Digitakt manual says these two things:
A sample contains a 16 bit, 48 kHz, mono audio file. (p.23)
D/A and A/D converters are 24-bit, 48 kHz (p.64)
It looks like audio is sampled at 24-bit and then has bit reduction applied before saving.
Huh?
If the audio is going to be turned into a 16-bit file, why are the A/D converters 24-bit?
Does this give some benefit?
24 bit recording is preferable, because it lowers the noise floor. So you can record at a lower level without adding noise to the recording. It’s pretty much the only reason why 24bit is useful. Playing back at 16 bit is enough, 24 bit is pretty much useless for playing back audio, unless you are going to play audio back at ear deafending levels. And I really do mean ear deafening, as in much louder then an aircraft taking off. 
Ah ok. Getting a lower noise floor when recording means that reducing the bit depth to 16 doesn’t degrade the audio much, eh?
Clever.
Thanks!
And I guess something similar happens in the D/A process, too ?
Yeah… The delay and reverb are probably working in 24bit 48 kHz along with the internal mixer before the D/A output.
12bit DA convertion can be sweet indeed. Recording @ 12bit is not 
unless you love noise or high level recording stuff that’s low on dynamics, and with a big chance of clipping issues as a result. 
And monomachine doesn’t sample right?
Sampling is recording a sample…
Misread your post, but you can’t sample with the drum either, and that was what this discussion was about. You can load a very small amount of samples to it, but that is not the same as sampling.
Good for you but this topic is not about the mono machine now is it?
Ah yeah. My bad, the UW version can. no offense taken. It’s just not what the topic is about
Settle down folks or I’ll have to mention more Elektron devices and bit rates, and we wouldn’t want that now would we… 
I guess I owe a more serious reply 
…
I’m not an expert at this but I think mixing in 24 bit gives more headroom and dynamic range and ability to sum and mix 16 bit files without degradation… Also fx and audio processing are more precise when done in a higher bit depth. Most daws mix and process in 32 bit float for reasons such as these and dither back to 24 or 16 but benefits are maintained…
Just to be clear and off topic too - if you peruse the MD manual you’ll discover that the sampling isn’t done at 12 bit at all. This seems to be a common misconception. The converters specs are 24-bit in and out.
Don’t remember the details but the manual will clarify - think bit reduction is applied for storage/playback.
apologies for my ignorance… can i load 24-bit wavs onto the Digitakt, or do they have to be 16-bit?
Transfer app will convert to 16bit 48kHz mono.