Let me make a complimentary comment on this effect.
Because of this oscillation feature, we have a 32kHz peak in the DAC spectrum rather than a 64kHz peak.
For advanced LIGO, the universal AI (D070081) was made to have 3rd-order 10kHz LPF with 64kHz notch.
If we have a higher peak at 32kHz (where the rejection is not enough) than at 64kHz, the filter does not provide
enough filtering of the DAC artifacts.
For the 40m, the original filter had the cut off of 7kHz as the sampling rate was 16kHz.
If we want to extend the frequency range by 4times, the correspoding cut off should be 28kHz.
The rejection is again not enough at 32kHz.
If this peak is an avoidable feature by using simple sample&hold the peak freq is pushed up to
64kHz and we can use the AI filters as planned.