (after having recorded at 24...)?

I have tried to go through the forum and get a clear answer to this, but I am more confused now! Any help or info would be most appreciated.

Thank you.

[FMiguelez]

Hello Everyone,

I have been at 24/96 for all of my work for a while. I fairly recently switched to a new Mac Studio. I have the horse power, I am wondering if I should start working at 32 float, now? I see that a few folks here have stated they are working at 32 float (Mike Halloran, for one), and I am wondering if this would be best? I believe someone also stated that Matt from Motu also said this was a good idea? Is this only for clipping issues, or something else?

The thing is that the noise floor is so low for either bit depth that you won't notice any difference (your DAC is 20-22 bits if you're lucky). 24 bits gives you so much headroom that is more than enough for normal audio purposes. I suppose you could use 32 bits if you're going to record something specialized or crazy like recording bomb explosions paired with quiet murmur noises. Otherwise, common sense and good gain staging take care of that.

If you're going to be processing audio files many times offline, then I suppose 32 bits are justified.

I do not recommend recording new audio at more than 24 bits except when the source is greater than 24 to begin with (I don't know of any sample libraries in 32 BFP. There's no point in recording a singer at this rate either because of the noise floor of your room, etc).
However, for stems and final mixes, 32 bits are fine (so you don't have to deal with dither noise VS quantization distortion at this stage).

I always keep levels moderate, and am very careful with gain staging at all times. Is there any other benefit other than protection from clipping I would get from the 32 float? Is it more taxing on the CPU, I know there will be larger file sizes...

Also, any benefit to switching existing 24 bit session to 32 float mid-work (after having recorded at 24...)?

Nope. No benefit at all. You would still be working with 24 bit files dressed as 32 BFP, so in that sense it would be a waste of resources. It would be like converting 8-bit files to 24 bits... it's still an 8-bit file, but over-dressed.

Mind that DP ALWAYS works at 32 BFP internally, even if your audio files are at 4 bits.
The setting in the transport window tells DP how to record/merge/process offline the new audio files.

You could do a bounce to disk of your final mix at 32 bits so you don't need to use dither and you keep your mix at DP's native precision; then you just convert it to 24 bits at mastering once (although multiple passes of dither are still harmless, since that noise doesn't add up as one would expect -and this is not true for the QD).

It's all about the noise floor.

[James Steele]

I have been changing over to 24bit/48KHz for all my personal stuff. Not that this adds much to the conversation, but I used to use 44.1 back when CD was a more popular format. Seems like video has always been 48KHz, so there's that if your audio will end up on video you would have had to upsample. Seems the conversion algorithms are so good these it's just better to downsample your mix to 44.1 in the end. I've never used 96KHz or higher, but I'm sure it has it's purposes. I don't think I could hear the difference anyway, given that I'm old enough to have some accompanying high end roll off in my hearing I'm sure. That's what standing a few feet from a drum riser and a crash cymbal 6 feet from your left ear will get you! Ha!

[mikehalloran]

During lockdown, I was making choir projects using audio files sent me by senior citizens. These were recorded in all sorts of formats including stand-alone Boss 44.1/16 recorders. Vocals were often sung into iPhones and Android, a few GarageBand and even one ProTools. I had DP auto-convert everything to 48k/24 bit.

My challenge was to take anywhere from 14 to 30 of these tracks and mix them till they sounded like a choir. Ignoring the tremendous number of hours it took to sync consonants, the big problem was keeping individual track levels low enough to eliminate distortion in the Master. Tracks recorded too close have intense saturation and wild dynamic range that DAWs and pop music plug-ins were never meant to tame — and couldn't, frankly. Certain plug-ins helped tame peaks such as bx_limiter True Peak (a life saver) while MOTU's MW Limiter made it worse. The Spectral Recovery and Declip modules from RX Advanced helped me get levels low enough—I could see how hard the plugs had to work depending on the levels and adjusted down accordingly.

After about a year and a half, I switched auto-convert to 48k/32bit float just to see what happened. Holy cow! I had about 30dB more headroom instantly. Yes, individual tracks can still distort if pushed (the nonsense that this is not an issue in 32bit float is just that) but the time savings was huge. Also remember that only the accompaniments were tracked 32 bit float. If I could keep track levels to -10 on the meters (still needed bx_limiter True Peak) I had a good chance of a decent mix in far less time. EIOSES E2 De-Esser became another essential plug-in.

I still bounce to 48k/24 wav. Adobe Premiere likes to crash when I would add 32 bit floating audio. I don't hear a difference bouncing to 32bit float in any case.

Do I think 32bit float is necessary? Normally, not for the instruments and voices used in pop music. On the uh… 'harmonically saturated' (the kindest term I can think of) source material that I had to work with? Oh hell yes.

It's a simple test: you hear the difference or you don't. If you don't, you can participate in all the gearspace boards where armchair experts discuss what they can't hear. If you hear the difference with your material, then it's easy: you know.

This choir gig has ended. A cost cutting decision was made once the churches went back to live and I'm out of a job. The choirs are feeling the pain of working with a director with no knowledge of the human voice — but that doesn't bring me back. Oh well.

It doesn't mean that I'll go back to 24 bit. I'm spoiled.

[mhschmieder]

Yeah, I switched to 32-bit during lockdown as well, and have noticed a huge difference in headroom as well as audio integrity across the signal path.

The thing is, so many products now work in 32-bits internally (e.g., iZotope RX), so I figure I may as well go to that format to reduce constant back-and-forth conversions.

This doesn't necessarily mean I'll record mics at 32-bits vs. 24-bits, but it's nice that we can now mix bit depths in the same DP project, reducing the need to convert older tracks if we overdub in a preferred format later on.

I did make one exception for mic tracking last year, as my North Carolina band had limited dates to record and so I had to overdub my bass parts after the fact. As they did it all as a live room take, I had a lot of work to do to try to get some instrument separation and intelligibility (though fortunately it was well-miked so was fairly well balanced), meaning that I was coming into DP at 32-bits from iZotope RX. At that point, I figured overdubs should go at 32-bits as well.

[sayatnova]

So, if I understand correctly: if my levels are conservative and able to be controlled, it is safe/fine to record at 24/96. Then goose the session to 32 for the mixing and mastering stages? Is this correct?

[labman]

Since we have a team spread out in various places, we moved to 32bit float many years ago, (32/48k) for cutting and mixing on everything. As cited by others above, that made it so much easier to deal with various peak levels and compressed levels coming in from various racks in the differing buildings. And our mastering guy of many years always wants us to give him 32bit float printed at conservative levels. 32/48

There were four on our team who used Logic in their work. They noticed thru time, that their work was sonically less desirable than all the peeps who were cutting at 32bit float in DP. Especially if they went thru plugins. (Logic cant do 32bit float to cut files from what i understand). Long story short, every one of them is now in DP. And they haven't looked back.

So FWIW we have noticed no downsides.

[sayatnova]

@Labman: are you and your folks recording their material in their studios using DP at 33 float? Or, editing already-recorded files?

Thank you.

[James Steele]

Hmmm.. maybe I'll have to switch over this year. Used to be storage space was a major factor in this, but I guess with drives continuing to get larger and cheaper, that argument against will be rendered moot soon enough?

[HCMarkus]

Correct me if I'm wrong here but...

1. Unless one captures audio using a device that combines dual A-to-D, assigning one converter to low-level signals and another to loud signals, there is no advantage to recording at 32 bits. And even if one has such a device, unless is capturing extremely dynamic material (generally field recordings), there is no benefit, as the signal being recorded is well within the 21 or 22 bits of resolution a quality A/D converter provides, respecting residual noise the mic pre (and condenser mic built in amplifier) adds to the recording.

2. Once audio is in your DAW (like DP), audio processes take place within the 32 bit floating point realm. So nothing is gained, to this point, by using 32 bit files.

3. It is not until an audio file has been processed in some manner (gain change or plug in) and the result of that processing rendered as audio within the DAW that any advantage is provided by 32 bit files. Freedom from potential clipping and/or loss of resolution are the advantages here.

Clearly, when transporting audio files from system to system or rendering and re-working audio within a DAW, there are advantages, mostly freedom from worrying about levels, to be gained by recording 32 bit float. In such a case, perhaps recording at 24 bits, then changing the file type within the DAW (without altering the file type for already-recorded audio) would provide the best of both worlds... more compact 24 bit files from tracking sessions, 32 bit mix audio file output.

For in-studio recording, I don't see other advantages. If I am missing something here, please educate me. Thanks!

[sayatnova]

@HCMarkus: This is exactly what I have come to, as well. In the spirit of wanting as many sources to draw from, I asked this same question to another forum of well-respected audio Folks and got this same answer... Here is a link to that thread https://www.tapatalk.com/groups/proreco ... 15344.html

Interesting topic, for sure.

[HCMarkus]

@HCMarkus: This is exactly what I have come to, as well. In the spirit of wanting as many sources to draw from, I asked this same question to another forum of well-respected audio Folks and got this same answer... Here is a link to that thread https://www.tapatalk.com/groups/proreco ... 15344.html

Interesting topic, for sure.

Indeed. Thanks!

[James Steele]

...unless is capturing extremely dynamic material (generally field recordings), there is no benefit, as the signal being recorded is well within the 21 or 22 bits of resolution a quality A/D converter provides, respecting residual noise the mic pre (and condenser mic built in amplifier) adds to the recording.

Yeah... I was thinking that for someone like my self that is basically almost exclusively recording rock/metal kind of stuff, there'd be no real benefit.

perhaps recording at 24 bits, then changing the file type within the DAW (without altering the file type for already-recorded audio) would provide the best of both worlds... more compact 24 bit files from tracking sessions, 32 bit mix audio file output.

Now THAT is an interesting idea! You're saying track at 24-bit and once "acquisition" is done and you're switching to mixdown mode, switch DP over to 32-bit? I'm too lazy to launch DP right now to check, but if samples rates are the same, it has no problem with sessions that use both 24-bit and 32-bit-float files at the same time, yes? This sounds like a good idea. The for final delivery you convert the resulting 2-track mix to either 48/16 or 44/16 depending on where it ends up? Like say if you're a dinosaur and decide to put it on a CD?

I need to go read the discussion at that link, but I'm liking your idea for workflow, Markus!

[mikehalloran]

(Logic cant do 32bit float to cut files from what i understand).

As far as I know, LPX has supported it since 10.7 over Catalina 10.15.1. The lated Zoom F series recorders record 32 bit float—here's a compatibility chart.

https://zoomcorp.com/media/documents/F8 ... 2MAR_E.pdf

perhaps recording at 24 bits, then changing the file type within the DAW (without altering the file type for already-recorded audio) would provide the best of both worlds... more compact 24 bit files from tracking sessions, 32 bit mix audio file output.

Nah... As I posted earlier, almost nothing I do is recorded at 32 bit float save for my voice and occasional instrument tracks. I auto-convert upon import to DP. The headroom increase and cleaner mixes in less time… haven't found a downside.

I mix to 24 bit, however, because a) that's what others ask for, b) I don't hear a difference in stereo masters and c) 32 bit float likes to crash Adobe Premiere (hoping that the 2023 update fixes this). I also believe in having the largest System drive possible so that storage space isn't an issue — 4TB now and 8TB in my next Mac — with a pair of 8TB SATA III SSDs in an external TB2 dock for archive and projects I may have to revisit someday (speed isn't a big issue but I won't use HDDs except for Time Machine over Ethernet).

[labman]

@Labman: are you and your folks recording their material in their studios using DP at 33 float? Or, editing already-recorded files?

Thank you.

They all record at 32 bit float. Just seems to make things easier for them to keep it all the same. BUt...... we made the switch so many years ago, i cant remember why. (after several strokes i forget a lot) In some of our rooms we do not use stock converters on interfaces, but higher end 3rd party ones.