32 bit Float question for Magic Dave

For seeking technical help with Digital Performer and/or plug-ins on MacOS.

Moderator: James Steele

Forum rules
This forum is for seeking solutions to technical problems involving Digital Performer and/or plug-ins on MacOS, as well as feature requests, criticisms, comparison to other DAWs.
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: 32 bit Float question for Magic Dave

Post by Shooshie »

Well, all I can say at this point is that as an experiment in recording digital audio, I wish I had not done it. 32bfp serves a very real purpose in mixing. As an audio file format, it serves no purpose whatsoever for normal recording. It does not improve the perceptible quality of the audio. Any extension of the dynamic range is way beyond the useful range of music, anyway — and I'm talking about jet-plane exhaust to bug footsteps in the same room. There is no musical advantage to 32bfp.

My thinking was that it might facilitate ease of mixing with so much headroom that essentially I would never have to worry about digital clipping, and I could put off all dithering until the final 2-track. Those ideas were poorly thought-out, hastily decided upon, and without merit once you start doing the math and studying what goes into bit resolution. I think that 24 bits/48K is pretty much audiophile quality, and all we need in recording small studio projects. If we start recording rocket launches, then let's talk.

Until then, my conclusion is that 32bfp as an audio file format is not really a viable tool for the studio under most circumstances. I'm glad it's there; there may be a use for it. I just don't know what it is. 24 bits is all we would ever need, and even that is probably overkill for most digital audio applications. Then why do we need 32bfp at all? As a mixing domain. Because when mixing and processing 24 bit audio, we must be able to do that from within a more capacious format, or we will quickly run out of room. 32bfp is essentially 24 bits plus an 8 bit mantissa, so that it focuses 24 bits of resolution where it is needed at any given moment. My understanding is that this reduces rounding errors.

Rather than being easier to mix, 32bfp audio file format seems harder. Levels clip easily, and when they do, you hear it, contrary to my original rationale. Also, when working with amplifying and synthesizing bass harmonics, I'm picking up a LOT of noise from the floor. Actually, it's more like stuff outside the studio. Our guitar tracks are not recorded with the same shielding as the vocals, and if there was a bulldozer outside, you might pick up some rumble. Every day there seems to be a bulldozer or some other heavy machinery working on all these tear-downs and new houses. It had not been a problem until this 32bfp project. So, I'm using Waves' Z-Noise to remove it.

That's it. In a nutshell: 32bfp as an audio file format means more careful mixing and levels, more low-level audio noise (but lower digital noise floor). That has been my experience with this project. I'm not sure I can explain it, and I won't try. You'd be better off consulting people who think about this every day and who know the math without having to work through it each time. I'm just stating the differences I've noticed in this project, vs. my normal way of working. My other problems appear to have been unrelated to using the 32bfp audio file format, and appear to be fixed since installing the latest build of DP 9.13.

Shooshie
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: 32 bit Float question for Magic Dave

Post by Shooshie »

terrybritton wrote:Long and short of it - 32-bit float is handy for archiving to hand off for a mix and for internal processing, 24-bit is great for recording as it is within the range that most plugins can understand, and 16-bit with slightly compressed audio is ideal for listening environments with higher environmental noise floors than the 16-bit digital "noise" floor is (-96db on the meters).

Like, most of the planet! :D

Perfectly stated. 100% true. And I wish I'd thought of archiving as a use for 32bfp file format, which makes sense. Especially for stems. As your primary audio format, I do not recommend it.

Shoosh
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
terrybritton
Posts: 1117
Joined: Thu Jun 04, 2015 8:45 am
Primary DAW OS: Windows
Location: Elizabeth City, NC
Contact:

Re: 32 bit Float question for Magic Dave

Post by terrybritton »

Shooshie wrote:32bfp is essentially 24 bits plus an 8 bit mantissa, so that it focuses 24 bits of resolution where it is needed at any given moment. My understanding is that this reduces rounding errors.
This is the point I was making in my last post - Magic Dave had gotten me on the wrong track with talking about crazy-high dynamic range from the 32-bit float... no, that is 32-bit integer there. The floating point version is, as you point out, 24-bit (thus that bit depth's dynamic range) plus a mantissa to reduce rounding errors.

Something in the file format might be quite different - perhaps it is saved as 32-bit integer format to be re-absorbed interpreting the extra 8 bits as the mantissa upon loading? Dunno... but would like to!

Terry
Computer: Sweetwater CS400v7 Intel Core i7-10700K CPU @ 3.80GHz | 64Gigs RAM | Windows 11 Pro x64 |
MOTU 828 mk3 hybrid

DAWs & Live: MOTU Digital Performer 11.31 | Cantabile Performer 4
Keyboard Synths: Kawai K5000s, Korg Wavestation
Controllers: NI Komplete Kontrol S-88 Mk3 & S-49 Mk2; Maschine Mk3 & JAM;
Akai MPK249 & 225, Alesis QX49, Behringer BCF2000 & FCB1010
Rack Modules: Ensoniq ESQm, Yamaha TX81Z, Wavestation SR

Tutorials: https://youtube.com/@CreatorsMediaTools
EMRR
Posts: 1079
Joined: Sat Jan 07, 2006 11:17 am
Primary DAW OS: MacOS
Contact:

Re: 32 bit Float question for Magic Dave

Post by EMRR »

If you are bouncing stems or mixes, 32f is a good idea because of the lack of gain penalty should you change the master later, no dither that can land in an audible range later.

I think Dave addressed the question of when dither is and is not applied in DP last year, can't recall the specifics right now. It may still be an open question.

Since I usually mix analog, the 24 bit soundbites headed to a DA converter get a post fader 24 bit dither plug, and I am of the OPINION it sounds a bit smoother and better that way. That re-dithers for any processing or level changes.

I've been in the habit of mixing even analog back in at 32f, even though individual tracks come in at 24. 32f capture may not be doing anything for me like this, even though it would if it were a bounce.

Using the 16A interface, the code stream appears to be 24 bit, with Audio/MIDI describing 32bit float paths for in and out, so if you send a 32 bit word from DP the 16A receives it. The output trims are digital and use the DAC's internal trim. Does it apply dither of it's own? The DAC chip is described as 32 bit input and filter architecture, selectable roll-off freq, soft mute and volume control. The rolloff and volume choice imply dither, but it's not specified that I can find. It's possible I'm double-dithering.
Doug Williams
Electromagnetic Radiation Recorders
The Martha Bassett Show broadcast mixer
Tape Op issue 73

DP 11.31
Studio M1 Max OS12.7.3
MOTU 16A and Monitor 8
M1 Pro MBP for remotes and editing
danatkorg
Posts: 19
Joined: Sun Jul 03, 2005 10:01 pm
Primary DAW OS: Unspecified

Re: 32 bit Float question for Magic Dave

Post by danatkorg »

terrybritton wrote:
stubbsonic wrote:Just out of curiosity (and some ignorance thrown in for good measure)-- I'd like to understand the difference between RESOLUTION and GAIN where the bit format is concerned.

When DP is using a 32-float, does the factor of it being "float" mean that it is interpreting the 32 bits to have a higher resolution as opposed to having 1500 dB of gain?

To ask a different way, if I have 16 bit audio normalized and the same audio as 24 bit normalized, I expect it to look same on the meters, but the 24 bit file will just have more resolution. I don't know what has to happen behind the scenes with the DACs to make that so. When we have 32-float files, should we expect them to look different on the meters (and clip internally?)
My understanding is that in any format, 0DB is still all ones or "full-on", and the resolution goes in the "quieter than full-on" direction, which is why 24-bit is better for reverb tails and piano string fades, and we have all that room to use -12db as our peak value. (The quieter the content is, the fewer bits it uses.)

So, yes, there is more resolution, especially for the quiet sections which, in 16-bit, become very low bit "words", while staying "longer words" with 24-bit even at the same low "volume" (all things being relative to "full-on"). 32-bit is 24-bit with room left for even more fine-tuning between levels with the mantissa/decimal fractions added. So, that being the case, higher resolution is what you are getting there, too.
Terry
Important to note that we're comparing 24-bit *fixed-point* with 32-bit *floating point.* It's that last part - fixed-point vs floating-point - that makes all the difference. One way to think of things, though it's a bit simplified, would be that the range of 24-bit fixed-point lives right in the middle of the much broader 32-bit floating-point range. Floating point lets you adequately describe both much louder signals and much softer signals. So, while the comment below is correct for fixed point, floating point is different:
terrybritton wrote:My understanding is that in any format, 0DB is still all ones or "full-on", and the resolution goes in the "quieter than full-on" direction
User avatar
mikehalloran
Posts: 15133
Joined: Sun Jan 25, 2009 5:08 pm
Primary DAW OS: MacOS
Location: Sillie Con Valley

Re: 32 bit Float question for Magic Dave

Post by mikehalloran »

Shoshie, as I read about your distortion issues, one thing that struck me was that they went away after you rebooted. That points to not enough RAM. There are ways to free up some RAM during the process like running sudo purge in Terminal but nothing clears it like a reboot. I know you also found a bad stick.

After Dave's post, I'm convinced of it.
There is no good reason to record at 32 bit float...
That was my conclusion, too, no matter how much nonsense I read on Gearslutz when I tried to research the subject.
DP 11.31; 828mkII FW, micro lite, M4, MTP/AV USB Firmware 2.0.1
2023 Mac Studio M2 8TB, 192GB RAM, OS Sonoma 14.4, USB4 8TB external, M-Audio AIR 192|14, Mackie ProFxv3 6/10/12; 2012 MBPs Catalina, Mojave
IK-NI-Izotope-PSP-Garritan-Antares, LogicPro X, Finale 27.4, Dorico 5.2, Notion 6, Overture 5, TwistedWave, DSP-Q 5, SmartScore64 Pro, Toast 20 Pro
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: 32 bit Float question for Magic Dave

Post by Shooshie »

I don't know, Mike. I've got 40 GB RAM, all working and passed all the tests of Rember, the RAM tester. My files just aren't that big. Anywhere from 5 to 10 audio tracks, some automation on aux tracks, and a few compressors, limiters, and Altiverb.

I don't really know where things stand right now. Around the time that I upgraded to the latest build of DP, I had gotten my levels pretty well adjusted so that I wasn't getting any clipping. The newer build solved the intermittent problems, as far as I can tell. I haven't gone back and overloaded my busses to see if I'm still getting the distortion when that happens. Maybe it ALL was due to the previous DP build. I'll be able to test things more when I'm done with this album, in a few days I hope.

Shoosh
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
User avatar
Timeline
Posts: 4910
Joined: Tue Nov 09, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Fort Atkinson Hebron, Wisconsin...
Contact:

Re: 32 bit Float question for Magic Dave

Post by Timeline »

With the latest build of DP I have not had this issue but I use the precision limiter from UA as a final PL. I wonder though, UA automatically sets a -.1 into the ceiling when inserting into a master or channel fader. Could be they would also have a comment or two about 32BF. I have seen the problem pop up curiously in the past BTW and I think I remember that it was intermittent. Thanks Dave for that input BTW.

Cheers,
GB
2009 Intel 12 core 3.46, 64GB, OSX.10.14.6, Mojave, DP11, MTPAV, Key-station 49,(2) RME FF800,
DA-3000 DSF-5.6mhz, Mackie Control. Hofa DDP Pro, FB@ http://www.facebook.com/garybrandt2
David Polich
Posts: 4827
Joined: Tue Dec 07, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Los Angeles, CA
Contact:

Re: 32 bit Float question for Magic Dave

Post by David Polich »

I found this lengthy discussion reminded me of a
recent exchange I had with someone who
sent me some stems to cut a new vocal over. The
exchange went something like this (condensed
here, of course):

YOUNG SONGWRITER: Ok I sent the stems. I
record at 32 bits. Can you please record the vocals
at 32 bits?

ME: 32 bits? That is your audio format?
The stems you sent are 24-bit 44.1khz WAV.

YS: Yes I record at 32 bits.

ME: Wait a minute. How old is your DAW?

YS: Way old, like 6 years.

ME: I think you mean your DAW is a 32-bit
application.

YS: Let me check...yes, it is.
2019 Mac Pro 8-core, 32GB RAM, Mac OS Ventura, MIDI Express 128, Apogee Duet 3, DP 11.2x, Waves, Slate , Izotope, UAD, Amplitube 5, Tonex, Spectrasonics, Native Instruments, Pianoteq, Soniccouture, Arturia, Amplesound, Acustica, Reason Objekt, Plasmonic, Vital, Cherry Audio, Toontrack, BFD, Yamaha Motif XF6, Yamaha Montage M6, Korg Kronos X61, Alesis Ion,Sequential Prophet 6, Sequential OB-6, Hammond XK5, Yamaha Disklavier MK 3 piano.
http://www.davepolich.com
User avatar
stubbsonic
Posts: 4600
Joined: Fri Dec 22, 2006 12:56 pm
Primary DAW OS: MacOS
Contact:

Re: 32 bit Float question for Magic Dave

Post by stubbsonic »

Hilarious!! :lol:
M1 MBP; OS 12, FF800, DP 11.3, Kontakt 7, Reaktor 6, PC3K7, K2661S, iPad6, Godin XTSA, Two Ibanez 5 string basses (1 fretted, 1 fretless), FM3, SY-1000, etc.

http://www.jonstubbsmusic.com
User avatar
Shooshie
Posts: 19820
Joined: Sat Oct 16, 2004 10:01 pm
Primary DAW OS: MacOS
Location: Dallas
Contact:

Re: 32 bit Float question for Magic Dave

Post by Shooshie »

David Polich wrote:I found this lengthy discussion reminded me of a
recent exchange I had with someone who
sent me some stems to cut a new vocal over. The
exchange went something like this (condensed
here, of course):

YOUNG SONGWRITER: Ok I sent the stems. I
record at 32 bits. Can you please record the vocals
at 32 bits?

ME: 32 bits? That is your audio format?
The stems you sent are 24-bit 44.1khz WAV.

YS: Yes I record at 32 bits.
Heck... I'm only recording at 64 bits now!
:koolaid:
|l| OS X 10.12.6 |l| DP 10.0 |l| 2.4 GHz 12-Core MacPro Mid-2012 |l| 40GB RAM |l| Mach5.3 |l| Waves 9.x |l| Altiverb |l| Ivory 2 New York Steinway |l| Wallander WIVI 2.30 Winds, Brass, Saxes |l| Garritan Aria |l| VSL 5.3.1 and VSL Pro 2.3.1 |l| Yamaha WX-5 MIDI Wind Controller |l| Roland FC-300 |l|
Post Reply