I did two different tests to PRCL today that would hopefully help improve the low frequency noise.
PRCL feedforward test:
I tried fitting a new feedforward with Gabriele last week. We saved this in FM6. However, the comparison of the injection without feedforward and with feedforward showed that the coupling was worse with feedforward.
PRCL offset test with new HAM1 FF:
I engaged the previously tested PRCL offset (alog 79989), and tried the new HAM1 feedforward I fit. Not only does this HAM1 feedforward perform better than the previous tuning, generally the HAM1 coupling is smaller than without the PRCL offset.
Comparison plot. Compare the gold traces with the blue traces. These are both HAM1 feedforward OFF times, except gold is with no PRCL digital offset, and blue is with PRCl digital offset at -58. There is improvement in the noise in CHARD P, CHARD Y, and INP1 P (left three plots). PRC2 P noise looks the same (right side, second down). Then, red shows the new HAM1 feedforward with the PRCL digital offset. The reduction is noise is the same or better than the noise reduction achieved in the previous HAM1 feedforward tuning (see alog 79799). Again, I find it suspicious that the signifcant change seems to come from the loops that use the RF45 demod signal (CHARD and INP1), while the loop using RF9 only do not change (PRC2).
I consider this test a success, and the lower noise in CHARD Y with the PRCL offset on (left, second plot down) is evident.
The PRCL offset is back in the guardian in LOWNOISE_LENGTH_CONTROL and SDFed for observe. I SDFed the new HAM1 feedforward settings in both safe and observe.
Some other results from the PRCL offset test:
No visible change in the sensitivity: plot (there are some excess peaks because the cal lines show up in calib clean for a bit in every lock). However, the CHARD Y/PRCL coherence is gone, and the CHARD Y/DARM coherence is reduced: plot.
I trended the REFL WFS NSUM around the time of the offset engagement, and turning on the offset corresponds to a drop in REFL power: plot.
Here is more data investigating the change in power with the PRCL offset engaged.
It appears that as the PRCL offset is increased negatively towards the chosen value of -58 ct, the power at both the LSC and ASC REFL diodes decreases, and the power in POP LF increases. The offset also appears to make the REFL A LF signal less noisy. Power trend plot
Similar behavior is evident in the March test, and the March test also showed that this behavior is consistent even with the thermalization which steadily increases the power at REFL early in the lock. Power trend plot
Reminder: we chose the -62 ct offset in March and the -58 ct offset now as a part of minimizing the PRCL/REFL RIN coupling. It looks like both now and back in March, increasing the offset beyond these values doesn't much reduce the REFL power further, however we didn't take a big enough step to confirm that.
Another note: in the past we have checked the POP45 phasing and the LSC sensing matrix, but we haven't checked the POP9 phasing. These are some alogs Sheila sent me: 77289, 77416. I think we looked into this a bit, and then the OFI broke, which drew our attention more than this issue.
