We have continued the recovery having the same issue in the lock of the DRMI that was experienced the day before.

The "lock" was achieved but the SRCL corrections start immidiately to drift

after long analysis we realized that no current was sent to the SR mirror coil. Fabio found a temporary box relay which went in the wrong state due to the shutdown. Fabio simply bypassed it.

Then we have tuned the phases up to carm step 3/3 but we experienced very low sidebands in the CITF so we decided to not waste more time in tuning and we did a step of DAS to anticipate the work of the morning (at least to understand the direction of tuning for tomorrow).

Noticing that the B1p beam was lower than usual on the camera (and the B1p galvo were opening) we removed the offset of 350 um in the B5_QD2_V signal. Offset is now 0. This correction was applied around 17h00 utc.

In addition we corrected the B1_PD3 offset at 17h19 utc by +5.3 uW.

This morning I found the two cavities locked; once recovered the SDB2 vertical position we went on with the ITF lock recovery, all the morning was spent trying to lock the DRMI, the central interferometer has finally stable locked at the end of the shift, the lock recovery will go on in the afternoon.

SBE
recovery of SDB2 vertical correction with F1 step motors from 6:50UTC to 7:05UTC.

This morning (March 5, 2026) around 8:30 UTC I went to the external cabinet hosting the electronic box to install a new battery (CSB UPS 12360 7 F2, see Figure 1), the installation was completed at 8:53 UTC. The box was closed and placed back inside the cabinet (see Figure 2).
At 11:45 UTC I went to DET EERoom to check the power supply and it was erogating 0.34 A, 24 V (see Figure 5), the hardware is operating correctly since it is not erogating the maximum current set (~0.8A), with the battery disconnected the erogated current was 0.15 A.
Figure 3 shows the comparison of ENV_EXT_MAG_N signal on May 25th 2024 (orange plot), Februaty 1st 2026 (purple plot), and March 5th 2026 (blue plot). The reference time are choosen to make sure that no tarins are passing trhough the Pisa-Empoli railway (see #68641).
Figure 4 shows the last 24 h of data.

Indeed, that effect was visible in the past also on the INJ/PSL signals (fig.1) but with a much smaller effect compared to what is seen now(fig. 2). 

This may be due to the fact that, starting from 22nd of October 2024, we are using one of the chiller borrowed by TCS team for the Neovan electronics (located in EEROOM). Thus, a check of the settings of this chiller is planned for today.

Between Monday and Tuesday morning a full recovery of the control system of the suspensions and BPC was required after the power interruption: all DSP boards have to be restarted and control loops verified. The recovery went smoothly, a part from two issues:

• The WI DSP board #31006 connected to vertical accelerometer failed to boot. Instead of replacing it (action that would require the change of both the other accelerometer board, the change of D-BOX, and the re-design of the controllers), we re-programmed the FPGA. That fixed the problem and the board was put back in the Sa_WI crate.
• TANGO servers were not available and the DC motors could not be used. For that reason OB and NI have to be recovered on Tuesday.

A similar feature was observed in the past from the cooling system of the CO2 main lasers, which produced a 1 Hz comb visible in the UPS current and voltage monitors. This comb disappeared when the chillers operated in “power-save” mode. The signal observed here could therefore be related to a similar mechanism.

It would be useful to verify how the devices were powered back on and check which operating mode they are currently in.

Past investigations:

https://logbook.virgo-gw.eu/virgo/?r=63358

https://logbook.virgo-gw.eu/virgo/?r=63357

https://logbook.virgo-gw.eu/virgo/?r=40113

Curiously, the same 1 Hz comb appears in the current of all three IPS phases, with an amplitude similar to the peaks seen in the UPS line’s R phase.

Nothing relevant is present in the MCB monitors.

An example of the locks of the CITF of yesterday. THere was an modulated oscillation present, that can't be blamed to a loop oscillation.

The IB tower magnetometers show the strongest 1 Hz comb magnetic component, indicating a possible source in the nearby area (LL or INJ_EE_room) powered from a single UPS phase (R phase).
The same 1 Hz comb is visible in the CEB magnetometers and in the External Magnetometer, even though the latter is located at a significant distance from the CEB.

It is also worth noting that the UPS is located very close to the IB tower, a configuration that may account for the higher magnetic level measured by the tower magnetometer.

At the beginning of the shift the  SDB2 intervention and the  MC transfer function checks (INJ) were ongoing; both activities were concluded around 16:00 UTC.

The SDB2 evacuation started around 16:00 UTC; at around 16:30 UTC I was able to close the SDB2 loops.

Then I tried to recover the proper status for DET_MAIN that did not go in SHUTTER_CLOSED; to solve the issue I contacted the DET on-call and he fixed the problen at around 17:00 UTC.

After that I locked the cavities and Julia started the lock recovery, see Initial recovery.

ITF left in locked arms.

Vacuum

The SDB2 pumping was stopped around 17:30 UTC.

We started the recovery by turning on the BPC of the green beams, and aligning also the signals at the level of the injection. Then it locked without problems.

We then aligned the central interferometer and we tried to lock. The signal of PRCL that uses the 6MHz seemed almost well tuned (initially was 1.6 and it seemed that 1.4 was better, but this had almost no impact). Instead the signal of 56MHz seemed wrong (when engaging the loops, SRCL seemed to diverge). However it is not easy to tune it, since we use I and Q for MICH and SRCL. We have tried to change it until we were able to lock (initially 3.24, we changed until 2.0 and we manage to get short locks). However the lock do not last.

In parallel we noticed that the alignment signals were also mistuned, so we tried to tune them better, however the phase that seemed good for the H didn't seem good for the V (initially 3.2 and seemed ideal one for H was 2.2 - we checked the sign). So I left the phase unchanged and I commented the alignment in the CITF (1829-1830 in DRM1_LOCK.py), to give us some margin to align by hand and see if the lock would last longer. I managed to tune the phases and to close the vertical alignment loop by hand, so I reimplemented it in the automation. I left the horizontal open for the moment.

In the meantime I called Ilaria to check the ChroC since we didn't expect to tune phases (the arms and the RFC didn't need further tuning). She made a small step of the chroc to try to match the previous temperature. I didn't see any change on the demodulation phases. To be checked better tomorrow.

The 1Hz "sidebands" observed in the laser power signal comes from the power subbly of the central building. Especially in the ENV_CEB_UPS_CURR_R channel. These sidebands were not present before the blackout.

We opended the SDB2 air tank in order to replace the Ethernet switch that was not reachable.

In order to access the switch, we had to remove the NetCom box and analog filters with their support plate.
The switch is attached to the bottom of the tank with two sqares with one screw per square. Before our intervention, it was attached with 4 screws, but we decided to mount it with two screws since the two others are really difficult to access. This will help for future interventions.
We also changed the optic fiber between the Ethernet switch and the tank feedthrought. Actually it was the culprit of the issue.
We removed:

• Ethernet Switch: 1231.9g
• Optic fiber: 8.68g

We added:

• Ethernet Switch: 1228g
• Optic fiber: 7.95g

After closing the tank, Fabio and Piernicola worked on the bench balancing. They added one screw two balanced it properly.

Probably due to the blackout of March 1, the laser beam on laser bench was slighly misaligned with the PMC cavity, which worsens the laser power stabilization loop and add some noise on the laser power noise.

On wednesday March 4 around 15:10 UTC, we re-aligned the incoming beam of the cavity by adjusting the orientation of one of the mirrors before, LB_M12.

The 2 figures show the time series and the spectral desities of the power stabilisation photodiodes. After realigning the PMC cavity, its output power increased a bit, as seen in the time plot. And the laser power noise was reduced, especially at low frequency below 2Hz, as shown in the FFT plot.

As seen in the FFT plot with the signal taken before the blackout. The current noise level is slighly higher then before the blackout, and some "sidebands" at 1 Hz around the 50 Hz signal appeared. This may be caused by the eletric system, but we don't know yet their origin.

This afternoon we tried to drive the M15_axis_x with a different PZT driver. 
We first brought the corrections to 0V with the command online and this had a positive effect on the RFC TRA (green arrow on the figure 1).
Then we connected the lemo1 cable of the PZT axis to the other driver that was also set to 0V and again this had a positive effect on RFC TRA (yellow arrow).
This strange behavior might be due to some hysteresis in the PZT.  

With the new driver we went from 0 to 100 V, adjusting the M16 axis by hand as well, but it did not have so much effect. (figure 2), RFC TRA was still between 3.0 and 3.1 V.  

At the end we reconnected the lemo1 of M15_axis_x to its driver and the transmission remained around 3.1 V. 

To go further on this, the idea would be to swap the RFC QD2 quadrant photodiode that it is not working correctly and then better investigate the problem : 
- Add some features to the online PZT driving to deal with the hysteresis
- Act on the picomotors in common between PSTAB and RFC and see if RFC situation can be improved without deteriorating the PSTAB one. 

7:00 UTC ITF found with arms locked
7:30 UTC beginning of SDB2 intervention
8:55 UTC I manually unlocked the arms to allow SAT measurements
9:00 UTC SAT vertical injections measurements (Palaia, Ruggi)
10:20 UTC WE chillers coolant refill (Andreazzoli, external firm)
11:00 UTC end of SAT measurements, arms realignment. I had to move NE tx+ a few urads away from the last locked position. NI tx then moved during the following hours and is now 4 urads from where it was this morning.
14:30 UTC MC transfer function checks (INJ)
ITF is left with arms unlocked with INJ and DET still working
 

With the help of Federico from remote, we made a temporary fix for the external magnetometer (ENV_EXT_MAG_N). We accessed the magnetometer electronics box located in the little shelter outside of CEB. This box contains one 12V 3Ah lead-gel battery which is connected, in parallel with the magnetometer, to one 24V voltage supplier inside DET-EEroom (the 24V is ment to compensate for the voltage drop along the >100m long cable). The box also contains proper amplification circuits to transmit the magnetometer signal back to the readout ADC in DET-EERoom. 

The battery was erogating just 7V instead of the 12V expected. This is probably due to aging of the battery. We removed the battery and left the magnetometer powered on the remote supplier. In this configuration the magnetometer is more prone to voltage spikes which might occur in the event of lightnings and has no back-up power in case of outages, but it works. For the time being we decided to leave it this way looking for a new battery. Disconnected cables have been secured with tape.

The magnetometer signal apparently looks similar to how it was before, including the drop of response below 10Hz, an issue that has been present since June 2024. Some noise at 5Hz and odd harmonics disappeared (last Figure) but this is more likely due to a disappered source than a magnetometer fixing.

The Phase camera FPGAs have been reconfigured and the  Phase camera channels are now received by the DAQ 

Operations done at  2026-03-04-10h47m-UTC

The two smart sockets in TCS room have been giving grey flags and absence of data on DMS after the shutdown. They are responsive to pings, I tried to disconnect and re-connect the network cables and to press the reset button (A) on the side. The manual then says to use the main reset switch to turn the sockets off and back on but, since they power part of the TCS configuration, it was decided by TCS team to keep everything as it is and to fix the issue at the next shutdown of the system, which is planned in the near future. Flags will be commented until then.

The main activity of this afternoon was the arms alignment/lock recovery (#68809); activity completed at 19:00UTC, I left the cavities locked on the infrared.

SBE
SDB2 angular and position controls opened at 15:25UTC;

Vacuum
SDB2 minitower venting started at around 15:30UTC;

I recovered quite all the FDS subsystem after the switching on of the hardware that was carried out by Fiodor.

1) Check if all the hardware was online, epsecially all the things connected via ethernet. This was done with the ping command and I checked: sqb1, sqb2, fcim and fcem mot switches, all the EQB1 PDUs, the 4 DDS board in DET Eroom and the phasecam1 PC.

I asked to Fiodor to switch off-on again all the Devices that were not responding: g25squeeze4 and SQB2 motswitch

2) I asked to Fiodor to restore the ethernet configuration of all the R&S power supply (6 in total)

3) Once all the hardware was on I did a trend of the main actuators set point: GR M5 and M7, SQB1, FCIM and FCEM MAR_TX and TY, EQB1_HD_M6 and I restored the main set points

4) I managed to close easilly the control loop of FCEM, for FCEM I had to ask to Fiodor to switch on again the SLED for optical lever

5) I had to recenter SQB2 in all the DOFs whereas SQB1 only in vertical

6) Once recovered all the suspended benches I could align the reflection of FC and close the galvo loop. After that I recovered also the FC flashes on FCEM GR PD

7) I managed to close CC PLL 

8) For SC PLL I had issues in switching on the PyServer. This was solved restoring the communication using an old interface without the PyServer and once the USB communication was restored I managed to switch on also the PyServer

9) Once the CC PLL was locked I asked to Fiodor to realign the MachZehnder in the squeezer. After that I managed also to realign the HD detector and close all the HD loops (CC, CCCoarse and AA) I measured 8.7 dBof ASZ and 5.8 dB of SQZ. With slighly different phases respect to the past 1.15 for ASQZ and 2.6 for SQZ. I had to unclip the beam in retroreflector with 380 steps in V and 600 in H.

Still missing Check alignment of SC into OPA, Lock Main PLL (still beat note not found, will be ckecked tomorrow) and lock the FC on green once MAIN PLL is recovered

We recovered the arms of the interferometer after the blackout.

North arm's recovery was quick, but West arm took many hours.

By aggressively misaligning the BS and following with the TMs, we managed to reach an alignment position that allowed the arm to lock and the dithering loop spent the next few minutes readjusting all components' alignment.

We leave the ITF with the arms locked.

The laser of the WE PCal was visible by the WE Camera. It was suspected to stop the WE arm from locking, so at 16h42m49s utc, the laser was switched off.

Two more actions done this morning:

1) turned on the SLED for FCEM local controls around 7:30 UTC

2) improved MZI alignment in the AEI squeezer to recover contrast around 13:10 UTC