This is not strictly related to WL Forum, or maybe it is since most professionals post here.
how many times and when apply Dither? If the file is still unsaved, should be 32 bit float. Let’s say I apply FX in two stages: in theory Dither should be applied only before saving?
And: every time I apply dither, the noise gets louder and louder?
Advice from pros appreciated (PG included, but he seems to be vanished lately, aside from very early in the morning!)
Thank you!
Dither should only be applied once, as the last step to saving a file to a lower bit-resolution. This also means that (mostly) anything done to the file in WL means you’re going from 32-bit float to the ‘save’ resolution.
Example: source file is 16 bit 44k1, you put one plugin on it and want to save to 16-44.1 again. Dither needs to be applied since the program uses 32 bit float internally, so 16-44.1 is stepped up to 32f-44.1 and then back to 16-44.1. So if you expect to be working on the same file again, saving to 32f is recommended. Applying dither multiple times will indeed increase noise.
Luck, Arjan
Thank you so much Arjan, much appreciated!
Interesting topic,
I’ve also asked myself that question lately.
OK, I know about Wavelab working internally with 32bit floating,
and Arjan’s answer makes sure sense with a finished track.
But, what about, if I just process some single samples (16bit) of a multisampled instrument
(which I lately did, removing some clicks with the Sonnox DeClicker plugin),
would you apply Dither on it ?
Say the sample has a constant Volume Level of -1 dbFS, does it make sense ?
Would you apply Dither if the sample had a lower Volume level, say -40 dB ?
What about processing samples with a Sample Rate of 24 bit ?
Would appling Dither make any sense in this case ?
Really curious…
Grateful for any thoughts
Cheers
You apply dither whenever you reduce the bit depth (not ‘rate’), and not otherwise. When dither is appropriate, there is never any reason not to do it, as it takes no effort.
OK, if the sample is solidly loud, then you may well not hear the difference, but it’s not worth the effort of deciding not to do it, frankly. If the sample if quiet enough to hear the slight added hiss, then it definitely needs it anyway.
If you are pressing 0dBFS very closely, you might need to check for overs as a result of the dither - but then perhaps you should reconsider your gain strategy.
Paul
Dear Paul, nice to meet you again on the forum. Would you be so kind to suggest me typical levels you work with (I know it’s a bit generic)?
I would appreciate it very much! (even though this might be out of topic, but I’m trying to steal secrets from pros!)
Thank you so much
My best
I’m not actually a pro, in this field; I know lots of theory, and use it just a bit. I would have thought that keeping peaks at no more than -1dBFS would be sufficient to avoid dither causing overs in practice, but I’ll let others give more real-world advice!
Paul
Much appreciated, Paul!
Your website is extremely interesting and your modesty makes you honor.
Thank you for replying!
Yes, I sure meant bit depth.
OK, if the sample is solidly loud, then you may well not hear the difference, but it’s not worth the effort of deciding > not > to do it, frankly. If the sample if quiet enough to hear the slight added hiss, then it definitely needs it anyway.
Well, makes sense to look at it this way.
Thanks for your replies.
Best
Bitmeter is your friend - I believe that whenever you see (even for a short time) higher bitrate than you target’s file to be saved (=result of any processing incl. fades, volume change - whatever), you should always apply dithering.