do you dither your 32 bit files before conversion to 24 bit?


The manual on page 193 says, “whether to dither when mixing down to 24 bits is a matter of taste”.

I’ve also read elsewhere, " … only dither at the end of the chain". So, if it’s all going to a 16 bit CD, one wouldn’t dither first from 32 bit to 24 bit, and then from 24 bit to 16 bit.

Of course a single 24-bit file can be drug from the pool and converted to 32 bits in the project, changed in some fashion (volume, etc.), then brought back to 24 bits and replaced in the pool … multiple times. So I guess it’s a good thing dither isn’t being applied automatically by Cubase each time, right?

I guess maybe it’s all academic … any sonic degradation must be small enough to be unnoticeable.


Now why would you dither to 24 bits when going to a 16 Bit mixdown…? And especially how would you do these two dithering steps in the one process that is audio mixdown…?

Well not quite all at once in the way you described, but the question refers to, for example, processing a 24 bit file in the 32 bit environment, for use some time later, then after that bouncing down to a 16 bit CD. If I’m understanding him correctly, Hugh Robjohns suggests that the 32-bit to 24-bit conversion should be dithered, in addition to the later 24-bit to 16-bit.

Noise shaped dither eg UV22 should only be used once at the end, general dither, (no noise shaping) technically should be used whenever the wordsize is reduced. going from 32 to 24, any re-quantization noise is so low as to be inaudible so it’s debatable if using dither is necessary?

Also the resolution change from 32 to 24 typically does not present a truncation problem like 24 to 16 or less. That is a generalization, not a fact by the way. There are scenarios where you SHOULD dither from 32 to 24, primarily if that is the target media format.

Thanks, everybody, very helpful!

Split, or anyone else - what non-noise shaped dither do you choose?

(I only saw the UV-22HR in the Cubase plug in manual).


Nice article about dithering

I use wavelabs internal dither when requiring non noise shaped dither. although you can probably find a plugin somewhere that’ll do it for you, just a good old triangle dither is probably fine.

Noise shaped dither will have a higher peak level than non noise shaped, due to the narrower bandwidth of the noise.

Thx much, Split!

I’m always uncomfortable to some degree loading up other plug-ins … hmmm, I wonder if my old SX3 has a non-shaped dither I can bring over.

Now that I learned the old “Only dither once” is wrong, I’ll monitor how often I actually do “bitter down” (e.g., 32 to 24 bit), maybe it’s infrequent enough that I don’t need to dither there (since artifacts are reportedly less than 24 bit to 16 bit conversion).

Thx again!

That is a highly debatable standpoint, and just cause one person says it convincingly doesn’t make it so. His arguments aren’t very scientific either (“analog also added noise in every step”). There are many sources still holding on to the very sensible standpoint of only dithering as the one time final process. Bob Katz for instance.

Dithering always sounds better than truncation without dither> . 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> .

Project wide yes, multiple generations of dither would indeed be bad. But for individual elements within a mix no.

Aloha guys,
Just to chime in on this.

‘Dithering’ is like throwing dirt on a clean pair
of white runners to hide some scuff marks.

No matter what you call it and theories be damned, any form
of dithering is just adding noise.

Works in some situations (like on paper) and not others
(like music with restricted/limited dynamic range).

And truncation is…?

we have a winner …

More noise.
Just a different form of ‘dirt’.

Except you can’t control where ruination appears in the nose floor. And it it usually well within the audio range.

To continue the analogy,

I can also pick which '‘street corner’ or ‘river bed’ from which to get my dirt
but it is still dirt.

Dirt that was not there at the time of the creation
of the original musical information (white scuffed runners)

No matter what form it takes, we are talking about adding noise.

Funny thing is tho’ in many cases this process works.

yes it works, sometimes beneficially. I’m not really arguing with you about aesthetics. You can even go straight from 32 or 24 to 16 without dithering or truncation if gain stage properly. However, when you go from 32 to 24 AND you dither properly, the introduced “noise” is WAAAAAAAAY below a humans hearing range. Truncation is NEVER below a humans hearing range. I’m not making a value judgment.

In graphics, the merits of dither work the same but are so much easier to see. Check out these photos from this website: Undithered and dithered. Now is dirt still really dirt?

Here’s an excellent video, in which dither is discussed and demonstrated.
It’s about 20 mins long, but well worth a watch. Dithering discussion is about 10 mins in.

In a nutshell: dithering gives you a few dB of extra dynamic range down near the noise floor.
Whether or not you need to use this extra range is down to the type of music you are making.
Is it possible to hear what is going on at the noise floor underneath a “normal mix”? Of course not. It is too quiet.

So the “dirt on white trainers” analogy breaks down here. Music is not one uniform “colour” for starters, it is dynamic, and “multicoloured”. And noise is not “black”, it is “multicoloured” also. So it is more like watching the passing landscape from the window of a train moving at 200MPH, while it is raining gently. Can you see the raindrops suspended in the air? Maybe when the train comes to a stop, but not at 200MPH (someone call the analogy police!)

Remember: “No great record was ever ruined by not dithering”

Anyway, grab a coffee and watch the video :smiley: