TITLE: 03/20 Eve Shift: 23:00-07:00 UTC (16:00-00:00 PST), all times posted in UTC
STATE of H1: Commissioning
OUTGOING OPERATOR: Corey
CURRENT ENVIRONMENT:
    SEI_ENV state: CALM
    Wind: 13mph Gusts, 10mph 5min avg
    Primary useism: 0.03 μm/s
    Secondary useism: 0.19 μm/s
QUICK SUMMARY:

Engineering Run #16 (ER16) started this morning and Oli handed over an H1 in NLN with Camilla doing a Squeezer measurement and there is a possibility we may try to "squeeze" in a few other measurements (i.e. no squeezing quiet time) before the 6pm target for Observing.

TITLE: 03/20 Day Shift: 15:00-23:00 UTC (08:00-16:00 PST), all times posted in UTC
STATE of H1: Commissioning
INCOMING OPERATOR: Corey
SHIFT SUMMARY: Commissioning all day with a few locklosses. Relocking took an average of 42ish minutes.
LOG:

15:00 Detector got to NOMINAL_LOW_NOISE
19:25 Lockloss

20:38 NOMINAL_LOW_NOISE
20:42 Lockloss in NOMINAL_LOW_NOISE from commissioners

21:31 NOMINAL_LOW_NOISE                                                                                                                                                                                                                                                                                                                                                                                                                                                                                             

Naoki, Nutsinee, Camilla

We turned the SQZ_ANG_ADJUST servo back to it's O3a nominal ADJUST_SQZ_ANG_ADF.  Changed sqzparams.use_sqz_angle_adjust to True and reloaded SQZ_MANAGER and SQZ_ANG_ADJUST.

Set H1:SQZ-ADF_OMC_TRANS_PHASE as 133deg, new scan (as in 76434) running now to check what angle is best for total squeezing.

Artem, Jennie W., Gabriele

Following up on alog 76516.

We compared DARM for following configurations:
* No squeezing O4a, 12/20/2023 18:10-18:20
* With squeezing O4a, 01/12/2024 01:35:15-01:45:15 
* No squeezing now, 03/17/2024 04:45:31-04:55:31
* With squeezing now, 03/17/2024 08:18:46-08:28:46

Specifically, following ASDs were calculated: sqrt(abs(no_sqeezing_now^2 - no_squeezing_o4a^2)), and sqrt(abs(squeezing_now^2 - squeezing_o4a^2)). These are shown as blue and red traces on the attached plot. The original DARM traces are shown in gray.

Our preliminary conclusion from this is that it appears that excess noise is ~same with and without squeezing.

We'll make this plot with longer time series and do some more tests.

Jupyter notebook is available here.

We have lost lock a few time duing the DARM transitions: 1394979344 1394954887

Both the lockloss tool and Elenna flagged that there's a big glitch in ADS durign the transition, so I've edited the guardian so it will be off.  It was already turned off for the coil driver switching, I've just moved the lines that turn it back on to after the DARM transitions.

I was trying to figure out which things are in the intersection of 1) could give a flat noise in DARM and 2) has been changed during the vent. One suihc thing is OMC lenght noise. This got me down the rabbit hole of understanding the OMC lock.

OMC dither lenght control

The 4.1 kHz we inject on PZT1 for the lock shows an amplitude modulation: this is normal since 4.1 kHz is high enough in frequency and close to 16384 Hz, so that the line produced by the DAC is a bit undersampled.

The OMC length locking error signal H1:OMC-LSC_I_OUT shows now a large 4 Hz bump that was not there in O4a: we think this is due to the fact that the I/Q demodulation phase of the OMC dither line is mistuned: we are coupling line amplitude modulation into I. So hopefully tuning the demodulation phase will help with this problem.

We retuned the OMC lenght demodulation phase while at 2W in PREP_DC_READOUT. The old value was 35 degrees, and gave us almost equal amplitude in I and Q of a 11 Hz PZT line I injected. The optimal phase that moves this line all in I and none in Q is 56 degrees. Decoupling is better than a factor 100. We'll see when we lock to NLN if this improved the 4 Hz bump in the OMC lock and DCPD, and if this mistuning has any effect on the OMC locking point.

I don't know where this phase is set (probably SDF), so we need to check that and update this parameter where it is set

Still to do: measure OMC lock open-loop gain and check the UGF. Lock loss at 13:42 LT caused by the OMC UGF template, the excitation was probably too large
 

Camilla, Nutsinee, Sheila

• No SQZ time 16:07-16:25  March 20th UTC
• Camilla measured NLG at this time: amplified seed 0.170 counts, unamplifed 0.0098 (dark offset too small)= NLG 17.3
  • while Camilla was doing we added the seed pzt scan to the SQZ Manager state and turning it off in down.
• moved ADF to 322 Hz, measured 5dB of SQZ with PSAMs at ZM4 150 ZM5 90 (strain guage 8.46V ZM5, -1.26 V ZM5) 16:39 UTC (no long data stretch), filter cavity ASC on, 42 MHz ASC off.
• mean sqz with filter cavity misaligned, ADF off, touched ZMs to maximize BLRMS.  This was confusing and didn't show clear changes due to our alignment shifts.
• We locked FDS anti-squeezing, and we spent quite a lot of time adjusting the alignment of ZM 5 +6, as shown in the screenshot.  We ended up with an alignment that was ~40 urad different in
• A SQZ 18:10-18:16 UTC 107.55 SQZ phase CLF servo polarity was negative ref 3
• SQZ 18:23-18:29 SQZ phase 207.1 clf servo polarity positive ref 9
• SQZ +10 18:30-18:35:29 217.53 SQZ phase clf servo positive ref 18
• SQZ  -10 18:36:30 SQZ phase 197.62 CLF polarity positive ref 19
• mid sqz 18:44-18:50 SQZ ang 240.83 polarity positive ref 20
• mid sqz 18:52:25-18:59 SQZ ang 95.76 polarity positive ref 21
• sqz clf flip 19:02-19:08 polarity negative phase 35.08 ref 22
• no sqz time: 19:12 until lockloss at 19:25:12 UTC ref 23

Screenshot of different sqz angles attached.  Nutsinee's final attachment compares sqz with two different CLF servo signs.

Oli P, Jeff K

As a continuation of the SUS model updates (76269), yesterday I updated the models, medms, and watchdog filter banks for the ITMs and BS. Everything looks to be functioning as intended. Changes made are basically the same as what was done for the ETMs and TMSs, with the exception of the pi model for the ITMs(attachment1). I will list the model files that have changed:
SUS
- QUAD_ITM_MASTER.mdl
- h1susitmx.mdl
- h1susitmy.mdl
- h1susitmpi.mdl
- BSFM_MASTER.mdl
- h1susbs.mdl
ISI
- h1isiitmx.mdl
- h1isiitmy.mdl
- h1isibs.mdl
The OVERVIEW and WD medm files have been updated for each suspension, there has been the addition of a new filter bank within the WD screen to accomodate for the addition of the RMSLP put into the WD models, and these filters, along with the BANDLIM filter banks, have been updated with the correct filters. SDFs have been accepted for safe(attachment2) and OBSERVE(attachment3). All changes are in svn.

The commissioners are aware of this now and some people are looking into it, but it seems that the new DARM offloading scheme has been sending big kicks (a few 10s to 200 microns) to the ETMX HEPI at almost every lockloss from NLN. This doesn't seem to be happening with ETMY. On attached trends, top blue trace is the ISC tidal signal sent to ETMX HEPI, yellow is the ISI watchdog state, bottom is the ISC LOCK state number. The ISI doesn't trip every time, but it's happend 5 times just over the last week. This isn't normal behavior, but it seems like we've just been ignoring this for a while. I recall having this happen a bit before the end of O4a, and we added a filter to the input to the HEPI model, but that just seems to slow the kick a bit. 

Wed Mar 20 10:11:21 2024 INFO: Fill completed in 11min 18secs

Travis confirmed a good fill curbside.

TCs started around -60C, I reduced trip temps to -120C. Note TCmins today are close to the -200C limit.

With the updated AS A WFS yaw offset, the most significant ASC coherence to DARM comes from CHARD Y, see Gabriele's bruco.

To try to improve this, I revisited the ITMY Y2L gain. I chose to inject broadband noise from CHARD Y from 10-40 Hz and adjust the Y2L gain to reduce the coupling. I was able to acheive a 20 dB reduction of the CHARD_Y/DARM transfer function at 20 Hz, and a 10 dB improvement in the DHARD_Y/DARM at 30 Hz. The CHARD Y coupling to DARM may be a few dB worse from 10-15 Hz.

The good news is that it appears that with the new WFS offset, the CHARD and DHARD Y coupling improve together over the same frequency band. The bad news is that there doesn't seem to be much appreciable change in the noise, suggesting that despite the high CHARD Y coherence with DARM, we are likely limited by other noise there.

Here is a summary of the steps I took and the notes:

Once I determined that a gain of -1.9 was better than nominal, I remeasured the DHARD Y coupling and saw that it was also better, although not the the same degree as the CHARD Y coupling.

First picture shows the transfer function measurement of CHARD_Y_OUT/DARM with different Y2L gains, second picture shows DHARD_Y_OUT/DARM with the initial and final Y2L gains.

I then measured DARM with no calibration lines to check if there was any improvement. I observed no significant change in the noise up to 100 Hz.

Since this Y2L gain appears better, I updated lscparams.py to change the gain from -1.65 to -1.9 (and loaded the guardian, but I did not do any svn commit). I also took SDF to safe and updated. I could not see any SDF diff in observe for the ITMY Y2L gain.

Although the DHARD/CHARD P coherence to DARM is less significant than the Y coherence, it is probably worth checking the P2L gains again to see if we can do any better.

[Jennie, Evan, Gabriele]

We tried again to move the input beam pointing in-lock, similarly to what was done in 76359.

First we moved IM1 and IM3 in pitch, looking at IM4_TRANS and POP_LF. We could increase the power in IM4, and we moved until we reached a maximum in POP_LF. Half way during the test we switched back to ADS from cameras, but it turns out that the PRM ADS was moving in the wrong direction. We went back to camera servos (restoring the original, correct offsets) and helped PRM along to speed up the camera servo convergence.

When we reached this maximum of POP_LF by moving pitch IM1 and IM3, we checked DARM: there was more noise in DARM at low frequency and around 100 Hz, the range was lower, but coherence with jitter was unchanged. Evan tried to improve things by tuning the SQZ, without luck.

We then moved to yaw, and we could improve IM4 and POP_LF further. Unfortunately we lost lock during the test. We don't believe we caused the lock loss.

The change in input beam alignment had a good effect on the optical gain, KAPPA_C increased by 2% (yaw being the most effective direction). For comparison IM4_TRANS improved by 6.6%, POP_LF improved by 2.8%

Naoki, Camilla, Evan, Sheila, Julian, Nutsinee

Many things happened with the squeezer this afternoon. A quick summary is we are back to 5 dB at kHz and we should be able to do this repeatedly. No PSAMs adjustment required at this time.

- When the IFO relocked this afternoon we adjusted ZM5 alignment to optimize ADF trans signal. By doing so we improved both the RF3 and the 42. However this made squeezing worse.

- We adjusted the SQZ angle. We couldn't make it better so we went the other way. This made squeezing worse and ADS TRANS went up with it. Note that this is the IQSUM channel. We didn't think that was a sensible behavior but we've seen it before.

- Sheila and Julian then optimized the crystal tempeature. The NLG for today was 17.3.

- Naoki measured the SQZ IFO sensing matrix. We found a big cross coupling between ZM5 P to AS42 B Y. Other than then everything else was sensible. A new improved sensing matrix has been implemented. 

- We lost hours tracking down why the filter cavity failed to lock on green. A reminder to check SDF next time we run into mysterious problems.

- After everything went back to normal we recovered 5 dB of squeezing. DARM plot said so. BLRMS seemed slightly off. We tried turning the ASC loop off and optimizing the ZM alignment by hand to see if we could do any better than the loop. We couldn't. SQZ IFO ASC loop now works as it should.

- We also optimized the filter cavity offset. Mostly to double checked that we were sitting at a good place. An offset of -28 (where we started) gave the best squeezing at low frequency.

We haven't got to increase CLF power today. 

After a new CLF VCO installed we should revert the CLF sign to make sure everything was the same as before. In theory this shouldn't do anything.