32 bit mixing?

[psycho_monkey]

Is there anyone actually working at 192k?!

I think you're confused as to the point of the 32bit floating point file format.

[PTbreaker]

care to elaborate?

[Lights]

I always mix in 32 float. For the most part you don't have to worry about gain (high or low) until the output of the 2bus. Much more forgiving as you say.

The only exception is you may get bad results overloading plugins. But in practice I find you hear that right away.

[rcn]

Never had any problems regarding gain structures mixing in tdm, there's so much headroom it's stupid..!!!

[PTbreaker]

Originally Posted by Lights

I always mix in 32 float. For the most part you don't have to worry about gain (high or low) until the output of the 2bus. Much more forgiving as you say.

The only exception is you may get bad results overloading plugins. But in practice I find you hear that right away.

Cool Thanks for the feedback it's new to me so I havnt really incorporated it into my workflow yet and was wondering if anyone out there had.

For sure about the plugins. Bad gain structure is something to be watched out for. I still catch classmates of mine tracking so hot that they have no headroom left on their tracks then not understanding why their EQ (or whatever) plugs clip.

RNC: your right compared to the older internal mix bus in native systems TMD systems have stupid amounts of headroom. As of right now we are in this weird transitional stage where a native PT10 system has more internal headroom than a TMD system.

[psycho_monkey]

Originally Posted by Lights

I always mix in 32 float. For the most part you don't have to worry about gain (high or low) until the output of the 2bus. Much more forgiving as you say.

The only exception is you may get bad results overloading plugins. But in practice I find you hear that right away.

Native systems have been 32bit float internally for years, and now many are 64bit float. You still only get a 24bit file out though. Unless you're in the habit of doing a really significant amount of offline processing, there's not much point working in 32 bit float file formats.

[Lights]

Originally Posted by psycho_monkey

Native systems have been 32bit float internally for years, and now many are 64bit float. You still only get a 24bit file out though. Unless you're in the habit of doing a really significant amount of offline processing, there's not much point working in 32 bit float file formats.

That's not true for Cubase. The WAV files in the project are 32bit float. Even the output WAV is 32bit float (which both Wavelab and Sound Forge can load and save as well). So theoretically, even if you had overs in your 2ch mixdown, they aren't going to cause clipping and you could manage them in the mastering phase.

Steinberg officially recommends using 32bit float as the project format in their tutorials. And I can absolutely verify that when you open a project folder all the WAVs in it are in 32bit format.

[PTbreaker]

Originally Posted by Lights

That's not true for Cubase. The WAV files in the project are 32bit float. Even the output WAV is 32bit float (which both Wavelab and Sound Forge can load and save as well). So theoretically, even if you had overs in your 2ch mixdown, they aren't going to cause clipping and you could manage them in the mastering phase.

Steinberg officially recommends using 32bit float as the project format in their tutorials. And I can absolutely verify that when you open a project folder all the WAVs in it are in 32bit format.

For sure Lights but as we are in the PT10 HDX sub-forum it is assumed we are all pro tools nuts.

This conversation is confusing for those outside of the pro tools world as there are actually two conversations going on here.

The first is the do you up-convert your wav files to 32bit for the benefits that can provide? See here for basic reference Pro Tools 10: The Benefits of 32 Bit Floating Point Audio - Pro Tools - macProVideo.com Hub

And the second, where people get lost, surrounds the internal mix architecture in PT. In PT 8 for example you had two different internal bit rates. one for le (someone remind me what it is) and the 48bit internal mix architecture of TDM.The new PT10 internal architecture is running 64bit math, even though the program is still a 32bit application. But if you have PT10 and are running HD with TDM hardware your internal mix bus is stuck at 48bit due to the hardware. It's time to go HDX anyhow.

[Lights]

Originally Posted by PTbreaker

For sure Lights but as we are in the PT10 HDX sub-forum it is assumed we are all pro tools nuts.

Ah ha! I never actually search by forum, I just click the "New Posts" button so I didn't even realize it was in the protool subforum.

[psycho_monkey]

Originally Posted by Lights

That's not true for Cubase. The WAV files in the project are 32bit float. Even the output WAV is 32bit float (which both Wavelab and Sound Forge can load and save as well). So theoretically, even if you had overs in your 2ch mixdown, they aren't going to cause clipping and you could manage them in the mastering phase.

Steinberg officially recommends using 32bit float as the project format in their tutorials. And I can absolutely verify that when you open a project folder all the WAVs in it are in 32bit format.

I mean you only have a 24bit converter, so you can have all the 32bit files you like, you only get the 24bit output.

It's only worth saving a 32bit file for mastering if you've gone wrong in your gain staging in your mix - if you've left headroom (or even if you've just made sure you're not clipping your mix buss) you don't need to save out a 32bit float file.

I guess it's ok as a further safety protection, but hopefully you shouldn't need it.

[Lights]

Originally Posted by psycho_monkey

I mean you only have a 24bit converter, so you can have all the 32bit files you like, you only get the 24bit output.

It's only worth saving a 32bit file for mastering if you've gone wrong in your gain staging in your mix - if you've left headroom (or even if you've just made sure you're not clipping your mix buss) you don't need to save out a 32bit float file.

I guess it's ok as a further safety protection, but hopefully you shouldn't need it.

The benefit of 32bit mixing is that during the mixing process you don't have to worry about gain, gain staging, etc. All you have to do is think about whether the channels and the mix sounds good to you. That is a benefit even if your A->D is 24bit. I don't know protools, but that is certainly true for Cubase.

As for the mixdown bit-depth, if you're already working in 32bit it's probably better not to downsample twice--once to 24 and once to 16 (for CD/MP3), and just output 32 and then do your downsampling/dithering when you master.

[psycho_monkey]

Originally Posted by Lights

The benefit of 32bit mixing is that during the mixing process you don't have to worry about gain, gain staging, etc. All you have to do is think about whether the channels and the mix sounds good to you. That is a benefit even if your A->D is 24bit. I don't know protools, but that is certainly true for Cubase.

As for the mixdown bit-depth, if you're already working in 32bit it's probably better not to downsample twice--once to 24 and once to 16 (for CD/MP3), and just output 32 and then do your downsampling/dithering when you master.

That's not really how it works...all DAWs are at least 32bit float in their processing, so all those benefits are inherent, all that happens is that you've got files 50% bigger. As I've said, the only significant difference could possibly be if you're doing a lot of processing in place - and even then, with good gain staging you've not got a problem.

As for your 2nd point, that would imply that the mix as you hear it "live" through your 24bit converters is different to what you supply to your ME. No thanks!

[UnderTow]

Originally Posted by Lights

The benefit of 32bit mixing is that during the mixing process you don't have to worry about gain, gain staging, etc.

Actually you do. Some plugins clip at 0 dB FS. (The Waves linear phase series comes to mind). And some plugins have weird artefacts above 0dB so it is best to always pay attention to your gain staging. (It isn't exactly hard...)

As for the mixdown bit-depth, if you're already working in 32bit it's probably better not to downsample twice--once to 24 and once to 16 (for CD/MP3), and just output 32 and then do your downsampling/dithering when you master.

There is no "downsampling" involved. 32 bit float has the same resolution as 24 bit fixed audio if you use sensible levels. You can take a 24 bit fixed point wave file, convert it to 32 bit float and back to 24 bit and the result will be bit for bit identical to the original.

Psycho also makes good points about only ever hearing 24 bits through your converters. If what you hear is good, there is no reason not to save/export/record/render at 24 bits. Using 32 bit float for files is a waste of disk space and puts extra strain on your system bus and memory. Not a good idea.

I have all my DAWs set to record and bounce at 24 bits. (That includes PT and Cubase). This is separate from the processing/mixing bit depth of the DAWs which is 32 or 64 bit float (for native DAWs).

Alistair

[Lights]

In my first post I noted that some plugins may by their design clip if your levels are too high, but this, I believe, is a different kind of clipping because if you set your project bit depth to 32bit float, you're still not getting what we know as "digital clipping" from them so if it sounds ok... it is. That's my understanding anyway.

Given what you say, Alstair, I wonder why Steinberg's own mixing tutorials have a whole section recommending 32bit float for all projects? They point out the points I was parroting above as the reason. Again, I can't speak to ProTools... so sorry for intruding on the thread unwittingly by seeing it in the "New Posts" list I use instead of looking at forums independently