Since the ITF was relocked on Thursday, the gain of the optical response estimated by Hrec for NE and WE differ.
On the VIM plot https://vim-online.virgo-gw.eu/resources/2025-06-06/resources/20250606_cal_hrec_gains.png (plot1),
we can see that the NE optical gain is much lower than WE or BSx300 optical gains. Hrec uses lines injected around 70 Hz
to do this optical gain estimation. Is there a problem with the NE line or more generally with the optical response estimation?
Also, the modulus of hrec/hinj (plot2), estimated online using the lines injected on NE, is lower by
a few percent with respect to the values on March.

The first part of the shift was dedicated to the ITF recovery, test of the Locking acquisition and TCS Tuning (See report #66924 and #66921).
ITF back in LOW_NOISE_3 from 18:42 UTC, at 19:27 UTC Grimaud launched a 6 hours long Calibration.
The ITF unlocked at 20:54 UTC, left relocking with AUTORELOCK_FAILSAFE engaged.

SUSP
From 14:55 UTC to 15:04 UTC Boschi decreased the correction of the WE IP with the ITF in DOWN State.

This afternoon, due to the problems in the lock acquisition reported in the entry  66921, tthe CH powers were adjusted in several steps. In the end, the following settings were used:

• NI CH pickoff (with all the actuators ON): 1.44 → 1.81 W

                        → 113+60=173 mW injected into the ITF

• WI CH pickoff (with all the actuators ON): 0.77→ 0.83 W

                        → 59+ 10=69 mW  injected into the ITF

 The CO2 powers settings are summarized in the table below:

* The CH values still need to be checked by closing the flip mirrors (TO BE DONE).

After the daily meeting we have dedicated some time to try to improve the lock acquisition which was unlocking repeatedly around CARM2MC. We have checked the following:

*B2 169MHz phase = -0.3rad

*B1p 56MHz phase = -0.6rad

*B2 56MHz phase = -0.18rad 

Still the transitions were not very stable so we have slightly modified the powers of the two steps CARM2MC and DARM2IR, to compensate for the increase in the intra cavity power. This seemed to help in the stability of the transitions. This mitigated but not solved the problem. Since the sidebands in CITF were low Ilaria adjusted the CH power and the lock acquisition worked well again.

To be rechecked the phases in the central part on Monday.

I tried to investigate the correlation between these glitches and the auxiliary channels. First, I identified a BRMS channel from Hrec that could serve as a glitch witness: V1:DQ_BRMSMonHrec_BRMS_HREC_HOFT_FREQ_BAND_85_95_Hrec_hoft_16384Hz seemed like a reasonable choice. I then computed correlations with all _mean and _rms trend channels, as described in a previous entry (#66922).
Fairly high correlations were found with the following channels V1:Sc_WI_FF50HZ_*_rms, V1:Sc_BS_CMRF_rms and V1:SDB2_POWERSUPPLY_DBOX_LEFT_DOWN_p12V_rms. However, these channels are flagged as “glitchy” and “Danger,” and are therefore very likely classified as unsafe in the channel safety study performed during ER16: link to the spreadsheet.

One possibility is that the observed correlations arise from these channels "reacting"  to the loud glitches in the strain channel, rather than being causally related to their origin. Therefore, this study does not appear to provide new insights into the origin of these glitches.

As with the range correlation study (#66922), more information may become available once we have access to longer stretches of undisturbed data in DQ Studies operational mode.

In parallel with the investigation of the loud mid-frequency glitches (#66895), I looked into possible correlations between fluctuations in the BNS range (possibly affected by the glitches) and the information available from auxiliary channels.
Specifically, I correlated data from the DQ Studies operational mode segments from June 4-5 with all _mean and _max channels from the trend frame (~14k), excluding obvious unsafe channels (e.g., BruCo excluded channel list).

The result was not particularly conclusive. The highest correlation (Pearson coefficient ≈ –0.35) was found with the following channels: V1:ENV_CEB_RF_8MHz_Q_FS_rms and V1:ENV_CEB_RF_8MHz_I_rms.

Figure 1 shows the time series of the BNS range and the first auxiliary channel over the entire analyzed interval.
Figure 2 shows a zoom into the final segment in DQ Studies mode. Some dips in the range appear to be correlated with this auxiliary channel, though not all variations are aligned. It’s unclear how informative this correlation is; I leave further interpretation to the experts.

In addition to these, a non-negligible correlation was found with the V1:NCal_WSN_motor_T_*_mean channels.
Looking at the final segment in DQ Studies mode, some oscillations in both the range and auxiliary channels do appear to be aligned: see Fig 3 and 4. Again, it's not clear how informative this observation is.

I plan to refine the analysis as soon as new data in DQ Studies operational mode becomes available, following the ongoing thermal tuning and the other Commissioning activities.

EDIT: Please note that the title at the top of the plots is wrong (a reminder of past correlation studies...)

To get the fiber noise suppression loop operating it was necessary to increase the light power deliverd to the fiber from Atrim to SQZ-DetLab. This required reducting about 25% (nominal) the power sent to the phase cameras. See Fig.1 of

https://logbook.virgo-gw.eu/virgo/?r=66789

This probably explains the observed power drop.

During last weeks there were some interventions on the distribution of the input beam pickoff towards the phase camera and squeezing. The power difference seems to be compatible with the different values seen by the phase camera.

ITF found in Prepare_Science and in relocking phase (Acquire_low_noise_3); Autorelock failsafe not engaged.

At the beginning of the shift Ilaria form remote was working at TCS tuning; unfurtunately the ITF unlocked in Acquire_LN3 because of the planned work NNN NCal replacement  at NE building by the Cal team. Cal activity concluded at 6:00 UTC; in parallel Michal performed a test of DET safety trigger adjusted by automation.

After that the commissioning team (on site and from remote) worked on the ITF recovery; still in progress.

DMS/TCS

TCS moni restarted many times to take into new configurations. Tuning to be completed.

On 04/06, the ML_AC channel showed an increase in RMS value as observed on the DMS (see attached plot).
Upon investigation, we found that this noise increase was not visible on any other PSL channels, including the ML_DC channel, which shares the same photodiode (see second attached plot).

This suggests that the most likely cause is a malfunction of some electronic components specific to the AC channel of the photodiode.

For now, the flag on the DMS has been shelved.

The B4 phase camera has been misleading, indicating lower powers on B4 for both the 6MHz and 56MHz than in March.

Figure 1 shows the power in the last two days, with 0.05 on PC B4 56MHz, 0.075 on PC B4 6MHz, up to 0.024 on B4 112MHz mag and between 0.03 and 0.04 for B4 12MHz mag.

Figure 2 back in March the power on B4 112MHz mag was 0.026 and on B4 12MHz mag was 0.036, so similar to the last two days, but the PC powers were 50% higher, with 0.08 on PC B4 56MHz and 0.11 on PC B4 6MHz .

Figure 3. looking at the long term trend of the phase camera, all of the powers including carrier power decreased, the ratio is about a factor 1.4. So that would explain why the sidebands as measured by the phase camera decrease. For example the phase camera alignment changed, on the reference beam power changed, and it shows value that are 40% lower now compared to two months ago.

This morning the NNN NCal (Box 10 rotor R4-10) has been replaced by the NCal (Box 06 Rotor R4-17) which was installed on the NWN slot.

This NCal swap required the removing of the box top plate for the metrology target holder (in order to fit in the vertical support), as well as the swap of the bottom spacer( to provide the right twist angle).

During this change, it was noticed that the power supply channels for the NNN NCal motor (32 V channel 1) and North setup position sensors (5 V, channel 2) was broken. More specifically, it could not be turned off by the “output” button, and once the main power (red) button was cycled off/on, these two channels could not anymore supply any current. They have probably been damaged during the fast stop of the rotor that might have send back more current than the shunt regulator could stand. Therefore, the position sensor cable was connected together with the 5V photodiode cable, and the NNN motor power cable moved to the channel 1 of the NCal East power supply. However, the channels 3 and 4 of the North power unit are still working properly.

The four nominal NE NCals have been restarted with their usual configuration at 5:58 UTC.

This morning I have checked a possible workaround to the increase in B1s power during SR misalignment: https://logbook.virgo-gw.eu/virgo/?r=66910

Figure 1. During the lock acquisition of the OMC a check on B1s power is a critical trigger to close the safety shutters for unlocks during that stage of the lock acquisiton

Figure 2. However once B1 is used for DARM control then the trigger on LSC_DARM and the trigger on B1 DC power are much more sensitive than B1s. So the trigger on B1s becomes un-necessary

I have added two lines in DET_MAIN  (line 486 and 662) to change the weight of what is in the trigger between the case when the shutter is open and closed, and tested that it works:

2025-06-06-05h46m34-UTC>INFO...-AcSumChSet> fast_shutter_trigger - rampTime 1s, nStep 7, nb 6/7:  1 B1_PD1_DC_saturation, 1 B1_PD2_DC_saturation, 1 B1_PD3_DC_saturation, 1 B1p_PD1_DC_saturation, 1 B1p_PD2_DC_saturation, 0 B1s_PD1_DC_saturation, 1 fast_shutter_manual_close,
2025-06-06-05h47m43-UTC>INFO...-AcSumChSet> fast_shutter_trigger - rampTime 1s, nStep 7, nb 7/7:  1 B1_PD1_DC_saturation, 1 B1_PD2_DC_saturation, 1 B1_PD3_DC_saturation, 1 B1p_PD1_DC_saturation, 1 B1p_PD2_DC_saturation, 1 B1s_PD1_DC_saturation, 1 fast_shutter_manual_close,
 

I have commented it back out, as it is the wrong place to do it. We don't want to turn off this trigger as soon as we open the shutter, but only once B1 is used for DARM control. If we need again the work around for the B1s increase when SR TY is misaligned, then these lines should be added in ITF_LOCK, with the reactiviation of the B1s check in DOWN state, and the deactivation of the B1s check for example when we arrive in LN1.

This morning, the automation was stuck in ‘locking arms on IR’, even though the arms were already locked on IR. To recover the lock acquisition, I put ‘down’ and restarted the lock acquisition.

The sidebands in CITF were quite low (B4 12 MHz ~0.27 mW), so at 05:54 UTC, I undid yesterday’s last power increase on the WI outer ring. After a few attempts, the ITF successfully re-locked in CARM null 1f at 03.42 UTC.

I performed two different steps on the NI DAS:

1) NI IN power increase:  +400 steps (03.53 UTC)

-pick off: 0.76 → 0.774

2) NI OUT power decrease:  -200 steps (04.17 UTC)

-pick off: 0.774 → 0.76

The ITF reached LN2 at 4.51 UTC.

The ITF unlocked during ‘acquire LN3’ at 05:32 UTC, but not because of excessive power on B1s (see figure below).

Then, Francesco told me that work had started at the NE, so the ITF likely unlocked because of that (o_o).

Both ITMs should have an additional ZnSe viewport (DN63 - 57791 and 57772) with a much better viewing angle—similar to the one used for the photon calibrators on the ETMs. However, a taller tripod will be required for the thermal camera.

The shift was dedicated to the continuation of the ITF recovery and CH / DAS tuning (see report #66906). Activity concluded at 17:50 UTC, further attempts were performed by Was from remote (see comment #66910).
From around 19:07 UTC, after an unlock at ACQUIRE_LOW_NOISE_3, it was not possible to reach CARM_NULL_1F again.
ITF left relocking with AUTORELOCK_FAILSAFE Engaged.

DAQ
From around 18:15 UTC the cameras B1p, B1, B4_Cam1 and B1s2 became unavailable. Experts informed of the issue.

About 1.5 hours after the last DAS step I have tried to lock in LN3, with a MICH offset to make the sideband balanced, and with the gain of the SR TY loop to reduce the DCP divided by a factor 2 to have a slowe motion.

Figure 1 shows that the reduced SR TY gain did reduce the speed of the SR misalignment, but B1s still increased and crossed the 42mW threshold, it only reduced the speed of the increase. This would disprove the theory that the issue is a gain of a loop being too low, the issue is somewhere else.

Figure 2. Comparing to last night, when the power on B1s was slightly smaller during the peak, the peak occurs at the same relative SR misalignment of slightly less than 1urad. 

Figure 3 shows the B1s image during the peak, Figure 4 a few minutes after (with SR more misaligned) and Figure 5 a few minutes before. The SR misalignment amplifies some spurious beam on the right side of the main one. That spurious beam is odd, and maybe teaching us something important about the interferometer.

We could remove B1s from the trigger of the fast shutter once we are in LN1. As in that case the B1 photodiodes are a better sensor of unlocks, and so is DARM, and the trigger on B1s is no longer critical as it is during the lock acqusition of the OMC itself.

Didier’s observation of the bump in the gated strain channel is very important. It left me puzzled and I spent a bit of time trying to understand what was going on (and ended up rediscovering something pretty basic in signal processing...). I’m sharing some notes here, especially for anyone working on Continuous Wave searches who uses the gated strain channel or applies gating.

This bump is an artefact of the gating process, and doesn't exist in the original data.

Explanation: When we gate a time series x(t) we actually perform a multiplication with a time domain gate window w(t) to zero out certain parts of it. This can introduce sharp edges if we do it "the hard way", that is, with no tapering to smooth the edges: w(t) = 1 - rect(t), with rect(t) the rectangular window centered around each piece of data to be gated. These sudden changes in the time domain lead to broadband energy in the frequency domain: x_gated(t) = x(t) · w(t) -> X_gated (f) = X(f) * W(f) (convolution theorem). That energy can cause a sharp spectral line to spread into nearby frequencies. For example, for a rectangular window w(t), W(f) = delta(f) - sinc(f). So, if X(f) is a spectral line at frequency f0, X(f) = delta(f - f0), and X_gated(f) = delta(f - f0) * [delta(f) - sinc(f)] = delta(f - f0) - sinc(f - f0).  That means, instead of a sharp peak at f0, we get a main lobe and side lobes. With many lines this spreads into the observed "bump": figure 1. 

Using tapered gates (like by means of a Tukey window) helps smooth out the edges and reduce this leakage, but it doesn’t completely solve the problem.

I’ve attached a plot from a simple simulation with three sinusoids (at 38, 40, and 42 Hz) in white noise. The blue line shows the Amplitude Spectral DEnsity of the original data. The orange shows what happens when I randomly cut out sections using the "hard gating" described above. There, the bump is very clear. The green line shows the same with tapered gating using a Tukey window: better, but the bump is still there, just smaller.

I’ve also attached the script I used to make the plot: feel free to play with the number and size of gated parts and the window.

EDIT: fixed an error in the formula for the window function after feedback by Samuel.

MORNING/AFTERNOON up to 19.35 LT

To verify the lock acquisition in the cold state of the ITF, we kept the interferometer unlocked for one hour.
Julia performed the unlock at 07:35 UTC.
The sideband behavior was monitored for one hour by repeatedly locking and unlocking the CITF. It was observed that the sideband level gradually decreased as the ITF cooled down (see figure 1).
After several steps increasing both WI and NI CH powers, we were able to recover some sideband power and proceed with the lock acquisition (see fig.2).

The new powers compared to the standard ones (pre WE mirror replacement) are summarized in the  table below .

The ITF relocked at CARM NULL 1F at 10:24 UTC. However, the interferometer experienced multiple unlocks at different stages: one occurred at CARM NULL 1F (at 11.06 UTC) — shortly after I opened and closed the flip mirrors of the CO2 actuators to check the power individually — and two others took place during the LN2 acquisition phase (at 10.51 and 11.44 UTC).

With Michal, we understood that many unlocks during the night and throughout the day were caused by the saturation of B1s during the acquire Low Noise 3.
We hypothesized that the issue could be related to the last step of the NI outer ring, performed yesterday at 17:00 UTC.
As a result, this step was reverted in the CARM NULL 1F configuration at 13:19 UTC.

However, although the fringe became brighter, we tried  to reach LN3 but the ITF unlocked due to a glitch in the BS.

The ITF relocked in CARM NULL 1F at 14:57 UTC, and at 14:59 UTC I increased the NI outer ring by 5%.

At 15:33 UTC, the WI outer ring was increased by 14% (from 0.36 W to 0.41 W on the pick-off).

Following this step, we observed an improvement in the 6 MHz. We then attempted to reach LN3, but the ITF unlocked during the acquire LN3 phase due to B1s exceeding the threshold of 42 (again).

Then we relocked at CARM NULL 1F and I increased the WI outer ring by 12%  (from 0.41 W to 0.46 W on the pick-off) at 17.21 UTC.

The main signals to date are shown in fig.3.

The sign of the NCal lines subtraction for the lines recently added at 36.52 and 36.56 Hz have been flipped at 17:14 UTC. NNF frequency which was left by mistake at 18Hz instead of 18.04 Hz, leaving lines in h(t) at 36.00 and 36.04 Hz, was fixed at 17:26 UTC. (Thanks to Samuel Salvador for reporting these issues)

On June 4, between 20:00 and 21:00 UTC, a set of "checkhrec" injections has been done. A summary of the analysis of those injections is available in the VIM pages
Plot1: https://vim-online.virgo-gw.eu/resources_archive/2025-06-04/resources/20250604_cal_daily_checkhrec1.png
Plot2: https://vim-online.virgo-gw.eu/resources_archive/2025-06-04/resources/20250604_cal_daily_checkhrecunbias1.png

Plot1: We can see that the injections done on WE give hrec/hinj modulus values about 4% larger than those on NE.
We can suspect that this overestimation on the reconstructed h(t) comes from an underestimation of the gain in the WE actuator's model used currently in Hrec and that this model needs to be updated. I have done a Hrec reprocessing that will confirm this or not.

Plot2: We can see also that, after the online unbiasing, the hrec/hinj modulus for NE is about 2% lower than what we had for instance for the 15 mars injections:
https://vim-online.virgo-gw.eu/resources_archive/2025-03-15/resources/20250315_cal_daily_checkhrec1.png
https://vim-online.virgo-gw.eu/resources_archive/2025-03-15/resources/20250315_cal_daily_checkhrecunbias1.png
Here, the problem may come from a wrong model of the ITF optical response in Hrec. We should check the ITF condition (TCS status and SR alignment for instance)
during those injections. And, whenever possible, we will do optical response measurements for WE and NE and do again checkhrec injections.

Here are the images of the WI mirror surface (the one facing the WE payload) done with the Optris 640i thermocamera:

- Fig. 1 shows the WI mirror surface with the ITF unlocked;

- Fig. 2 shows the mirror with the ITF locked in CARM_NULL;

- Fig. 3 shows the temperature measurements of three diffrerent areas of the mirror (in CARM_NULL): the highest part of the mirror ("area 1") is hotter and has a temperature of about 20.3°C. In the middle, there are two cold spots and they measure around 19.3°C ("area 2"). The lowest part of the mirror has a temperature of about 19.8°C. I couldn't find any point absorbers present on the mirror. However, this seems strange compared with the last measurements done with the HWS in May 2023 (#60377).

- Fig. 4 shows a slightly different color bar range that gives a better visual perspective (CARM_NULL_1F). The temperature in the right upper side of the figure is the background mean temperature of the mirror surface and the temperature written in the middle of the mirror corresponds to the cold spot value. 

ITF found in DQSTUDIES with autorelock enabled.

At 7:35 UTC the commissioning crew started to work on ITF recovery and TCS tuning.

ISYS

The DMS was reporting a red flag for ML- PSL_ML_AC_FS. I informed Matthieu; he decided to temporarily shelve the flag.

SUSP

Two red flags on DMS:

• WE_Electr - SAT..WE_Sc_DSP_STATUS. The DspServerWE is reporting the following message: ERROR..-Tower WE - wrong filter in DSP #43157.  Configuration filter: /virgoDev/Sa/DSPCode_Adv/WE/LC/WE_PSDm.map. Running: /virgoDev/Sa/DSPCode_Adv/WE/LC/WE_PSDf.map
• WE_IP - Sa_WE_F0_COIL_H1 coil close to saturation

Experts have been informed by mail.

PAY

Temporary installation of thermocamera at WI

the etalon down condition for the NI and WI have been set back

I've hacked a Virgo DQR script to zoom onto those glitches over a few minutes -- see plot in attachment

• Top: the Omicron triggers (before time clustering)
• Middle: the standard Omicron clusters (after time clustering; what we usually call Omicron "triggers")
• Bottom: an attempt to make 2D (time-frequency) clusters

=> Most of the strong glitches (dark brown) are short in time and very wide in frequency.