During the commissioning window this morning, I worked on the St0 to St1 feedforward for the HAM1 ISI. This time I borrowed some code from Huyen to try the RIFF frequency domain fitting package from Nikhil. This required using matlab 2023b which seems to have a lot of computationally heavy stuff like code suggestions added so it was kind of clunky to use and I'm not sure what all of the different fitting options do, so each dof took multiple rounds of fitting to get working. I also had to add an ac-coupling high pass after the fact to the filters because they all went to 1 at 0hz. Still the results I got for HAM1 seem to work pretty well. Attached spectra are the on-off data I collected for the X,Y and Z dofs. Refs are the ff off, live traces ff on spectra. Top of each image are the asds for the ff on & off, bottom is the magnitude of the st0 l4c to st1 gs13 tf. The improvement is broad ~10x less motion from 5 hz up to ~50hz. I'm looking at the rotational dofs still, but there is less coherence there, so not as much to win.
Elenna has said this seemed to have improved chard asc, maybe she has some plots to add.
There is about an order of magnitude improvement in the CHARD P error signal between 10-20 Hz as a result of these improvements, comparing the NLN spectra from three days ago versus today. Fewer noisy peaks are also present in INP1 P. I included the CHARD P coherence with GS13s, focusing on the three DoFs with the most coherence: RX, RZ, and Z. The improvements Jim made greatly reduced that coherence. To achieve the CHARD P shot noise floor at 10 Hz and above, there is still some coherence of CHARD P with GS13 Z that is likely contributing noise. However, for the IFO, this is sufficient noise reduction to ensure that CHARD P is not directly limiting DARM above 10 Hz. I also compare the CHARD P coherence with OMC DCPD sum from a few days ago to today, see plot.
In terms of how this compares with our passive stack + L4C feedforward performance, I found some old templates where I compared upgrades to our HAM1 feedforward. I compare our ISI performance now with the passive stack, no L4C feedforward to ASC, and passive stack with the best-performance feedforward we achieved: the results. It's actually a pretty impressive difference! (Not related to the ISI seems to be a change in the shot noise floor- looks like the power on the REFL WFS may have changed from the vent.)
The coupling of CHARD P to DARM appears to be largely unchanged, so this generally means we are injecting about 10x less noise from CHARD into DARM. from 10-30 Hz.
