I'll keep this short and sweet. It's prompted by a discussion today in the Home Mastering Masterclass Facebook group, asking when to dither, and why.
[Update: I've made a video demonstrating what dither noise is, and how it sounds - to check it out, take a listen.]
I've written about this subject before before, so if you want the nitty-gritty details, click here - but for today, I'll make it as simple as possible:
Always use dither when saving out to 16 or 24-bit files.
That's it !
The single most misleading piece of advice, which I keep seeing, is "only dither once".
I do agree you should only use it once within a single DAW application, on the main stereo output - so you don't need it on every channel, or in-between plugins, for example - but you do need it once when saving your file before mastering, and then again when exporting the final 16-bit master.
So a better rule would be:
Always dither - unless you're staying at 32-bit floating point, in which case once is enough.
Other versions of this advice I often see include only dithering when saving to 16-bit, or only at the mastering stage. Again, I disagree. It's true that the truncation distortion caused by not using dither at 24-bit is much harder to hear than at 16-bits, but it's there, and it's horrible.
So when you're saving out at 16 or 24-bit before mastering a file, especially more than once, correct dithering is essential.
And if you're bouncing tracks internally, to remove plugins and reduce CPU overhead - you should definitely dither ! (Unless you're saving at 32-bit floating point)
To get into more detailed reasons why I say this, you can click here. But for now, just remember - the only time you don't need it is if you're saving at 32-bit floating point.
A few people have questioned my advice here, but I stand by it. Especially since in the replies when I posted this video on Facebook, Paul Frindle weighed in to say he agrees. (If you haven't heard of Paul, he used to work at SSL and later Sony, where he helped design their legendary Oxford console - he knows his stuff !)
You can read the whole conversation here, but I'll post a couple of the most interesting comments:
"The thing is that there is actually no difference between digital and analogue signals - all have a dynamic range set by the ratio between the max level and noise. The difference is that analogue comes with it's own noise (caused by the reality of signal in the physical world) whereas any digital representation in math requires us to re-insert the physical random component the math does not provide us.
It is a theoretical requirement of the system, it doesn't mask the distortion - it removes it... ANY digital data representation of a signal in the real world has artificial certainty (which reality doesn't) and it has to be removed for the signal to be harmonically accurate - i.e. like a signal in the real world... It's a deep subject that shows our math is an artificial human approximation of reality - but the approximiation has too much certainty. Fascinating implications to that concept...
Noise-shapng tries to push the dither energy out of band so you will hear it less. But in math you cannot get something for nothing - so the trade off is higher peak signal values - because the noise restricted over a smaller band has to have more peak energy in proportion to be effective. My experience? It's a waste of time...
Triangular dither is just fine - it's enough to remove ALL distortion. So more complex probability densities [noise-shaped dithers] do not remove more distortion at all - they cannot theoretically. It's another myth."
