Output Dither

[NewHere]

Originally Posted by JFreak

Which is why you should not worry too much about it.

We can completely agree on that! It's not really that important, but nonetheless I want to consciously make the decision wether to use it or not or at least know what's going on.

Originally Posted by JFreak

Because the output of dither plugin goes to 64bit mixer that feeds the physical output

If you apply dither post any processing (incl. gain changes) it should work perfectly fine. According to that logic any dither plugin on the Master (eg prior to bouncing) would be useless because it would still be going through the 64 bit mixer.

Originally Posted by JFreak

You can think of it (overly simplified) that when you create a fixed-point word from floating-point you get 23 solid bits and the last bit is calculated from the floating point mantissa. So in a sense it is dither in itself. Whatever "errors" there may be are about -140dBFS

I am not the brain to argue with about the theory and I am sure you know your stuff, but I can 100% assure you that the signal not dithers itself (I know you didn't mean it literally). Try it for yourself, take a testsignal at -140, bitcrush it to 24bit (or bounce it at 24) and look at an analyzer of choice (or gain it up and listen). Distortion. Now do the same with dither noise applied. No distortion. If you go even lower you it could make the difference between content is there or content is lost. And again, you are right, we are talking about the noise floor at -140ish dbFS which is completely masked by the noise floor of my desk. I am aware of that.

The simplest thing would be to just check out the Dan Worral video about dither on Youtube, where he does all the bouncing and gaining for us and you can actually listen to the differences.
Spoiler: he also concludes that it doesn't really matter all that much. I think in the end that decision should everyone make for themselves.

[JFreak]

Last insert slot of master fader is the only place where dither has any use. Last insert slot of aux goes back to mixer and even though the fader was set to unity dithering there is futile.

[Wolfgang Eller]

Originally Posted by NewHere

We can completely agree on that!

Perfect closing words for this thread.

[NewHere]

Originally Posted by JFreak

Last insert slot of master fader is the only place where dither has any use. Last insert slot of aux goes back to mixer and even though the fader was set to unity dithering there is futile.

Okay, unfotunately it's impossible to attach audio clips. Here are three files. A snare with some reverb at -122 db, then bounced and gained up again to make it audible. First bounce is no dither, second is dither on an Aux FIRST SLOT, to prove my point, and lastly dither on a Master, last slot (I don't know why it sounds chopped, I didn't change the verb). If that is not what dither is supposed to do and "futile" then what is it's purpose?

No dither: https://www.dropbox.com/s/gh0qziixch...ither.wav?dl=0

Dither on Main Aux, first slot: https://www.dropbox.com/s/4dqu1289sa...tSlot.wav?dl=0

Dither on Master, last slot: https://www.dropbox.com/s/7qm4iy6cz9...aster.wav?dl=0

[Bob Olhsson]

My understanding is that dithering an aux is OK provided you don't add any signal processing downstream. Ideally, you want to dither the master fader tied to the same output.

[NewHere]

Originally Posted by Bob Olhsson

My understanding is that dithering an aux is OK provided you don't add any signal processing downstream. Ideally, you want to dither the master fader tied to the same output.

Exactly, this is actually my take away from this thread. I often ride the stem auxes which would „break“ the dither. From now on I will simply add a master to each aux with dither on the last slot and hide it.

[Bob Olhsson]

I use either Ghz Good Dither or MAAT LINearise because they have no latency. Any other suggestions would be appreciated.

[Wolfgang Eller]

Originally Posted by NewHere

A snare with some reverb at -122 db...

You really know what you did?
You didn't record with 24bit, you recorded with less than 4 bit.
And if you gain it up to 24 bit it still is a 4 bit recording.
The dither noise you added is working on the LSB. You get 3 bit of Signal and 1 bit of noise.
After gaining you still have just 3 bit of signal but 21bit of noise!

If you are really working within 23 bit (Peaks at -6dbfs), anywhere you have to go analog and the best DA converters are around a S/NR of 120 to 126db.
So you have at least 18db of noise which are 3 bit and that is way more than you can add with dither.

On the other hand, if you prefer to record lower than -100dbfs I would use dither in any case.

[NewHere]

I didn’t record anything. I dragged a snare sample from a commercial sample library into PT. Then I put Sonsig on an insert and pulled the channel fader down 122 db (in a 64 bit environment!). I couldn’t get it to -140 and thought it was close enough (which it was). I then bounced the clips to 24 bits. They never left the box. The clips got automatically imported back into the session and gained up with max clip gain and six trim plugins to make them easily audible. Without dither you get a crunchy snare and no reverb tail, with dither you obviously hear the dither noise but also that the crunch is gone and reverb tail is preserved (which is the whole point of dithering), and all that with the dither on an aux which supposedly doesn’t work. The whole point of the clips is to illustrate what happens at the noise floor, what good would a sample at -6 dbFS be in that case?

I wrote several times that a trip through my desk adds analogue noise that is well above dither noise (around -110dbFS). That is not the concern here.

[JFreak]

Originally Posted by NewHere

The whole point of the clips is to illustrate what happens at the noise floor, what good would a sample at -6 dbFS be in that case?

Point is to be able to hear this example you made you need to boost your listening environment 116dB to get comparable levels to -6dBFS sample. And by doing that you are boosting all noise that much whereas -6dBFS sample has the noise 116dB lower level. And when you listen to the -6dBFS sample you surely do not hear the noise at all, whether dithered or not.

Best AD/DA conversion today is in the range of 124dB dynamics which means for best gain staging and headroom you should target -20dBFS to get full dynamics and least noise. If you have stuff -100dBFS you have a problem.

It's like... imagine a mile long straight highway, you are on one end and i am on another. While in theory I could see your car, in practice I just cannot. Need to wait until you are closer and then I can tell which colour it is.