Bit Depth Question for mixdown...

[musikman1]

My goal is to try to get my final mix at 24 bit with no dithering, prior to using mastering software on that mix. The mastering software will apply dithering at the final stage, as it will be converted from 24 bit to 16 bit.  When I looked at the bit depth of all my audio tracks, some are 24 bit, and some are 16 bit.  My record bit depth in CWbB is set to 24, and I use drum loops that are mostly 24 bit, but sometimes I need to use some that are 16 bit.  Plus there are other tracks at 16 as well, I'd say it's about 60% 24bit, and 40% 16bit. 

If there is always going to be a mixture of 16 and 24 bit audio clips/tracks, can I still bounce everything down to a final mix of 24 bit without dithering?. ..   I'm trying to avoid applying dither more than once.  I'd rather have a 24 bit final unmastered mix to put into my mastering software, and then dither down to 16 bit at the final master export.   Is there any way to do this if my tracks are always a mixture of 16 and 24 bit?  Are my 16 bit tracks negatively affected if I bounce all tracks to 24 bit mixdown?

[scook]

Don't worry about the mix of different clip bit depths.

The DAW audio engine uses 32 or 64-bit floating point depending on the 64-bit Double Precision Engine setting.

By default, the DAW is setup to render 32-bit.

Why not let the DAW export at the default 32-bit depth?

 

 

 

[musikman1]

Scook, I don't know tons about the bit depth and dithering. But I did watch a few videos on it from another thread awhile back.  I seem to remember that going from a higher bit depth to a lower one....ie...24 to 16 it is recommended to use dithering.  Yesterday I realized that I had my Audio Data settings at Record = 24 and Render = 16, and I think I had changed that Render setting from 24 to 16 a while back prior to this project, and then forgot to put it back at 24.  So any tracks that got bounced automatically went to 16 bit, even if it was just a couple of guitar tracks and not a final entire mix bounce.  As a result I now have a mixture of 16 and 24 ( so that's how I got here)

The reason for my question is because I noticed on a project that I had mastered I had a sort of a glitch or stutter on one vocal part on the master but not on the unmastered mixdown.  The only thing I could think of that may have caused it was maybe using dither to export the mixdown to wav file, then again using dither in my mastering software may have been too much. I had asked someone knowledgeable who told me they wouldn't recommend using dithering twice like that, only on the final master output.  In this particular project I had bounced the mixture of 24 bit and 16 bit tracks down to a 16 bit mixdown track without dithering, then when I exported that mixdown to a 16 bit wav file I used dithering, then when I ran that through my mastering software I used dithering again upon exporting there. So I basically used dithering twice, once exporting the final mix out of CWbB to wav, then again exporting from the mastering software.

So am I to assume the best way to do this as far as a step-by-step is to Step 1..... first render/bounce all my tracks ( a mixture of 16 and 24 bit) down to one 24 bit track final mix without using dithering, then Step 2.... exporting that final 24 bit mixdown track to a 24 bit wav file without dithering, then Step 3.... put that 24 bit wav file in my mastering software and export using dithering to 16 bit 44,100?    (The end result if I want to say burn a CD is why the 16bit 44,100 final master)

 

Edited July 9, 2022 by musikman1

[scook]

Put the render bit depth back to the 32-bit default and leave it there.

Export using the default 32-bit.

Load the file in the mastering software and dither to 16-bit when exporting.

[musikman1]

Alright then, so the 32 bit setting covers everything I render without the need for dithering.  Thank you my friend.  Does that mean even if I just bounce a couple of guitar tracks to save space, just for example, that bounced/mixed guitar track is going to also be 32bit?...(or any other smaller groups of tracks I bounce for that matter?)

Ok, then NO dithering at all at any time in CWbB, and only use dithering one time in the mastering software export.  Got it.    I wonder why CWbB  Bounce to Tracks dialog always shows the dithering setting in the drop menu by default? 

[scook]

28 minutes ago, musikman1 said:

Does that mean even if I just bounce a couple of guitar tracks to save space, just for example, that bounced/mixed guitar track is going to also be 32bit?...(or any other smaller groups of tracks I bounce for that matter?)

yes

 

28 minutes ago, musikman1 said:

Ok, then NO dithering at all at any time in CWbB, and only use dithering one time in the mastering software export.  Got it.    I wonder why CWbB  Bounce to Tracks dialog always shows the dithering setting in the drop menu by default?

The dither setting is used only when needed. The DAW only dithers when necessary.

[Terry Kelley]

When is that?

[scook]

When rendering at a lower integer bit depth such as 32-bit floating point to 24-bit or 24-bit to 16-bit. Going from 64-bit to 32-bit does not dither.

[musikman1]

2 hours ago, scook said:

The dither setting is used only when needed. The DAW only dithers when necessary.

 

2 hours ago, scook said:

When rendering at a lower bit depth.

Interesting...So even if the setting in the Bounce to Tracks dialog box is set to dither, that function will automatically remain inactive unless it detects one or more tracks will be getting converted to a lower bit  depth?

According to what I've read, rendering at 32 bit will create a larger file size than rendering at 24 bit, but I would think the difference would be minimal, and probably would not outweigh the quality benefit that 32 bit would provide in comparison.  Related thread:

I have recently noticed there are also some plugins that also apply dither, like one of my limiters has two presets which include dithering, one preset is 16bit and the other is 24 bit.  I'll have to remember not to use those presets.

Scook, do you think that the way I had previously rendered and mastered, not using 32 bit and dithering twice could have caused that vocal stutter on the master that I had mentioned? 

 

Edited July 10, 2022 by musikman1

[scook]

45 minutes ago, musikman1 said:

Interesting...So even if the setting in the Bounce to Tracks dialog box is set to dither, that function will automatically remain inactive unless it detects one or more tracks will be getting converted to a lower bit  depth?

yes

45 minutes ago, musikman1 said:

According to what I've read, rendering at 32 bit will create a larger file size than rendering at 24 bit, but I would think the difference would be minimal, and probably would not outweigh the quality benefit that 32 bit would provide in comparison.

Size should not matter. It's a temporary file for mastering.

45 minutes ago, musikman1 said:

Scook, do you think that the way I had previously rendered and mastered, not using 32 bit and dithering twice could have caused that vocal stutter on the master that I had mentioned? 

Doubt it. Dither adds noise to mask any artifacts created by bit depth reduction.

[scook]

Here are a couple of posts from the old forum where Noel goes into more detail about the audio engine, bit depth and dither.

From http://forum.cakewalk.com/64bit-Audio-engine-vs-double-precision-engine-and-a-dithering-question-Noel-please-m3804976.aspx

Quote

The 64 bit mix engine does all mixing / summing math in double precision 64 bit. double precision has 53 bits of precision and 11 bits of exponent. See https://en.wikipedia.org/wiki/Double-precision_floating-point_format
SONAR and CbB are identical.
 
Dithering is performed at the final stage before output is sent to the driver. Dithering is an option in preferences but its on by default...

Quote

if the 64 bit mix engine is off in the audio settings then all mixing is done in 32 bit floating point irrespective of the audio word length in the track data. The mix engine setting is controlling how mixing is done and all audio data is up/down converted to the mix engine bit depth prior to mixing.

and http://forum.cakewalk.com/Bits-Freezing-and-Dithering-m1204271.aspx#1204822

Quote

1. Render bit depth defaults to 32 bits which will capture the result of bounces, freezes and other intermediate processing to a floatint point wave file. i.e. there will be no integer truncation loss so no is dithering required. (you couild even set it to 64 bit but thats overkill for most scenarios)

2. When you freeze it is converted to the specified render bit depth irrespective of the source bit depth.

3. If its bouncing to a floating point format dither is not in the circuit. i.e. if the render bit depth is 32 or 64 you dont need to worry about toggling the dither since it wont be applied. In general you can leave dither on unless you really don't want it even applied to integer formats.

 

[scook]

36 minutes ago, musikman1 said:

do you think that the way I had previously rendered and mastered, not using 32 bit and dithering twice could have caused that vocal stutter on the master that I had mentioned? 

 

Stuttering is usually a buffering problem. Maybe a CPU load issue.

[musikman1]

I doubt it's a CPU issue, I have upgraded my PC and the CPU meter shows it's barely working, 12 threads in all and 32GB RAM (those are old specs on my post footer, I haven't revised that yet).  I do have a lot of plugins going, it could be one of the plugs I have on the vox buss.  Btw, thank you for taking the time to reply, always appreciated. 

The stutter was just on one word, the rest was fine. It's probably not an audio skip or drop from lack of resources, it sounds more like a glitch where the word I am singing gets a little garbled for a second.   Going back just now and listening again, the original vox tracks were good. The glitch can be heard on both the final mixdown, and on the master, so whatever caused it happened within CWbB on the final mix bounce.  I'll see what happens tomorrow when I bounce it all again to 32 bit.  You almost can't notice it unless you really listen close.  I do have that lead vox track doubled, so it is possible that one word is slightly out of alignment between the two tracks.  Plus the word is "sleep", so it could be a sibilance issue.   I'll check a few things more closely tomorrow, I'll find what's causing it.

Edited July 10, 2022 by musikman1

[musikman1]

19 minutes ago, scook said:

Here are a couple of posts from the old forum where Noel goes into more detail about the audio engine, bit depth and dither.

thanks for the links, I'll have a read through, much appreciated!

[scook]

7 minutes ago, musikman1 said:

The glitch can be heard on both the final mixdown

Is this a fast or real-time export?

If fast, try real-time or see if you can isolate the plug-in causing the problem.

If real-time, bump up the ASIO buffer.

[musikman1]

Fast.  I don't remember ever doing a Real-time export. 

After listening more closely today, and isolating that vocal track with all FX plugs turned off, the glitch is there.  Almost 100% certain it was something I did in Melodyne when I edited all the vox tracks.  Sometimes Melodyne will garble a word or syllable if you edit with the modulation or pitch tool and go a little too far, I've had it happen before, and then had to back off a bit to get it to sound more natural.  So what I did was go back to an older version of the same project and I grabbed the original piece of that vocal and brought it into the newest version of the project.  A little repair job I guess you could say.  So  now it's a bit better, but I still have a little smoothing to do.  At least I now know it's not anything to do with a plugin or a render issue. I had a feeling if it were a render or plugin issue, more glitches would have occurred, not just on one word.  

I did a draft mix as you had recommended at 32bit, then ran it through the mastering software.  Still a few tweaks to do, but at this point in time it's sounding pretty good.  ?

Edited July 10, 2022 by musikman1

[John Vere]

Melodyne is amazing software but definitely can make a mess of things if overly used.
I alway copy the vocal track first to make a working copy track that gets processed. I chop it up into short clips and remove silence between lines.
If I hear any glitches after rendering I simply delete the garbage clip and drag copy with Shift/Alt held down from original track to the working track and try again. 
And I always work in 32 bit no dithering when creating stem or bouncing tracks. 

[musikman1]

1 hour ago, John Vere said:

Melodyne is amazing software but definitely can make a mess of things if overly used.
I alway copy the vocal track first to make a working copy track that gets processed.

Yes it can!  I do the same thing, make an extra copy of any vocal tracks that I have slated for Melodyne, just in case, that way I always have copies of the originally recorded vocals. 

1 hour ago, John Vere said:

And I always work in 32 bit no dithering when creating stem or bouncing tracks. 

I will be working with 32 bit all the time now. For awhile I was using 24 for rendering, and not long ago I had changed my audio data render setting to 16 just because that day I wanted to export without mastering, and burn a quick CD to listen to the mix in my car, then forgot to change it back to 24!  Now it's at 32 and that's where it's gonna stay.