worth transferring 16 bit tracks to do 24 bit mix

[stephentayler]

.... yes but if you add some flavourings to the water, it's much more effective to shake them up and mix them in the 2 litre bottle!!

Cheers

Stephen

[Kubi]

He he. Good answer, Stephen!

@mirmidon: If you play back only a single track, and only hit play, without blending in another track, without changing levels, applying EQ, compression, reverb, or processing in any way, then the 1 liter/2 liter example applies. You are absolutely correct that playing back a 16bit track in a 24bit environment (without doing anything else to it) will sound exactly the same as playing back a 16bit track in a 16bit environment (without doing anything else to it.)

BUT: As soon as you do as much as apply a fade, or a level change to a single track, things change. Also, as soon as you blend even two tracks at unity level, things change.

Now, most mixes involve a lot more than just blending two tracks at unity gain. They involve dozens if not hundreds of tracks, with lots of EQ, compression, delays, processing, fades, crossfades, etc. etc. etc., resulting in tons of acoustical interactions between elements, and between elements and processes.

At what quality do you want all this processing to take place? Sixteen bit allows for 65536 values to describe the level of a sound wave at each sample. 24bit allows for over 16 million values to describe the level of a sound wave at each sample.

[gaucho]

got it, kubi
I always wanted that clarified a bit

[mirmidon]

@ Stephen > lol (very good answer!)

@ Kubi > Yes of course, I understand. Thank you for your answer.

[mikehalloran]

At what quality do you want all this processing to take place? Sixteen bit allows for 65536 values to describe the level of a sound wave at each sample. 24bit allows for over 16 million values to describe the level of a sound wave at each sample.

Although that answer looks great on paper - and it does look great, BTW, there are many who disagree on whether it means anything when 16bit and 24bit are mixed in the real world.

[Kubi]

The way I put it may have been clumsy or not very clear, but there's no question that once a 16 bit source file gets processed further -- whether with EQ, compression, distortion, level changes, or any number of processes common to mixing -- it makes a huge difference whether that processing is done inside a 24bit project, or whether it's done inside a 16bit project.

Because once that file gets processed, the data immediately seizes to be 16bits deep - it becomes 32bit float, so 24bits deep with a sliding 8bits for dynamic range. And then the result of the process gets resolved to the project bit depth. And that's where the extra 8 bits make a huge difference, the difference between 65000 and 16 million. That's my only point.

Of course, as you pointed out, with DP allowing for varying bit depths inside the same project, the source file itself does not have to be converted to 24bits, since the source file itself doesn't contain any information in those last 8 bits.

[TnMike]

I dragged 6 songs (44.1/16bit) from an audio CD into an empty DP file. The DP file format was set to 44.1/24 bit. I put Ozone 3 on each track to add some volume and EQ. I exported each track by BTD. I did not dither. Should I have?
Thanks, TnMike


G5 2.7 GHZ, DP 7.24, MOTU 2408, Ivory, SD2, B4ll, UAD-1, Melodyne, Ramza DA-7,

[Dan Worley]

I dragged 6 songs (44.1/16bit) from an audio CD into an empty DP file. The DP file format was set to 44.1/24 bit. I put Ozone 3 on each track to add some volume and EQ. I exported each track by BTD. I did not dither. Should I have?
Thanks, TnMike

Technically, yes. But in reality, don't sweat it. Not in the case you explained. No one hearing it will ever know or be able to tell those files were truncated.

[Prime Mover]

I see some misconceptions, even among people giving the answers.

DP, as an environment, runs at 32bit float, no matter what setting you have it at. Fades, automation, processing... all done at 32bit float. So Kubi, I'm gonna call you on the fader stair-stepping assumption. Upping a project to 24bit won't make any difference in how the effects handle incoming audio. What you CAN change is how audio is recorded at by default. If you set it to 16bit, that means that an armed audio track will record a 16bit audio sound file. However, it will still run through all of its automation and effects at 32bit float after the fact. Similarly, at bounce down, you choose what quantization (if any) the engine outputs to. But everything between tracking to final bounce is 32bit.

So my answer to the original question: if you're completely done with tracking, then setting the bit rate won't make a lick of difference.

[TnMike]

Thanks you guys!


G5 2.7 GHZ, DP 7.24, MOTU 2408, Ivory, SD2, B4ll, UAD-1, Melodyne, Ramza DA-7,

[Phil O]

When in doubt, test it.

So here's what I did:
Test 1: Three 16 bit files (400Hz tone, 800Hz tone, 1600Hz tone) in a 16 bit project. Schwa's Bitter bitmeter on Master fader (post fader). With all faders set to 0dB the bitmeter shows 16 bits. Moving any of the faders results in over 16 bits showing on the bitmeter.

Test 2: Same setup, but with a 24 bit project. Same results.

Test 3: Same setup, but with a 32 bit floating point project. Same results.

Test 4: Same setup as test 1 (16bit project), but 1600Hz file at 24 bit resolution. With all faders at 0dB, bit meter displays 24 bits.

BTW, toggling Dither in the audio menu did not change the results of any test.

So it seems that changing the project's bit depth does not affect the OUTPUT resolution. It seems this only affects any audio FILES created (i.e., tracking, destructive edits, bouncing, etc.)

But there is one thing that bother's me, and perhaps someone who understands the math better than I do can explain. Watching the bitmeter, DP's output appears to be 32 bit floating point. As I move the master fader up and down, I see a range of about 24dB on the bitmeter sliding up and down the meter's scale. So, for example, I see bits 1 through 24 displayed, or bits 9 through 32, or 14 through 37 - which is what I'd expect. But how does DP then send this to a 24 bit DAC? Truncated? So it sends 1 through 24, 9 through 24,
14 through 24, etc.? Or something else? And if it does truncate, does the project bit depth affect THAT truncation which would be post, post fader, and would not be seen on the bitmeter.


Phil

[Prime Mover]

Interesting questions. I'll have to go back and check my Bob Katz, but I'm curious if there *IS* such a thing as a 24bit output DAC. It strikes me that since all commercial recordings eventually are processed to 16bit, that audio output is ALL truncated to 16bit. The only situations I could see needing it is when sending audio to external hardware processors. Since fewer and fewer people are doing that these days, interface manufacturers may very well be only putting 16bit converters on their main outs... who knows, maybe the mains are 16bit, but sends are 24.

By this point in the process, it's hard to know at what level these operations are being done: DP? Core Audio? The interface? At some point, the DAW says "good bye" to the signal and lets lower level subroutines take over. If there is output truncation, is it the DAWs job, or is it further along?

In any case, I kind of doubt that the bitrate setting in DP has any bearing on the output resolution, I'm not sure what purpose that would serve. After all, the reasons for changing the input resolution would really have nothing to do with the reasons for wanting different output resolutions. Most likely this is something arbitrated behind the scenes between DP, the OS, and the Interface.

Then there's freezing... oh boy... that's a weird one to try to fit into all of this. Freezing is such a bizarre process, it's hard to know what's really going on. There's always something a little eerie about freezing because it seems to be using parts of the program you wouldn't normally expect to cooperate. It seems to be using the bounce down process (or maybe it's just simply the bounce down GUI), but actually doing an internal record off the bounce down! This wouldn't be so strange if it weren't for the fact that it ALSO prints to tracks in realtime, and you can even see them being made. So some recording procedures seem to be working at the same time as bounce down. That's a completely different data pipe than anything else we know about. Who knows where this stream is coming from, does it go all the way out to the OS level and loop back around? Does it print to audio bitmap files in realtime and DP is simply accessing the file as it is being created on the disk? Whatever it is, I bet it's pretty complex and probably took a lot of work to setup.

[Phil O]

...I'm curious if there *IS* such a thing as a 24bit output DAC...

Well first of all its a DAC, so it's output is analog. Its INPUT is digital and your MOTU 828mkII has a 24bit DAC. The questions really are 1)Does that DAC see the full 24 bits when coming from a 16 bit project? and 2)What bits are sent from a 32bit floating output?
From the tests that I did, DP's output "appears" to always be 32bit float, but what actually gets sent to the DAC? i.e., are my tests a valid representation of the actual output?

Phil

[Dan Worley]

Hey, Phil. I don't have any bit meters that can measure above 24 bits. But I do have RME's Digicheck, so I'm able to measure after DP passes on the signal. No matter what the project bit-depth is set to, it reads 24 bits unless I purposely dither DP's outputs, then I can make it read 24, 20, 16, 8, or whatever.

I'm pretty sure (okay, it's a guess) DP only works internally in 32-bit float (and you can save files with that), but passes on only 24 bits unless purposely dithered down.

[Armageddon]

Since DP can mix 16-bit and 24-bit tracks inside of a 24-bit project, and since you pretty much don't gain anything by upconverting your bit rate (aside from some extra headroom), you're likely fine just where you're at. It's no more perilous than printing a MIDI track containing 16-bit samples being played through Kontakt in a 24-bit project. However, just to upset Dan Worley, I am gonna suggest upconverting your entire project to 32-bit.

All joking aside, the first time I mixed a score in DP 5.13, I wasn't aware that you could mix 16-bit and 24-bit files at the same time or, moreover, that you wouldn't get any alerts regarding this, like you did in previous versions or in other DAWs I'd used. I was bouncing down tracks to 96 kHz/24-bit, but inadvertently mixing and freezing audio at 16-bits. I didn't actually catch onto this until my very last cue, but since the rest of the score had been delivered and no one had complained, I decided to just re-record and remix the last cue 24-bits all the way through to hear what it sounded like. There was actually a shocking amount of difference -- a lot more than you'd hear between differing sample rates. These days, I actually do everything in 32-bits, but that's more for peace of mind (and the fact that I insanely subscribe to the "straight division" theory of quantization) than actual sound quality. 24-bits is probably plenty, especially if you have a decent quantization algorithm.