I reported in LHO aLOG 82376 that there were aliasing artifacts as well as a large number of line artifacts not predicted by the offline anti-aliasing analysis. Digging deeper into these unknown artifacts, these may instead be due to the fact that DTT normally holds data as single precision. The DCPD data has a large DC component, so if the data is processed with the "Remove mean" unchecked, then there may be artifacts owing to the large dynamic range. My hunch was confirmed when I tried two things:

1. re-check the "Remove mean" option. Many of the unknown additional line artifacts disappeared
2. using the diaggui_test program, which holds data as double precision.
 2a. Using "remove mean", the unknown additional line artifacts disappeared
 2b. Disabling "remove mean", the unknown additional line artifacts still were gone

Check out LHO aLOG 78559 for other experiences.

Attached is a plot comparing the 524 kHz data of H1:OMC-DCPD_A0_OUT (red) with the 16 kHz of H1:OMC-DCPD_A_OUT_DQ (blue). We don't see any significant differences now that the additional anti-aliasing filters have suppressed high frequency artifacts. The second figure is the now updated PSD comparison before the additional AA filtering in DTT using the "Remove mean" option for the 16k data. Additional artifacts consistent with the expected contribution from aliased artifacts are visible, but now the rest of the spectrum seems more-or-less in agreement.