A Question About Mixdown / Dithering

When in the export window of Cubase, it gives me the option to choose my output resolution. If I have a 24bit project file, and I chose 16 bit as the output resolution, does that mean that all the plugins throughout the project will be rendered at a 16bit depth, or does it mean that it will render the plugins at 24bit and THEN downsample to 16bit? If the latter, I find it odd that no dithering options are available in the export window.

Can anyone clarify/explain? I’m not so bright with the sample rate / bit depth thing.

Ha! You asked a good question. As far as I know, if you have recorded in 24 bit and do not apply the UV22 dithering plugin to the stereo out bus, setting it to 16 bit, in this case, I think you are left with a 24 bit recording. The mixdown window does show this option but there is nothing in the Cubase program that will apply dithering from the mixdown window. Try it.

All audio processing inside Cubase is done at 32-bit floating-point format, no matter which format is selected for export nor which format you decided to set your project to record audio. It’s just final step of the export/bounce, which truncates this 32-bit stream into whatever bit-depth you have chosen.

Wrong. Cubase truncates samples to whatever bit-depth you have chosen. No dithering needed. Do you even know, what is dithering?

If you do want to apply dither, there is a dithering plugin in Cubase called UV22hr that you can put in the master channel insert 7 or 8 (post fader)

Theory.

Usually done during the mastering process,
‘Dithering’ is the act of adding desirable noise
to mask/hide/cover-up less desirable noise
made when going (rendering/downsampling etc)
from a higher bit depth to a lower state.

IMHO (like ‘Dolby’ back-in-da-day), works well on some songs
hardly noticeable at all on others.

YMMV
{‘-’}

hmm, Im confused now.
my projects are all in 24bit, so when I export to 16bit CD I usually dither with the UV22hr from 24 to 16bit. Is this correct?
Or does Cubase automatically dither when export a 24bit project to 16 bit?

No.

Always happy to receive an attaboy from Jarno. :smiley:

So the answer to the OP’s original question is that Cubase truncates or down samples the final stereo mixdown, and does not directly truncate the various plugins used in the mix. Is this correct, Jarno?

And here’s a question or three about mixdown from me then. If I record a project in 24 bit, and export in 24 bit, do I need to dither? Since Cubase performs its tasks in 32 bit float, is this why the UV22 plugin has 24 bit dithering? Should all mixes sent out to be mastered be sent out in 32 bit float? What is the benefit of having a track exported in 32 bit float? It can’t be used in Cubase unless it is down sampled, correct? Can other DAWs work with 32 bit float tracks without down sampling?

Up- / Downsampling is about samplerate - Dither is about Bit depth / wordlength.
And without Dither applied, Cubase truncates all signals where wordlength reduction is applied.
The advantage of exporting 32 bit files is, you can have signal levels over 0 dB FS without clipping the audio.
Your Mastering engineer will tell you, what bitdepth you should send your files.
And of course a 32 bit FP files can be used in Cubase without converting anything.

When going from 32fp to 24 once, it’s debatable whether dither needs to be applied!

More reason if going 32fp to 16 for final product.

The correct among us would say you should apply dither whenever the wordlength is changed downwards, but then the UV22hr is probably not the best choice.

There are reasons that become fairly technical and can be found with a good old Google search.

My project files are generally 16bit. Since 32fp resolution is applied throughout my project, and my ending resolution is 16bit, then I SHOULD be dithering, correct? (I understand dithering, itself, is debatable, but let’s assume we wanted it) I just figured that since I’m beginning and ending at 16, I didn’t need to, but considering 32 bit fp, then I should?

But according to Jarno, this is not correct.

So which is it?

However, If my project is 24bit, and I’ve chosen to export to 16, THEN it would make sense to dither. But not 24 to 24. Correct?

Yes, you should.

Read the sentence in the context it was posted. Jarno´s quote is not contrary to what you wrote above.

Yes it is. Cubase’s signal path is 32bit floating-point from the beginning to the end. When signal enters Cubase (from A/D, track recorded as WAV or whatever) it’s converted into 32-bit FP and stays in that format until it exits Cubase (into a new WAV or D/A converter) and is converted to whatever bit-depth required.

You don’t need to dither. Ever! It’s up to you if you want to dither.

You don’t have to worry about signal level (keeping peaks as close to 0dBfs as possible, but not exceeding it), when using 32-bit FP. But in my opinion, it’s waste of space. 24 bit dynamic range (144dB) is more than enough for everything and forever. Why? Because, as Mr Scotty would say: “You cannot change the laws of physics”.

Most (if not all) modern DAWs can.

True. But even when going to 16, it’s still debatable.

We have to remember what is dithering about. It’s about getting rid of quantization distortion by adding some (semi-)random noise to the signal.

So how great problem is quatization distortion? Well it’s RMS value is 1/sqrt(3) of least significant bit. Which means, it’s RMS is around -100dBfs in 16-bit audio. Now, what kind of distortion level is disturbing? Let’s say you have golden ears and are able to detect 0.1% distortion level. That’s -60dB of your signal level. This means, your audio must have RMS level of -40dBfs or below before anyone can notice, it’s there.

Now, which kind of modern music has RMS levels below -40dBfs? Even in quieter passages.

EDIT: corrected my lousy maths :blush:

This was good info, thanks. :slight_smile:

Just to clarify: While it may look like I’m against dithering, I’m not. What I’m against is 2 things:

  1. “Dithering” becoming a synonym with bit-depth reduction (and even downsampling), which it definitely is not.
  2. Attitude of “having to” or “needing to” dither. You don’t have to or need to dither. It’s optional.

Dithering is basically about converting distortion (bad thing) into noise (not as bad thing). And when using noise-shaped dithering, the noise dithering creates is moved into frequenies in which your ear is too insensitive to hear it. And this means dithering (noise-shaped) is a “free lunch”: You always improve sound quality without sacrificing anything.

But can you hear this improvement of quality or not? Well … as I already demonstrated with simple math: depends on your music. With classical music: definitely. With modern dance music (or Metallica): no way. Something between: your choise.

And for everyone, who are not sure about what dithering is, there’s a great article on Wikipedia: Dither - Wikipedia
While the article is more about dithering images than audio, it’s extreamly good, because dithering of images is easier to understand than dithering of audio, but with a tiny bit of imagination it’s easy to understand why the same methods applies to audio.

Totally agree with Jarno with that. Although for headphone users at high level (bad :smiling_imp: ) truncation distortion is noticeable at the end of fades, so dither is wise to use on the final master. IMO

Yes, agree totally with Split, with one addition: IF you have files mastered separately, do not apply dither at mixdown. Leave it to the mastering phase, where it belongs.

Another great source of info is Bob Katz’s Digital Domain:

1 Like

Oh, you’re right, and when you’re right, you’re right, and you, you’re always right!

Well … it doesn’t matter if you properly use 24-bit dithering to create 24-bit mix. With 24 bits you have so much headroom (approx 20dB more than human hearing range), that even multiple generations of dithered audio doesn’t generate audible dithering noise. Just be sure, you don’t select 16-bit dithering algorithm on your dithering plug-in, when creating 24-bit audio file.

Yes, it’s less important with 24 bit files, but cumulative dither is worse than no dither until the final phase, as stated by Bob Katz at the end of his very enlightening article, linked above: “But to avoid adding a veil to the sound, avoid cumulative dithering, in other words, multiple generations of any dither. Make sure that redithering to 24- or 16-bit is the one-time, final process in your project.”

Another interesting quote, concerning audible range and dither: “By using dither, ambience and decay in a musical recording can be heard down to about -115 dB, even with a 16-bit wordlength. Thus, although the quantization steps of a 16-bit word can only theoretically encode 96 dB of range, with dither, there is an audible dynamic range of up to 115 dB!”

Thanks to everybody who contributed. Learned a lot about dithering in the process. So to summarize,

*ALL plugin processing is done at 32bit float, regardless of project bit depth.
*Subsequent to the plugin processing at 32bit, wordlength is truncated in the export window to 16 of 24 (or can leave at 32).