For those needing to create 16-bit files from 24-bit/32-bit float masters, I came across a new dither from Chris at Airwindows. I guess the best way I can describe it as a layperson is intelligent truncation, i.e. no noise added via the usual triangular or shaped dithers but an attempt to get a rich, “dark” silence by truncation choices that follow the trajectory of the waveform. It’s best I don’t put my foot in it further and just share the link: https://www.airwindows.com/dark/
Thanks to @x42 for enabling Ardour to read the labels on the generic UI of the airwindows plugins. There are 16-bit and “HD24” options depending on where the slider is moved. And, yes, I now realize that 24-bit dither is completely and utterly overkill based on real-world and existing ears/technology. Chris just likes to dot his i’s and cross his t’s
Actually it is not a new dither, it is just a biased rounding. The examples he shows clearly have distortion as the level decreased down to the lsb (moved up to 8 bits word length so it would be easy to hear). I can’t see why you would ever want distorted truncation when you can completely remove the distortion with proper (i.e. tpdf) dither.
Hence my OP title using the rather ungrammatical ‘“un”-dither’
I hear you but I disagree. For my classical needs it works really nicely. This from an email exchange with Chris (airwindows):
“…Dark works like an EQ, but at super-faint levels where dither lives. TPDF is mathematically correct, but unrealistic: at whatever level it resides, it produces a precise white noise. In air, you won’t hear that white noise as it’s too bright and won’t remain white noise over propagation through even twenty feet of normal air: the high frequencies don’t persist, especially over long distances. So you’ve got to have quantization, but if you can use it to produce white noise or use it to apply a treble-rolloff EQ at a faint level, the latter (Dark) is going to sound more realistic on classical stuff, which is heard at many feet of distance from the performers (especially orchestral works).”
We should probably gently remind ourselves that we are talking about the LSB of full-scale 16-bit or 24-bit audio here. Any sort of dither won’t be a dealbreaker and pure unbiased truncation won’t either. Chris is simply providing yet another option that just so happens to gently aid—at the lowest levels—the affect of listening to music at a distance which in my case is most desirous.
