16bit to 24bit?

[johnportrash]

Doing a project at someone elses house. They have 8x 16bit analog inputs on thier soundcard. Drag huh? Anyway, after i record these files (drum takes)I was going to convert them up to 24bit before processing and mixing. And of course, the last step would be to dither entire mixdown back to 16bit. Is this a waste of time? Will converting up to 24bit just add 8bits of nothing and not benefit further processing? Thank you in advance.

[Plec81]

It would be useless to convert the files into 24 bit files since you would get 8 empty bits in the bottom of the file itself.

The thing that you must understand is that when working on a DAW like Cubase, Nuendo, Protools or other system, your mix is already being processed at 32 bit float point resolution (48 bit for ProTools). This is because the system itself works at that resolution. The mixer, eq, kompressors, all your plug-ins etc.. They all work att 32 bit float or higher. Let's say that you have recorded 10 tracks at 16 bits. After you've mixed the project you might be mixing it down to 16 bits. If you're not dithering, this will truncate (cut off) the last 8 bits of your mix which results in distorsion.

Even if you record something at 8 bits your mix will be 32 bit float in the end (for native systems).

[johnportrash]

Thank you for that information. It is perfectly logical now. It leaves me wondering if my multitrack(Vegas- sonic foundry) works on a 32bit float or not. And if my editor (sound forge) works on a 32 bit float. I will proably have an answer to this within a couple of clicks, but if you have an idea please let me know.

[Plec81]

Some editors does work at 16 bits. But then all you can do is just that.. EDITING. If you're going to apply effects or just change the gain by 0,2db your system must work at 32 bit FP or higher in order to handle all the calculations.

I think that all Sonic Foundry software is 32 bit FP systems. I'm not one hundred % sure though.

[jecahn]

Most software processes at 32 bit, now. Most did before, too. It was only recently when you became able to record usably at 32 bit that they actually started advertising "32 BIT!!!"

[juppu]

...and 56 bits for Pro Tools master faders, if I don't remember wrong.
[Luckily Digidesign has the decency not to advertise 56-BIT RECORDING!!!] http://www.audioforums.com/forums/biggrin.gif

Juppu

[Sonic Valley]

Yup SF6 is 32....Well it's late...gonna 32bit float myself off to the sack...lol

[Plec81]

Juppu: Interesting! Are you saying that ProTools is handing off 56 bits on the master faders??

[juppu]

Yeah, the internal resolution is 48 bits, but the master fader has an extra 8 bits of headroom.

For more information on the subject, check out http://duc.digidesign.com/cgi-bin/ubbcgi/ultimatebb.cgi?ubb=get_topic;f=2;t=005504 There is 510 replies, so you'll have to do some serious reading...

So here's the key bit from DUC:

"...fader values are calculated at 48bit and passed on to the mix buss without truncation. So the mix buss is 48bits wide. However, there is also an additional 8bit overflow built into the mix buss to provide a little extra headroom when summing 48bit values. So in total the mix buss has 56bit resolution."

http://www.audioforums.com/forums/smile.gif

Juppu

[jecahn]

That's weird. I'm not saying that you're wrong. But, if you're already using 32bit FP processing notation, shouldn't that (in theory) give you an infinite (for practical purposes) amount of headroom? Doesn't make a lot of sense. Maybe my info is wrong.

[Plec81]

No. The more bits your processing use, the more accurate the outcome will be. It will sound slightly better to process at 48 bits instead of 32 bit FP.

Thanks for the info juppu. The thing I'm concerned about is the fact that the ProTools environment hands off a 24 bit word between your internal TDM processors. I don't know if this has been changed in the HD version.

Does anyone know how ProTools manages the 56 bits on the masterfaders into a 24 bit wordlength? Is it a built in dither, is it rounding or is it just truncating the output?

[jecahn]

48 bit uses exponential notation, then?

[Plec81]

I think that pro tools use 48 bit fixed point, but not sure..

[juppu]

The extra 8 bits are added for extra headroom to prevent clipping when you add 32-128 (and that's just the voices...) 48-bit values together.

Not sure how it handles it dithering/truncating/rounding-wise...

If you're interested, check out the thread in DUC I pointed out, despite the HUUUUUGE amount of replies, there is some good info there.

Juppu