Output Dither

[NewHere]

Hi,

I looked up dither in the Reference Guide 2020.3 and want to make sure that I have understood it correctly. It says:

Pro Tools systems use 24-bit audio input and output signal paths, and internal 64-bit floating point processing for mixing and audio processing.

and

When mixing down to an analog destination with any 24-bit capable interface, do not use a dither plug-in on the main output.

So whenever I route something out to the real world (including inserts and auxes) 24-bit dither is automatically applied to the output stream?

[danyg]

hmmm... when I print in 32 float it comes out at 32 bit float though

[JFreak]

You only need to dither when you change FIXED-POINT word length (bits) from your session to bounce file.

[NewHere]

Thanks for your reply!


I don't want to bounce a file and know that I would need dither then if it'S a 24bit bounce. I am sending stems out to analog and use outboard inserts via my 24bit adda. Usually you would need to dither the stem outputs (auxes) and before audio goes to any outboard via inserts. The manual states explicitly that I should not do that (see my second quote, first post), hence my question: does PT automatically apply 24bit dither to its outputs or does it simply truncate?


Again, I don't mean internally (printing, bussing, bouncing), I am talking of leaving the box.

[JFreak]

You would never dither when you go 24bit to 24bit

[NewHere]

Okay, I am obviously no DAW expert, that's why I am asking here.



I am mixing and processing my audio in PT, that happens in 64bit floating point. Then I am sending it out of the box in 24bit fixed. Where exactly in PT is the word length changed from 64bit fp to 24bit fixed and is dither applied? Why do you think I would want to apply dither going from 24bit to 24 bit?

[JFreak]

Going from fixed-point to floating-point to fixed-point does not require dither. You only dither when you go from fixed-point to fixed point and change word length. As in having a 24bit session and bouncing a 16bit file.

[Bob Olhsson]

Actually, you do need to dither when you go from floating-point to fixed. While every sample is 24 bits deep in 32-float, it's at full level and first must be converted to somewhere between 64 and 80-point fixed. You also need to dither any signal processing no matter what the depth of the source because all DSP (with the exception a decimal shift) expands bit depth.

[Wolfgang Eller]

Originally Posted by Bob Olhsson

While every sample is 24 bits deep in 32-float, it's at full level...

OK, it is 24 bits deep but does the signal have full 144db dynamic without noise?
You can treat the noise with dither noise so that the noise sounds better as it is truncated but I don´t think it does make any hearable sense.

I just dither below 24bit.

[Bob Olhsson]

It's not about changing the sound of the noise. You dither to prevent distortion from truncation!