Our frequency noise and contrast defect (measured at the OMC) are lower than they ever have been in O4 (contrast defect, frequency noise). As a part of trying to track down what may have caused this reduction, I tracked the coherence of DARM with the LSC REFL A RIN, using that as a witness to frequency noise, as it was used in OAF for online frequency noise cleaning in O4a (source: 72276 and poking around the OAF screens).
I tried to pick times that were a few hours into the lock to avoid thermalization confusion. I also used only observing data.
There are a couple ideas about what could have caused this improvement. A short list:
- ring heater/power changes: we have stepped up and down the ring heaters by 0.1 W to avoid PIs, and briefly stepped up the input power by 1 W for the same reason
- AS WFS yaw offset: this was turned on due to an investigation in the O4 break, and then turned off after the OFI vent recovery
- OFI change: this doesn't help match the arms, but changes the mode matching to the OMC which could effect the contrast defect
These are the times I used. All of these times were at the same 60 W input power from the PSL:
Dec 10 2023: O4a time, no WFS offset in use, EY ring heater set to 1 W (1386244818)
Apr 11 2024: O4b time, before the OFI problems, WFS offset in use, EY ring heater set to 1 W (1396950886)
Jul 10 2024: O4b time, just before OFI repair vent, WFS offset in use, EY ring heater set to 1 W (1404656647)
Sept 14 2024: O4b, after OFI repair, WFS offset in use, EY ring heater set to 1 W (1410432000)
Sept 20 2024: O4b, after OFI repair, WFS offset set OFF, EY ring heater set to 1 W (1410855190)
Oct 22 2024: O4b, after OFI repair, WFS offset set OFF, EY ring heater set to 1.1 W (1413635762)
Here are some notes for the comparison of these times:
- the frequency noise coherence dropped significantly from O4a to O4b (red trace to blue trace)
- the frequency noise coherence increased again after the problems with the OFI (blue trace to green trace)
- the frequency noise coherence improved again after the OFI vent recovery (green trace to brown trace)
- the frequency noise coherence improved further after we turned off the WFS offset (brown trace to pink trace)
- the frequency noise coherence stayed the same when we changed the EY ring heater to 1.1 W (pink trace to light blue trace)
All of these O4 times show less coherence than the O4a time. Based on this data, it seems like the WFS offset did have an impact on the frequency noise. It also seems like the various vents with output port changes could affect the frequency noise, but the overall beam alignment in the arms could have changed during the vent. For example, we did adjust the camera offsets/ADS gains during the vent commissioning times. The change in frequency noise during the OFI problems (between April and July) could have a similar source, since we had to change a lot of alignment during that time. I'm not sure if any arm alignment was significantly changed though. Finally, it seems like this small ring heater change had no effect on the frequency noise.
Just a quick look at how the input jitter coherence has changed with these changes as well. For estimating input jitter, we generally use the IMC WFS A DC pitch and yaw channels, both for the noise budget and for jitter cleaning.
I used the exact same times as the freqeuency noise traces for ease, and I matched the colors too for easy comparison. Note: I used coherence with a channel that has no jitter subtraction to also avoid confusion.
This first plot compares the jitter coherence with DARM from all the times across the whole band in pitch and yaw. It's a bit hard to read. To make this easier, I'll break down the trace comparisons:
- comparing Dec 10 2023 O4a to Apr 11 2024 O4b shows that the jitter coherence got worse
- comparing Apr 11 to Jun 10 during the time of OFI problems, the jitter coherence gets better
- comparing Jun 10 to Sept 14, before and after the OFI vent, the jitter coherence stays about the same
- comparing Sept 14 to Sept 20, when we turned off the AS WFS offset, the jitter coherence gets worse
- comparing Sept 20 to Oct 22, when we changed the EY ring heater, the jitter coherence is about the same, with a bit of worsening around 10-20 Hz
- the jitter coherence now (Oct 22) is worse than the coherence in O4a (Dec 2023)
- the jitter coherence before the OFI problems (Apr 12) and now (Oct 22) is worse in some places and better in others
The most interesting point here (in my opinion) is that the improvement in frequency noise from the brown trace to the pink trace, that is, when we turned off the WFS offset, is opposite to the effect on the jitter noise. We've seen this before: improvements to frequency noise worsen jitter coupling and vice versa. I still don't understand that mechanism.
These are saved in the same template as the frequency noise plot: /ligo/home/elenna.capote/freq_noise_coherence_compare.xml
Daniel rightfully pointed out to me that REFL A RIN is a better measure of intensity noise than frequency noise. In the process of thinking about this, I realized that we found that PRCL had some offset present that increased DARM coherence to LSC REFL RIN. We first fixed this by applying a digital offset, and then Sheila rephased POP9 which had the same effect of reducing the coupling to REFL RIN and increasing power in the PRC. The digital offset was applied to PRCL from March 30 to May 6, and then again applied from Sept 16 until Sheila rephased POP 9 on Sept 23.
Since the PRCL offset/POP9 rephasing effects the coherence of DARM and REFL RIN, and to better ensure I am capturing frequency noise, I reran these measurements using coherence with the REFL SERVO CTRL channel.
Figure 1 We see similar improvement and worsening between O4a, Apr 12 O4b and Jun 10 O4b. Then, the three traces after the OFI vent show the same coherence, Figure 2. This indicates that the improvement between the brown (Sept 14) and pink (Sept 20) trace in the REFL RIN plot (original alog) is likely due to the changes in PRCL and not due to changes in frequency noise. In fact, the frequency noise coherence since the vent for these three times looks about the same.
This leads to the conclusion that the something about the OFI vent itself changed the frequency noise. We might be able to attribute it to an alignment change, but my sense is that whatever alignment change that occurred is small compared to the significant change in the output mode matching. I still don't full see how this then couples back to frequency noise, but it's worth some thinking and modeling.
