The new high pass filter for the SRCL FF (73622) has improved the DARM RMS between 2 and 4 Hz a bit, although it's not clear why there is a new ~4 Hz peak. Now the largest contribution to the DARM RMS are the lines at 1.3 Hz, which we know are caused by the ETMX ISI (73625) and by the peak at 3.44 Hz which is coherent with CHARD_Y.

The lower DARM RMS in the 2-4 Hz region seems to have improved a little bit the non-stationary noise above 10 Hz, see spectrogram and scatter plot of DARM RMS in the 20-50 Hz region vs RMS in the 2.5-4 Hz region.

It's worth trying to further improve the DARM RMS by

• fixing the ETMX ISI problem
• trying to reduce CHARD_Y motion at 3.4 Hz
• increase DARM gain at frequencies < 5 Hz and see if that's enough to reduce the non-stationariity (depends on where's the source of the non-linearity)

Looking at the DARM bicoherence, there are three main components to the noise modualtion above 20 Hz: 1.3 Hz (ETMX ISI), 2.7-2.8 Hz  and 3.4 Hz (CHARD_Y). Not sure what's causing the DARM motion below 3 Hz, but there is a lot of coherence with ETMX M0 and R0 damping signals.