32 bit bounce

archtoper · Post by **archtoper** » Tue Dec 27, 2005 6:42 am

I decided to bounce my mixes at 32 bits since DP manual says all the internal processing is done at 32 bits, and because every book on dither/mastering says dither once and only once at the very end. If I bounced to 24 bits I'd have to use DP's dither (which is what I've been doing up till now). Anyway, I had a problem with my 32 bit bounces when I tried aiff or wave files (they were totally fried-digital hash), but when I bounced as SDII files they come through perfectly fine. I'm opening the files in Peak-5, BTW, and doing the mastering there.

I can't say yet for sure if the 32 bit files sound any better since I'm not done mastering, so I'm wondering two things: (1) anyone here been down this path and have an opinion of on 32 bit vs. 24 bit bounces, and (2) anyone have any idea why SDII works when wave and aiff wouldn't?

Thanks
www.markkleinhaut.com

giles117 · Post by **giles117** » Tue Dec 27, 2005 7:28 am

HAdnt bothered but this next song I mix I will do 3 bounces, 1 16, 124 and 1-32 to see what I think.....

archtoper · Post by **archtoper** » Wed Dec 28, 2005 11:04 am

Did you try the 32 bit bounce? I'm listening to my test master CD made from 32 bit SDII files (dithered to 16 bit aiff) using OZONE 3 (in Peak-5) and I swear it sounds better.

David Polich · Post by **David Polich** » Wed Dec 28, 2005 8:06 pm

Ya know, I guess I'm not getting this. If your project is at 24bit, what difference would bouncing out at 32 bit make?

Trying to do the math here...

kelldammit · Post by **kelldammit** » Wed Dec 28, 2005 9:49 pm

David Polich wrote:Ya know, I guess I'm not getting this. If your project is at 24bit, what difference would bouncing out at 32 bit make?

Trying to do the math here...

because, he's trying to preserve as much of the original audio as possible. dithering should only be done once, and as a last step prior to burning to glass.
yes, the source files are recorded at 24 bit, but the mix engine is operating at 32 float. so take a track, bump it up one dB (a multiplication of 1.xxxx out to umpteen decimal places in itself) add some eq and a reverb...and you're quite possibly into 32bit float land already (at least with the rounding errors, etc).
so at best, if he doesn't dither, and just exports to 24 bit, he gets truncation at 24 bit. if he dithers, he gets 22-23 (or less) bits of audio, plus dithering noise, and now that the last bits are dithered, they're technically errors (it's what dithering does...try to make more tolerable-sounding rounding errors...often by masking/shaping them with noise).
the mastering engineer then has to take this, make any changes he has to, thus adding more errors, and then has to dither again, adding more noise/errors...you can see it's not really the ideal.

but if the audio's exported at 32bit float, he gets good relatively undamaged 24 bit audio, with the artifact noise, etc down below 24 bit (in theory). the mastering engineer then works his magic at 32float, so his errors, etc are still down there, not harming the 24 bit level signal.
when finished, he performs the dithering, hopefully thusly introducing only one round of "audible"(i.e. falling within the bit depth of the final product) errors.

kell

Timeline · Post by **Timeline** » Wed Dec 28, 2005 10:09 pm

I agree kelldammit although it is prolonging the inevitable so that a mastering house with a gazillion bucks worth of blessed spiritual audio gear can sanctify your mix by reducing your bank account significantly.

chrispick · Post by **chrispick** » Wed Dec 28, 2005 10:48 pm

kelldammit wrote:
David Polich wrote:Ya know, I guess I'm not getting this. If your project is at 24bit, what difference would bouncing out at 32 bit make?

Trying to do the math here...
because, he's trying to preserve as much of the original audio as possible. dithering should only be done once, and as a last step prior to burning to glass.
yes, the source files are recorded at 24 bit, but the mix engine is operating at 32 float. so take a track, bump it up one dB (a multiplication of 1.xxxx out to umpteen decimal places in itself) add some eq and a reverb...and you're quite possibly into 32bit float land already (at least with the rounding errors, etc).
so at best, if he doesn't dither, and just exports to 24 bit, he gets truncation at 24 bit. if he dithers, he gets 22-23 (or less) bits of audio, plus dithering noise, and now that the last bits are dithered, they're technically errors (it's what dithering does...try to make more tolerable-sounding rounding errors...often by masking/shaping them with noise).
the mastering engineer then has to take this, make any changes he has to, thus adding more errors, and then has to dither again, adding more noise/errors...you can see it's not really the ideal.

but if the audio's exported at 32bit float, he gets good relatively undamaged 24 bit audio, with the artifact noise, etc down below 24 bit (in theory). the mastering engineer then works his magic at 32float, so his errors, etc are still down there, not harming the 24 bit level signal.
when finished, he performs the dithering, hopefully thusly introducing only one round of "audible"(i.e. falling within the bit depth of the final product) errors.

kell

Interesting. This is a concept I'd been pondering for a little while now, and was planning to test this weekend. My only difference is I'll be conducting mastering myself using Peak, pow-R dithering, UAD-1 Precision plugs and a slight touch of Ozone3. Oh, and a less-than-pristine monitoring environment (which might render all efforts moot or close to it).

My logic is the same as above. I don't know if it's really valid since I only have a basic understanding of 32-bit calculation and its effects, but I wanted to find out what my ears would tell me.

I guess I'll post my impressions next week, should anyone care to see.

sdfalk · Post by **sdfalk** » Wed Dec 28, 2005 11:46 pm

I guess I'll post my impressions next week, should anyone care to see.

I would.

qo · Post by qo » Thu Dec 29, 2005 2:38 am

Me too!

archtoper · Post by **archtoper** » Thu Dec 29, 2005 6:18 am

kelldammit wrote:because, he's trying to preserve as much of the original audio as possible. dithering should only be done once, and as a last step prior to burning to glass.
yes, the source files are recorded at 24 bit, but the mix engine is operating at 32 float. so take a track, bump it up one dB (a multiplication of 1.xxxx out to umpteen decimal places in itself) add some eq and a reverb...and you're quite possibly into 32bit float land already (at least with the rounding errors, etc).
so at best, if he doesn't dither, and just exports to 24 bit, he gets truncation at 24 bit. if he dithers, he gets 22-23 (or less) bits of audio, plus dithering noise, and now that the last bits are dithered, they're technically errors (it's what dithering does...try to make more tolerable-sounding rounding errors...often by masking/shaping them with noise).
the mastering engineer then has to take this, make any changes he has to, thus adding more errors, and then has to dither again, adding more noise/errors...you can see it's not really the ideal.

but if the audio's exported at 32bit float, he gets good relatively undamaged 24 bit audio, with the artifact noise, etc down below 24 bit (in theory). the mastering engineer then works his magic at 32float, so his errors, etc are still down there, not harming the 24 bit level signal.
when finished, he performs the dithering, hopefully thusly introducing only one round of "audible"(i.e. falling within the bit depth of the final product) errors.

kell

Kell, right on...that's exactly what I'm talking about. The only thing I'd mention is that the export function i'm using is 32 bit fixed not floating point. I don't know if this matters. I'm still wondering why the wave and aiff formats for 32 bit are coming out garbled while SDII workes perfectly. Maybecause SDII is DP's native tounge? Just a guess. Anyone else going there? Anyone else bothering with 32 bit bounce. I gotta say- the results are REALLY worth it as I've had more time to listen carefully.

TOD · Post by **TOD** » Thu Dec 29, 2005 8:41 am

It's funny, I've always done my projects at 24/44.1 and then bounce to 16/44.1. I don't use a dither plugin on the master channel. DP just dithers automatically during the bounce process and I've been very happy with the results.

kelldammit · Post by **kelldammit** » Thu Dec 29, 2005 8:54 am

32 bit fixed is even better than float, especially in this case. you wouldn't have to worry so much about as many float rounding errors piling up throughout the process...

yeah, "pro mastering" does come at a cost...and for most of us, the $ just doesn't allow. the end result may or may not be all THAT much better, strictly mathematically speaking, but then again, odds are a mastering engineer has tools geared specifically to that task...and hopefully plenty of experience insofar as how to use them...whether the corresponding bank account drain is really worth it...well...

also, i'd suspect that a lot of the "improvements" would be fairly subtle...i mean, in theory you should only really be seeing improvement at the very bottom of the dynamic range...

kell

Kubi · Post by **Kubi** » Thu Dec 29, 2005 10:40 am

32bit SDII, well I'll be. Didn't even know this was an option!

The "garbled" aiff and wav files may be due to a format restriction. aiff and wav may simply be unable to hold 32-bit words. Or else DP writes garbage when set to that setting and it's a bug. Or else the files are fine, but DP is unable to read 32bit aiff or wav files. But obviously I don't know...

I'm actually surprised SDII is able to carry 32-bit info, but I'm not surprised it's fixed, not floating. The only file formats I'm aware of that can carry 32-bit floating information are the likes of SUN and NeXT (I use this option in Soundhack, since it allows for a virtually unlimited dynmic range and thus makes sure my esoteric process doesn't just result in "all full level" samples that blow my speaker to bits...)

I always turn my dither in DP off (I print IDR via a plug-in when needed instead.) AFAIK, it only kicks in when you export from the soundbite window, and not when you print your 32bit floating into a 24bit fixed audio track. I have a 16-month old e-mail exchange with MOTU on this topic somewhere in my inbox, where they enlightened me in detail on when exactly DP uses dither (if checked) and when not... I really should re-read that.

I'd really be curious as to MOTU's official recommendation on using 32bit SDII for bounce, and also (another?) clarification on whether the dither option affects BTD/Freeze/Record when merely printing a 24bit track inside a 24bit project.

Magic Dave, care to elaborate?

Kubi · Post by **Kubi** » Thu Dec 29, 2005 10:48 am

PS: Silly question:

Are the resulting files really 32-bit SDII's? Call me paranoid (which I am), but it may be a good idea to open them in Peak or some other program and read the detailed file info.

I should really try some of this myself...

archtoper · Post by **archtoper** » Thu Dec 29, 2005 11:28 am

Kubi wrote:PS: Silly question:

Are the resulting files really 32-bit SDII's? Call me paranoid (which I am), but it may be a good idea to open them in Peak or some other program and read the detailed file info.

I should really try some of this myself...

That's what I'm using them in- Peak-5. Not at that computer now but I'm pretty sure Peak sees them as SDII 32 bit raw files. I don't think DP could open these files. I do use DP's dither when freezing a track, and actually, it's on all the time while mixing so that my MOTU 896HD will see 24 bit imput. I mean, if DP is doing 32 bit internal and you output to a 24 bit DA you'd get the dreaded truncation otherwise. Dither thou goest:)