I checked the taken images and confirmed that we don't need to update the reference positions.

ETMX is cooled rapidly (51K: 8K reduction from 59K last morning). So it is better to turn on the heater.

The increase rate of the EX BRT Head temp became so slow. The vacuum level at EXT is also very gradually increasing.

Just a report

I found that ETMX_TM_WIT_P suddenly oscilated when only X-arm locked. Judging from the timeseries of WIT_P_DQ, the oscillation frequency seems to be 7.44 Hz. It is Pit resonance. Today, since there is OMC REFL work, X-arm has locked and ETMX has been LOCK_ACQUISIION from morning, Fig. 2 shows the 11 hours trend. ETMX_TM_WIT_P seems to raise up occasionally.

Michimura, Tanaka

## Abstract

We completed the alignment of the OMC REFL path and their characterization (TFs and phasing). Also, we installed the BS (R:T=90:10, Thorabs BSX11) for the QPD protection from 10W lockloss.

## What we did

• This morining, we aligned the OMC REFL path from the upper mirror of the periscope. During the alignmen, we also installed the BSX11 between BS5 and L5 (AS optical layout) to protect QPDs for the high power beam at the moment of 10W lock loss. Fig.1 shows the current layout around the installed BS. The 90% power of the beam reflected by this BS and dumped  by the dumper, then the 10% power directs to both QPDs.
• This afternoon, we measured the TF from TEST1 to each SEG output with Moku:lab. Measurement setup is the same as the previous measurement of REFL WFS (klog21025). All of the results are stored in KAGRA dropbox/Measurement/IFO/ASC/OMCREFL/TFs. Fig. 2 shows the example of TF from TEST1 to SEG3 of QPD1, The QPD1 result seems to be fine judging from the previous measurement by AEL (JGWdoc). QPD2 results are also fine. (Unfortunately, I could not find the previous measurement in JGWdoc)
  • I'm sorry, the result of SEG2 seems to be lost. However, the results are the same as far as my memory.
• After that, we performed the phasing of each I and Q segment with AM laser. Measurement setup is the same as the previous measurement of REFL WFS (klog21025. This time, we injected AM at 16.88 MHz + 100Hz. When the magnitudes of the I and Q signals are different and their phase difference is not exactly 90 degrees, their trajectory forms a tilted ellipse. By performing an ellipse fitting on this trajectory, we determined the relative gain and relative phase difference. PDF.1 shows the IQ trajectory and the fitting result for SEG4 of QPD2 (K1:ASC-OMC_REFL_QPDA2_RF17_{I,Q}4_OUT) before adjustment. The gain obtained from the fitting, which is 1.017 in this example, was entered into OMC_REFL_QPDA2_RF17_Q4_GAIN, and the phase difference, which is 89.102 in this example, was entered into OMC_REFL_QPDA2_RF17_SEG4_PHASE_D. PDF.2 also shows the IQ trajectory and its fitting result after this adjustment, observed in the ERR channel (K1:ASC-OMC_REFL_QPDA2_RF17_{I,Q}4_ERR), where the phase difference has been corrected. From the fitting results, we can see that the gain ratio is now approximately 1, and the phase difference is close to 90 degrees. The same procedure was performed to the other segments. Fig.3 is the current situation after the phasing.

At 20:23:32 JST, ETMY was tripped suddenly and went to PAY_TRIPPED. At this moment, H3_RMS value seems to jump third time (fig.1), and surpassed the thereshold of WD at last (fig.2). After that, the H3 signal in view of K1:VIS-ETMY_IM_OSEMINF_H3_OUT16 seems to be strange compared with other sensors' outputs (fig.3). Fig.4 is the same signals 17 days before in PAY_FLOAT. Current, H3 seems not to observe ETMY motion for some reason.

I'm not sure how to deal with it, so I leave it tonight.

Abstract

Details

I started the lockloss investigation on 2025/03/05 8:58 UTC ~ 20:45 UTC. Previous lockloss investigation was posted in klog32803.
During this period, there were 18 lockloss. The plots are summarized in wiki.
 

Saturation of OMC PD

Just before the lockloss, we observed the saturation of both OMC PD. In the previous investigation, we saw similar saturation with reducing the AS 17 (which indicates the reducing of the optical gain). But, this time, we did't see such reduction. The AS17 signal looks healthy.
Although the OMC PD signal was oscillated with 20 Hz in most lockloss, the amplitude was not increased. As I heard from Ushiba-san, this frequency came from the suspension and is not related to the lockloss. Figure 1 and 2 show the example.

For futher investigation, I checked channels related to OMC LSC control (Figure 6). One finding is that the typical value of K1:OMC-TRANS_DC_DIFF_OUTPUT is -0.45, which is different from the previous typical value. For example, the value of 12/17 is -0.26.
After reparing the OMC PD B, we did't perform the dark noise subtraction. We need to do it in future. But, as chated with Ushiba-san, this will be not the cause of the lockloss.

Although I checked the channels to monitor the mirror motion, I didn't find the strange motion. I plan to compare the spectrum of OMC PD B is same as previous.

Transient change of ISS feedback

Probably this is not the cause of the lockloss. But, I found the sudden drop of ISS feedback signal (K1:PSL-ISS_FIRST_SERVO_AOM_MON_OUT_DQ) and K1:LAS-POW_PMC_OUT_DQ, even though the error signal of ISS was not changed.
Figure 3 and 4 show the examples.
 

1Hz seismic motion made the IMC lockloss

One lockloss was caused by the 1 Hz seismic motion. That makes the 4 Hz oscillation of the feedback signal of IMC length and IMC lockloss, as shown in Figure 5.

## Abstract

I completed the replacement itself of OMC REFL QPDs but the alignment of OMC REFL path is on going because I found that the beam position on the upper mirror of the periscope is shifted. Also, the characterization of replaced OMC REFL QPDs is necessary.

## What I did

• I heard that a diode on an RFQPD can be pulled out by hand, at least it can be done for RFPD . So I tried it this morining.
• Before pulling out diodes, I confirmed that both diodes on OMC REFL QPD1(fig.1) and QPD2(fig.2) were burnt.
• Unfortunately, QPD1 diode is too firmly fixed and could not be pulled out. Moreover, in the QPD2 case, one of screws on the black lid cannot be also removed due to the screw strip. Therefore, I gave up replacing the diodes. I decided to replace QPDs themselves. I took 2 spare QPDs (s/n006 and s/n012) from the deciccator next to the IMC duct to OMC REFL.
• This afternoon, I replaced the broken QPDs to the new ones (fig.3 and 4). The broken QPDs were brought back to Mozumi by Miyoki-san. On the other hand, we need the characterization and phasing because we replaced QPDs themselves...
• After that, I tried to the alignment to the OMC REFL path. I found that all of beam postions in the path shifted in the vertical direction. Fig.5 shows the beam position on the upper mirror of the periscope and seems higher than thee center of the mirror. According to Ushiba-san, when OMC shroud was installed, the OMC output beam was directed downward. it was then readjusted by tweaking some steering mirrors so that its height at the chamber viewport? matched the original height. In that case, it is natural that the beam postion on the upper mirror is shifted. Anyway, I found that we need to perform the alignment in OMC REFL path from scratch. Today, I have no time, so I continue tomorrow.
• I put the beam dumper in front of the second lens in the path not to hit the high power beam by lock loss on QPDss.

## Next

• put the beam splitter (R:T=90:10, thorlabs, BSX11) for the protection to reduce the power directed to QPDs
• OMC REFL path alignment
• new OMC QPDs' characterization and phasing with AM laser

With YamaT-san, Oshino-san, and Marco-san

Summary

We are maintaining the existing network configuration of cal-gst0 and cal-gst1 while enabling the test server, cal-gst0, to run the cal pipeline for testing Kafka and other components. Today, as part of the preparation, we copied the missing data from gst1 to gst0.

Pre-Work Conditions and Approach

For the time being, we will maintain the current network configuration and proceed with development and testing of Kafka on gst0. The current network configuration allows gst0 and gst1 to have different IP addresses on the DGS network, enabling access to GitHub and other necessary resources.

Work Summary

Our goal for this week's work is to run the calibration pipeline on gst0. To achieve this:

• We confirmed that data1 on gst1 is used in the pipeline.

• Currently, data1 on gst0 was initially empty.

• We copied data1 from gst1 to gst0, and this task has been completed.

Current Status and Next Steps

Now that the data copying is complete, we have started testing the pipeline on gst0. The next step is to verify that the pipeline runs successfully and to troubleshoot any issues that arise during execution.

Memo:

• cal-gst1 = Production server

• cal-gst0 = Test server

Absract:

I checked current OMC ASC performance and found that the performance is not good due to two reasons: CARM offset value and ISS.
So, I modified INITIAL_ALIGNMENT guardian and CARM offset,
It improved the SNR of dither lines by a factor of 10, so it would improve the alignment to OMC.

Detail:

Recently, even performing initial alignment to OMC, alignment to OMC with PRFPMI seems not good, whcih prevented the stable lock.
So, I investigated the reason why the initial alignment doesn't have good performance.

Figure 1 shows the OMC trans PD spectrum when INITIAL_ALIGNMENT guardian was at ALIGNING_TO_OMC.
Blue lines show the spectrum of OMC trans PD signals with the current nominal condition for initial alignment.
Green lines show the spectrum with ISS, which improved the noise floor by a factor of 2 aaround dither frequencies (13-17Hz).
Brown lines show the spectrum with ISS + manual tuning of CARM offset, which improved the noise floor by a factor of 5 around 10Hz.

So, I changed the CARM offset value from -0.8 to 6.7initial alignment.
I also modified the INITIAL_ALIGNMENT guardian to turn on ISS for OMC alignment.
I haven't checked it improves the alignment to OMC, so it should be checked with IFO.

Note:

I also modified the OMMT2T trans QPD loops to reduce the jitter noise during initial alignment, which might be contaminated the OMC trans PD spectrum according from the OMC commissioning in 2024 (klog30453).
Figure 2 and 3 show the OLTF of pitch and yaw loops, respectively.
Though they are not drastically changed, gain peaking around UGF becomes small while keeping the larger DC gain, which would reduce the rsidual motion of beam spot on OMMT2.

[Ushiba, Kenta, Yuzu]
This is a report of last Friday.
During the lockloss investigation (klog32803), we found the strange causality about the IMC and LSC_LOCK guardian and K1:IMC-CAV_IS_RESONATING channel. I guess this strange causality is not related to the lockloss. For the future works, I put the memo about this phenomena. Although we checked the data, we are not sure the cause.

Details

• Figure 1 shows the example of the normal lockloss. The first line is the K1:IMC-CAV_REFL_OUT_DQ and its threshold. The scond line is the K1:IMC-CAV_TRANS_OUT_DQ and its threshold.
  When the channel value crosses the threshold, K1:IMC-CAV_IS_RESONATING changed to zero. After that, IMC guardian and LSC_LOCK guardian catch this change via the decorator and shifted to the DOWN (LOCKLOSS) state.
• Figure 2 shows the example of the strange causality. The chaDetailsnnels are same as Figure 1. The IMC_REFL and TRANS values cross the threshold. At the same time, K1:IMC-CAV_IS_RESONATING changed to zero. But. before the shift, the IMC and LSC_LOCK guardians already changed the LOCKLOSS state. As I checked, the decorator message was "IMC seems to be unlocked". So the decorator catched the shift of K1:IMC-CAV_IS_RESONATING, before actually K1:IMC-CAV_IS_RESONATING changed to zero.
  • In the case of lockloss 2025/02/19-23, we saw this phenomena at least 6 times.
  • Figure 3~7 shows the other examples. The plots are available on wiki.
• I thought I saw this phenomena previously and tried to find the klog post. But there was no klog post.

Yesterday I replaced the connectors from the AI chassis (D-sub 9) to the PEM injection amplifiers to the pannel type ones.

I prepared a new model library of the common mode servo on which the latch system was applied to all IN1, IN2, and FAST gains (see also Fig.1).

The latch system was already applied only to IN1 for IMC and CARM servo as shown in Fig.2 and it works fine. So this new model should help to mitigate a loud glitches coming from a timing gap between turning on and turning off of BIO and probably allow to change a gain also for IN2 and FAST as 16dB <-> 15dB and 7dB <-> 8dB with keeping IR lock.

For applying this new model to CARM servo (= k1alspll), we need to take care about Dolphin issue. So I haven't installed the new model yet. It can be applied on next maintenance day.

-----
An amount of the timing gap is slightly different in each BIO card. So a waiting time on the latch system was implemented as a variable named as K1:***-SERVO_BO_DELAY_N in the past work. Though I had considered to prepare independent 3 variables for IN1, IN2, and FAST gains, one variable should be enough because one common mode servo is driven by using one BIO card. So in this update, an amount of the waiting time is a common variable for IN1, IN2, and FAST gain.