This morning we went in the Atrium of the Laser lab to investigate on this additional noise, in particular concentrating on the fiber EOM electronics. 

We thus looked at the spectrum of the EOM corr directly at the output of the amplifier (see fig. 1).  The spectral noise level in the MHz band showed a lot of spikes, where we normally should only have flat noise of the order few tens of nV/sqrt(Hz).

However, we noticed that, manipulating around the electronics box, we were able to remove or make appearing this noise (see fig.1 between 9h30 and 10h UTC). We reconducted this to a grounding problem, which probably appeared after the lightning that caused an unlock of IMC.

To solve the problem, we switched off the (homemade) power supply of the electronics, plugged on a different UPS wall plug, and restarted everything. As a result, the noise disappeared. Moreover, we were not anymore able to induce it by manipulating the box.

For the time being, the noise is not visibile on the RFC_I_8MHz.  

From Tuesday morning, October 20, to Wednesday morning, we worked at North End to investigate the issue of power supply channels failing after power cycling.
The affected units are Rohde & Schwarz HMP4040 four-channel power supplies used by NCAL and PCAL. On several occasions over the past few years, individual output channels have failed after a restart following a power shutdown, or after switching off the power supply using the "red" power button on the HMP4040.
Each output channel on these power supplies is independent and electrically isolated from the other channels. The units have a "red" power button to disconnect them from the 230 V mains supply and an “Output” button to enable the output channels. See photo.

NCAL

Two HMP4040 power supplies are used to power the NCAL system:

• Four +30 V channels to power the motors.

• Two +5 V channels to power the sensors.

Motors drivers

Each +30 V channel powers one NCAL motor. Each channel includes a shunt regulator (Maxon 309687) designed to absorb the energy generated by the motors during braking, thereby protecting the power supplies against potentially destructive reverse current. These protection circuits were set to 56 V, which is too high to provide effective protection for the power supplies.
The voltage at the motor controller inputs was measured during motor braking. No voltage increase was observed, indicating that the protection circuits remain inactive under these conditions. Therefore, no evidence was found that these braking phases pose a risk of damaging the power supply channels. The behaviour during an uncontrolled shutdown may differ from that observed during normal motor braking. At the end of Tuesday morning, the jumper JP1 (see the purple circle in the picture) of the shunt regulator of all NCal motor controllers have been set to the open position, meaning that the voltage is now limited to 27 V. The voltage for all motor power supply has been reduced from 30 V (NE) or 32 (WE) to 26 V in all buildings

Electrical grounding and isolation

Each motor and its associated sensors are mounted on a table suspended from the supporting structure. (photo) The following observations were made:

• The impedance between each table and the main tower structure is 100 kΩ, except for one table that is not electrically isolated (on the left-hand side, near the equipment racks).

• On the non-isolated table, disconnecting the photodiode connector located on the motor side results in a measured impedance of 50 kΩ between the table and its support.

• The GND of the +30 V rails is connected to the tables.

• The GND of the +5 V rails is completely isolated from the tables and the supporting structure. The only apparent path from the +5 V ground to the structure is through the ADC inputs used to read the sensors, since the ADC grounds are connected to the structure. During operation, a voltage of approximately 150 mV was measured between the +5 V ground and the structure.

When the power supplies are switched on using the “Output” button, some of the +30 V channels enter overvoltage protection. The affected channels vary randomly between restarts. The power supplies' overvoltage protection thresholds are set to 32.5 V, their maximum setting. No overvoltage transients were observed on the supply rails at startup using an oscilloscope. This behaviour therefore appears to be related to the internal operation of the power supplies. When the output voltage was reduced to 26 V instead of 30 V, this behaviour was no longer observed.

Finally, the protective earth and the output grounds were connected together on the power supplies, as shown by the purple wires in the picture. This modification affected the position measurements of the NCAL tables powered by the +5 V channels ( grounding-north-west.png ,  grounding-south-est.png), changing their offsets. However, it should provide a better-defined ground return path for the +5 V supply.

It would probably be preferable to electrically bond the tables to the supporting structure to improve equipotentiality around the electronics. However, this would not explain the channel failures. Since these connections could introduce untested electromagnetic compatibility (EMC) side effects, they were not implemented.

PCAL

One HMP4040 power supply (No. 32) provides the following voltages:

• +12 V and −12 V rails for two photodiodes.

• +15 V and −15 V rails for the photodiode in the optical sphere and its signal driver. The signal from this photodiode is designated Rx_PD1

This image shows how the PCAL bench is wired. Within the PCAL enclosures, all electronic equipment is electrically connected to the supporting structure. No modifications to the existing wiring appear to be necessary.

After switching off the laser, we repeatedly switched the power supply outputs on and off using the “Output” button while monitoring the PD1 signal. No abnormal behaviour was observed.
We then switched off the outputs using the “Output” button and completely powered down the unit using the "red" power button. After switching the unit back on, we found that channel 2 (+12 V) had failed.

Previous failures had affected channel 3 (−15 V) on two occasions and channel 1 (−12 V) on one occasion. The failures therefore appear to affect different channels rather than being confined to a specific output.

We also observed that when an HMP4040 is switched off using the red power button, a residual voltage of approximately 1 V appears on each output channel for a few seconds. The available current is approximately 2 mA. We do not understand this behaviour of the HMP4040. Normally, before the outputs are enabled, each channel should be disconnected by an internal relay. It is therefore unclear why these residual voltages are present, but this behavior is present on all HMP4040.

The failed power supply was replaced with unit No. 20. The protective earth connection on the front panel was then connected to the output grounds and to the rack chassis. The PCAL setup was restarted and returned to operating points identical to those observed before the intervention.

Conclusion

The measurements carried out have not provided a convincing explanation for the failure of power supply channels following a restart. We modified the GND and protective earth wiring, but there is no guarantee that this will resolve the issue. To switch off the output channels, it is preferable to use the “Output” button rather than the "red" power button.

At 09:35 UTC, the SR RH was switched back on at a power of 7 W (V = 23.13 V).

I forgot to explain what went wrong initially.

Some checks performed directly on the actuator, with the mask in place, revealed that the center of the aperture was not aligned with the center of the thermal pattern produced by the actuator during free beam propagation, which we know to be aligned with the center of the mirror.

To correct this misalignment, we raised the mask by 8 mm from its original position using a spacer.

Today, we repeated yesterday's measurements,

To have more flexibility in selecting the aperture diameter, I asked Alessio B. and Angelo E. to prepare a new mask allowing us to switch between two apertures, with diameters of 15 mm and 30 mm.

They managed to manufacture the new mask in just two hours (see Fig. 1). Many thanks to both of them and to the mechanical workshop for their prompt and valuable support!

In the afternoon, we installed the new mask and performed some preliminary checks directly on the actuator, without identifying any anomalies.

We then started the HWS-INJ measurements using the larger, 30 mm diameter aperture. This first measurement was performed during the cooling phase (20261021T1804), which explains the negative OPL variation, appearing in red in the HWS maps (see Fig. 2).

Afterwards, we switched to the smaller, 15 mm diameter aperture.

We then started a second HWS-INJ acquisition, this time during the heating phase (20261021T1928).
The beam appered visible in the HWS map.
At 18.45 UTC, the WI PAM was switched OFF and HWS measurements interrupted.
I left the ITF with arms locked on IR.

Figure 1. When the noise increase the 6MHz sideband on the RFC photodiode gets two additional sidebands at +/-1.05 MHz around it. That is the IMC free spectral range. This would indicate that the noise has a bandwidth of several MHz, that we see it at +/-1MHz away directly in the photodiode sample data, and at +/-2MHz less directly on the demodulated 8MHz signal. 

Most INJ/PSL loop have an UGF of ~100kHz or less, so it cannot be a photodiode that saturates and propagate noise in a loop to create this noise. As the loops will have some kind of low pass filtering to keep the UGF at 100kHz or less. It could be a fast actuator having saturations, as that is after the filtering. 

There are several actuators that are able to act with a MHz bandwidth:

• Master laser noise eater loop, that is fully analogic and embedded in the master laser acting on the current of the pumping diodes
• The fibered EOM, the Op Amp of the fibered EOM electronics have a bandwidth of 70MHz. There should be a monitoring output for the fibered EOM electronics, but I don't know if that is acquired somewhere
• The solid state EOM, which should have a resonant circuit for the PMC sideband generation, but also be able to have a broadband feedback to act on the laser light, as it was used for laser frequency control before the installation of the fibered EOM

The fibered EOM electronics seems to be most likely issue, as they were modified to lower their output noise but increase their risk of saturation. Maybe there is a way of understanding that by acquiring the monitoring EOM corr output. The alternative is to reduce the RMS of the signal at the input of that electronics, while keeping the same gain, which means locking the IMC with a lower gain in the loop. Either by adding a passive attenuator at the input of the fibered EOM electronics, or by adjusting the IMC ramp auto.

Figure 2 During the night there have been small fluctuations in the level of the pole at 500Hz noise while the RFC/IMC remained lock. Maybe this can give us further ideas on where the issue is.

ITF found DOWN in UPGRADING mode.
Planned activity: PAM pre-commissioning (Nardecchia)
In order to allow the planned activity I relocked ITF on the arms on IR using ARMS_LOCK Metatron node, with a realignment of NI, WI and WE mirrors. In principal along tx dof.

Activity in progress...
 

The noise reappeared after the unlock of yesterday 20/10 at 8:17 UTC (plot 1, RFC_8MHz has been band pass filter in the 10-40Hz band).

The noise appeared during the unlock of the 07/10, on the operator report it is said that the ITF unlocked because of the thunder around 18:51 UTC. 
I could not find anything that makes those two unlocks different from other unlocks in which the noise did not appear. (signals checked : IMC alignment, MC_Z, PMC_PZT, ML_TH, EOM_CORR_max...) 

This morning I tried to make a cycle of IMC unlock/lock but it did not make the noise disappear. 


EDIT : I added a channel to have an easy access to this addiotnal noise. The RMS of the RFC_8MHz signal betwenn 15Hz and 45Hz 
V1:DQ_BRMSMon_BRMS_INJ_RFC8MHZ_15Hz_45Hz_SIB2_RFC_8MHz_I 

As reported (here and there) few days ago, the LNFS installed on September 10th (SN: 06FS02-03), which at that time was the only fully working sample, appears to be too noisy for the low noise configuration of the ITF.

We thus decided to install the spare one that has been recently repaired (SN: EG-LNFS-002), yesterday afternoon between 12.30UTC anc 13.30UTC.

After setting the last modulation frequencies and amplitudes, I put back in remote control and check the noise level for the three output looking at the dfreq channels (see fig.1). It seems that noise is reduced everywhere.

We will keep tracking its behavior in the following days.

During this morning's maintenance, I installed the mask with a 20 mm diameter aperture. A new WI PAM measurement was then performed this afternoon with the mask in place (fig.1).

The main optics were aligned in the same configuration used for yesterday's measurement by Piernicola ( fig2- entry 69948). 

14.50 UTC: HWS-INJ START with a time sleep of 100 s (20261020T1650)

14.58: WI PAM ON  at 5 V with ramp of 15 min (WF 6)

At 15:58 UTC, I increased the voltage from 5 V to 6.5 V. Unfortunately, I then noticed that both MATLAB and the HWS-INJ acquisition had crashed.

The acquisition was restarted at 16:10 UTC (20261020T1810), using the reference images from the previous measurement (20261020T1650).

However, something seems to have gone wrong with the new acquisition. One of the correctly acquired wavefronts is shown in Fig. 3.

The measurement will probably need to be repeated tomorrow.

The actuator was switched off at 18:30 UTC.

AIm : replace NEN and NWN rotors, survey all rotor position from North End Building

Situation on arrival :

• NEN: box 2024-12 (led "away" from the mirror)
• NWN: box 2024-10 (led "away")
• NSN: box 2021-08 (led "Away")
• NNN: box 2021-04 (led "close" to the mirror) 

Situation after replacement:

• NEN: box 2026-16, R4-11 (led "close")
• NWN: box 2026-15, R4-15 (led "close")
• NSN: box 2021-08 (led "Away")
• NNN: box 2021-04 (led "close") 

Remark: Some tests and changes in the electronic (see logbook...). Tests on LV supply do affect position measurements ( observed between 10 :18 :44 and 10 :29 :02). Pause these tests to finalize survey in stable conditions

Survey

• 11:55 UTC : Place jigs under NSN and NEN to prepare for survey measurement 
• Measurement from survey group from 12:00 UTC to 13:05 UTC (with 2cm spacer between rotors and sphere)
• Measure horizontality of NSN and NWN plates (see pictures "HORIZ*" by pairs: first is global view, second is a zoom on the bubble )
• ~13:15 move jig and bars to NNN and NWN setups to prepare for Survey
• Measure horizontality of NNN and NEN plates (see pictures "HORIZ*" by pairs: first is global view, second is a zoom on the bubble)
• Measurement from survey group @ 13:51 UTC  with first station and 14:09 with second station on this side.
• Remove jigs and test vertical direction of sensor between 14:22 and 14:29 UTC
• Values of NCal position sensors are saved in attached excell file

Tuning

• Correct merchanically for too low position for NNN and NSN (was -1.48 for NSN, -0.8 for NNN). Done at 16:15 UTC. New sensorpositions are in same  excell file. together with the corresponding displacement.
• Correct for phase asymmetry in NEN done at 16:48 in NEB_NCal.cfg

Today we carried out the survey for measuring the position of the NCal benches @NE. We measured the point on top of the calibrators (Fig.1, 2, 3 and 4) for the two sides of the tower (Staz01, Staz02 for the East side, Staz03, Staz04 for the West side).

Tab1 reports the result of the survey in VRS coordinates, while the scheme of the survey is shown in Fig.5

Following the completion of the electrical work in the TCS chiller room and the filter cleaning carried out by Cecilia, all CO₂ lasers were switched back ON.

The main lasers were restarted at around 13.40 UTC (see fig.1).

Some issues were then identified with the auxiliary cooling system, which were successfully solved by Nicola and Cecilia.

The auxiliary lasers were restarted at around 15.35 UTC (see fig.2).

Note that the WI CO₂ beams are currently blocked on the bench to allow the ongoing WI PAM measurements.

ITF found in UPGRADING mode and DOWN state with arms locked on the IR via ARMS_LOCK Metatron node.
All times are UTC.

06:00 MAINTENANCE mode set and ITF unlocked (ARMS_LOCK set to DOWN state).
07:00 Begin of NEB power supply check (Letendre, Frappez, Lafrasse) and Ncal installation (Mours).
07:07 PAM/TCS Room activity (Nardecchia).
07:26 - 09:32 Laser Lab checks (Spinicelli, Craighero).
07:47 CEB Clean Room Level 1 PC retrivial (Margarita, Fede).
08:10 IP noise injections (Ruggi).
08:38 UTA INJ Lab set to full speed (Portata Nominale). 
09:39 UTA INJ Lab se to normal speed (Portata Ridotta).
10:01 End of Maintenance.
10:18 End of Switch-off of TCS chillers room.
13:07 TCS chilles cleaning completed (Zaza).
Restart of the TCS laser is in progress..

ITF left in UPGRADING mode and DOWN state, with arms locked on the IR via ARMS_LOCK node.

Activity during maintenance:
- Cleaning with external company at WEB and SWEB.
- HYMO maintenance at CEB.

The aim of this shift, and the subsequent ones, is to perform the fine alignment of the WI PAM and characterize it using HWS-INJ.

Last Friday, I noticed that both TCS_PAM_WI_PowerSupply and TCS_PAM_WI_pico processes were grey in the VPM. I attempted to restart them, but without success. I then asked Giulio to check the devices, and he confirmed that neither the pico driver (picocltcs2) nor the power supply (rstcs3) was responding to ping requests.

This morning, Giulio restarted the WI pico driver and pico server, restoring communication with the devices (see 69946).

After lunch, we went to the TCS room to restart the WI PAM power supply process. 

Since the WI mirror will be repeatedly heated during the PAM tests, the etalon control loop was switched off at 12:34 UTC by Madda and Francesco.

For the HWS-INJ measurement, I retrieved the positions of the optical components along the beam path used during the CO₂ actuator centering on the West CP (see 69694). The goal is to reproduce the same optical configuration, for which the position of the WI mirror center on the HWS-INJ wavefront maps has already been determined. The corresponding reference measurement was performed on September 4, 2026, between 17:19 and 17:35 UTC. The data are stored in folder 20260904T1919 (see Fig. 1).However, Francesco noticed that the current alignment differed significantly from that of September 4. Therefore, I decided to repeat the HWS-INJ measurement by switching on the CH to determine the current position of the WI mirror center on the HWS-INJ map.

% CH switching ON
I switched ON the HWS-INJ SLED at 1.42 V
ITF in WI SINGLE BOUNCE +PR aligned 
13.29 UTC: HWS-INJ START (20261019T1529)
13.35 UTC: CH ON at high power (WF 8)
13.48 UTC: CH OFF (WF 33 ) and HWS acquisition stopped

The CH appeared clearly in the HWS map (see fig.2).
%%%%%%%%%%
Then, I took advantage of the thermal transient to go to the TCS room and switch off the lasers (69955). 
%%%%%%%%%%

% WI PAM switching ON

The first measurement with the WI PAM was performed with the beam in free propagation, without any mask installed.
The voltage applied to the source was gradually increased using voltage ramps lasting several hundred seconds.

Data acquisition started at 14:57 UTC (20261019T1657). At approximately 5 V (T ≈ 430 °C), the OPL induced by the PAM became visible in the HWS maps only after piston and tilt removal (Fig. 3).

At 6.5 V (T ≈ 620 °C), the PAM-induced OPL was clearly visible in the HWS maps even without removing piston and tilt (Fig. 4).

At the end of the shift (20:00 LT), I switched off the actuator. Tomorrow, we will proceed with further characterization measurements.
I left the ITF in WI single bounce+ PR aligned.

This afternoon, between 14:00 and 14:30 UTC, all CO₂ lasers (Main and CH) and their chillers were switched off in preparation for the electrical work scheduled for tomorrow morning in the TCS chiller room (Fig. 2).

The chillers and the microphone were also covered with protective sheets to prevent dust accumulation.

This morning we did some tests to try to understand the excess of noise
Here are the GPS :

09:38 UTC - INJ AA loop open (from DSP card)
09: 59 UTC - IMC unlock (from VPM)
13:18 UTC - PSTAB open

Surprisingly the cycle IMC unlocked / locked arounf 10 am made the noise as seen on RFC_8MHz disappear. This might go in favour of some saturation in the frequency loop actuators. 
However, I checked some IMC unlocks of the previous days and the noise was present both before and after the unlock. 
I checked if there was other changes in the INJ parameters but could not spot anything obvious. 

ITF found in locked arms IR.

Below the activities communicated in control room:

• SR electrical check (Fabio);
• HYMO maintenace at WEB/MCB;
• from 8:30 UTC noise injections on the suspesions with ITF unlocked (Manuel and Paolo);
• from 12:00 UTC PAM pre-commissioning (Ilaria) with the ITF in WI single bounce with PR aligned. Activity in progress at the end of the shift.

SBE

SIB2 and SPRB vertical position recovered with stepper motors; loops properly closed.

LSC

at 12:34 UTC WI etalon disabled under request of Maddalena.

Since the access to SR tower is over, today at 10:15 UTC I removed the beam dumps from EQB1.

To prevent any unintentional change to the CO2 configuration, I put in PAUSE the TCS_MAIN node; ITF_LOCK is already paused and should not be revived before we start relocking on Thursday afternoon.

Laser frequency noise has been a major limitations during these measurements, making any conclusions on the effect on the BNS range in general, or on the HOM noise (mystery noise) difficult to make. The higher laser frequency noise is due to something that happened on Oct 7 around 19:00 UTC and above 300Hz due to the more noisy LNFS: https://logbook.virgo-gw.eu/virgo/?r=69929.

Figure 1 shows a noise budget in the starting tuned SR configuration. The fitted HOM noise level is comparable to the case of January, when the first stage of SR diaphragm was already installed, for the times when the etalon effect was making the B1p higher. It is unclear in what position the etalon effect was on Oct 14. The estimated frequency noise contribution is comparable to the HOM noise, so the fitted level of HOM noise depends strongly on whether that frequency noise estimation is correct or not.

Figure 2 shows the SR detuned configuration. The frequency noise estimation decreases, because the coupling was lower by a factor ~2 in detuned SR. The estimated HOM noise also decreases by about 20%, which is less of decrease than what was measured in March with the SR diaphragm. It is hard to draw a conclusion from that, as it may be due to many reasons. A tentative conclusion is that the SR detuning is not dramatically better at reducing the HOM noise without the SR diaphragm. At low frequency the sensitivity is much worse, because this data is in LN2 (with many dither lines on), unlike the figure 1 in LN3. 

The 2026-10-18 I found  some SBE SPRB DMS flags red with the follozing comment : SBE_SPRB_ACT_F0H0_raw loops ringing . The SBE SPRB controls was in this state  since the 2026-10-17-08h00-UTC 

I made some actions trying to recover the correct running conditions but without success : 

• actions performed between  the 2026-10-18-06h25m05-UTC and 2026-10-18-06h35m33-UTC

At the end the SBE SPRB loops were left open 

In preparation for the HWS-INJ measurements with the WI PAM, I blocked all CO₂ beams (DAS and CH) on the bench.

• 2026-10-19 08h06m03 UTC    ilaria 'WI_Bench:OUT UP [byte=2]' sent to TTL_co2
• 2026-10-19 08h06m04 UTC    ilaria 'WI_Bench:CH UP [byte=3]' sent to TTL_co2