93
by seby
Dither nerd here (I have a thesis student writing her own dither in C++ as I write this - she is in conversation with Chris from Airwindows and he has been super duper generous with his time).
A couple of raw statements, then an attempt at a non-technical explanation.
When downsampling, you absolutely need to dither. Not doing so will result in truncation distortion and quantising errors.
Your DAW _does_ dither for you automatically. When you export (and downsampling involves exporting), you will always - well nearly always - be dithering. Unless you turn dithering off that is.
Dither will be set to ON, most likely triangular probability density function (tpdf) dither, by default in your DAW. In most DAWs, the dither settings are available under the export settings. If you are not tragically obsessed with this sort of stuff, your eyes have probably just glazed over these settings as you exported. This is an appropriate response.
This is why no one talks about dither much any more, because it is baked in and you can ignore it.
Or you could not.
Here is an experiment that everyone can try. Get a high-res file of a track with some good dynamic range. I like to use a 96/24 render of Dire Straits' "Ride Across the River", but most things will do. Do it to Shellac. Import the track into your DAW. Then export it at 44.1/16, and with _no_ dither. (You might need to do a quick google search to find out how to turn off your DAW's built in dithering function if it is not somewhere obvious at export).
Now give it a listen at a nice loud volume. Sounds like crap right?
There are two things at work making it sound like crap:
1) Truncation distortion at the loud parts.
2) Quantising errors at the quiet parts and transient tails (which are really just other quiet parts).
(1) is happening because the loss of dynamic range resulting from the move from 24 to 16 is cutting off the peaks. There is no dither to smooth things out, so you can hear the clipping.
(2) is happening because the rounding errors are far enough above the noise floor that we can hear them.
By way of an analogy to help explain (2) a little more, think of digital images or a digital display or a digital movie file. It is when the image is at its most simple - the sky, someone's skin, water on a lake, and so on, that we start to notice the limitations of the digital render. You can see the wobbly parts. As it is with digital images, so it is with digital sound.
A full explanation of how dithering fixes (1) and (2) involves a bit of a deep-dive into Word Length, and Least Significant Bits. I'll spare everyone unless you want the gory details. Think of it as adding stuff so that the weird digital things happen beyond the range of human hearing.
Here is a second experiment that everyone can try. Do again as above - but this time add a dither plugin as the very last plugin on your master buss. It is important that you have done both of the following - (a) downsampled from high-res to 44.1/16, and (b) turned your DAW's in-built dither OFF. You need to turn off your DAW's in-built dither so that you can hear the result of the plugin dither that you have added at the end of your master buss. Otherwise you would be adding one dither on top of another dither, which is just perverse.
Once you have done this, you can start exporting the downsampled file repeatedly, each time with a different dither (Chris from Airwindows makes many), and compare the results with each other, as well as with exports dithered by your DAW's built-in dither. If you use Ableton, you will have the option (under "Dither options" in the export pop-up window) between rectangular and triangular dither (as well as some others).
Then report back here! You might be surprised just how different some of the exports sound from each other.
If you really are surprised and can hear the difference, then you are pretty much fucked because now you will start worrying about dithering.
: )