PT2399 uses sampling similar to DSD but at lower clock which clock rate is also dependent on delay time because it uses fixed 44Kb FIFO memory buffer. This is exactly why at shorter delays it is pretty high fidelity but as you increase delay to its limits it becomes distantly noisy at higher frequencies. PT2399 at longer delay times could use stronger low pass filter on both input and output to smooth out the noise. Most designs utilize higher fidelity at lower delay times and this unfortunately leads to audible quantization noise when you increase delay time. That said PT2399 noise has very high degree of entropy (read: is very random) and therefore sounds kinda analog. I got one in Korg Minilogue and this high frequency noise is part of the charm of delay effect on this synth. It fits some patches quite well actually.
Thank you for this great explanation! The doepfer is basically a dream module, it's a shame the clock noise is so intense... Have you tried using it for clock signals? i.e. inputting a clock division and then using the single outputs to to trigger things around your system, in a manner that is clocked but still not perfectly in time. Not sure if it's possible but I think it would be a neat application.
Great idea, never thought about that. It's currently not in the rack because space is too limited, but I will give it a try when I rearrange next time!