Cheryl, Stuart, Arnaud, Rahul
EDIT: (04/09/2020): I have re-run these (ETMY/ITMY) analysis after Arnaud suggested that I offset the Damping ON/OFF gps time by few mins and let things stabilize. The results from yesterday looked like damping ON/OFF measurements were taken with a transient in the signal.
Attached below are the amplitude spectral density for the Top mass OSEMs for the ETMY and ITMY suspensions. The plot compares the state when the damping was OFF (for an hour) and then ON. For comparison, I am also attaching the results from 2013-14 for ETMYand ITMY.
The noise floor for both ETMY and ITMY above 10 Hz is of the order of ~5e-11 m/rtHz (for the LTV dofs) and ~e-10 m/rtHz (for RPY dofs). There are no resonant peaks observed above 10Hz for the M0 stage osems. The resonant peaks are lower when the damping is ON.
However on comparing the 2020 results with that of 2013-14 (posted by J. Kissel alog 13184), I see that the noise floor in 2013 drops off by several orders of magnitude after 50Hz. I don't see this drop in my results taken today (maybe Jeff or Stuart/Arnaud can comment on this).
The spectrum were taken using a Matlab file stored at the following location,
/ligo/svncommon/SusSVN/sus/trunk/QUAD/Common/MatlabTools/plotquad_spectra.m
The matlab script used over here had few issues which was resolved swiftly by Stuart and Arnaud (from LLO). 24 out of 39 channels worked fine and the rest had some issues in the .m file (which is now fixed according to Stuart's email later in the evening).
Comments from J. Kissel: These OSEM spectra scripts are grabbing data from the frames (nds1 probably). In 2013 we were only storing the top-mass OSEMs at some ridiculously low sampling frequency — probably 128 Hz. As such, the Nyquist frequency was around ~50 Hz. Near Nyquist, the data aquisition system always applies a very aggressive digital down sampling / anti-aliasing filter. The filter is computed from the *ratio* between the rate at which the front-end model runs (in the SUS’s case this is 16 kHz) and the requested rate that the data is stored (in the this case 128 Hz). Thus, the data is filtered with the 128x filter show in the figure on page 3 of <https://dcc.ligo.org/LIGO-T1600059>.
Since then, we’ve up’d the rate by 2, to 256Hz, which means the Nyquist is *just* off the edge of your plot which goes to 100 Hz. And thus you don’t see the influence of the 64x filter.