An interesting comparison can be done with old data taken with an intermediate SR TY misalignment. Fig 1 shows the comparison when the DCP is about 270 Hz: the mystery noise is clearly lower. Fig 2 shows that the working point is quite similar, according to the DCP and the optical gain; now the dark fringe is much darker.

As we know, in the past the mystery noise was lower with a higher misalignement (fig3); now, going further with the misalignament the curve does not change in the region where the mystery noise is supposed to be limiting (fig 4). To be noticed that there is still an improvement of the dark fringe when the DCP is pushed down (fig 5).

A rough estimation of the noise components (fig 6) says that a pessimistic estimation of the thermal noise (0.7e-23  @ 100 Hz) is almost sufficient to explain the curve.

ITF found unlocked with both cavities misaligned; to realign them I had to move the PR. The ITF relocked at the first attempt in LN3_Aligned.

At 9:30 UTC Paolo made some test to acquire data with SR misaligned in different steps.

At 13:28 UTC, following Maddalena's instrunctions, I changed the Etalon setpoint:

•  NI:NI Set point [NI_RH_SET=18.66,rampTime=3600]' sent to LSC_Etalon_Acl
• WI:WI Set point [WI_RH_SET=18.82,rampTime=3600]' sent to LSC_Etalon_Acl

ITF left locked in LN3_Aligned.

Same plots again, where yellow dots now include this morning's scan of the SRM angle. It seems that the strain noise at 110÷133 Hz is flat below 300 Hz DCP frequency. And the effect above 300 Hz DCP is mostly due to optical gain, i.e. the net broadband noise in power units does not change significantly when misaligning SR. 

Starting form 9:30 UTC, for two hours some data have been acquired with SR TY misaligned in many steps, up to the usual 2 urad which sets the DCP below 200 Hz. The aligned state with SR TY in loop has been recovered at 11:35 UTC.

To continue on this cleanup exercise also the GUI applications Chassis and CenteringUI have been removed from the SUS IceWM menu since not working or not in use anymore.

At the beginning of the shift I found the ITF in LOCKED_ARMS_IR, TROUBLESHOOTING, with inj team working on the RFC aligment
15:30 UTC beginning of relocking activity: west green laser's thresholds adjusted (Bersanetti), high sea activity  SR guard opened, NI F0 H3 had to be adjusted (Palaia). ITF kept unlocking between LOCKED_ARMS_IR and CARM_NULL_1F. After a thorough investigation by Bersanetti, Pinto and Was it was necessary to move SR by +10 urads tx and +5 urads ty and the ITF locked
21:00 UTC ITF in COMMISSIONING, LN3_ALIGNED
21:37 UTC unlock
22:00 UTC end of shift, ITF left relocking with AUTORELOCK_FAILSAFE

On-call intervention as the H3 coil on NI F0 was approaching saturation (Sa_NI_F0_COIL_H3 value around −9 V).
After the intervention, the correction values were restored to about 0 V.

• Start time: 18:25 LT
• End time: 18:40 LT

We made a characterization of the noise changes related to the replacement of the UPS at the NEB occurred on Dec. 9 2025, https://logbook.virgo-gw.eu/virgo/?r=68330.

The noted changes are very similar to what noted for the similar replacement of the UPS of WEB occurred on July 1st 2025, and described in the elog "Noise measurement of the new UPS units installed in the NEB".

We noticed:

• the largest noise change is seen in the UPS monitors, magnetometers see tiny ones
• Figure 1, Figure 2: a peak at 1.4 Hz and +-1.4 sidebands in odd 50Hz harmonics disappear form voltage and current monitors, some reduction at 1.4 Hz is seen as well by magnetometers
• Figure 1: the low frequency (below 200Hz) part of UPS Voltage and currents monitors clean significantly, up to the point that current monitors are now dominated by sensor noise at all frequencies
• Figure 3: the high frequency part does't clean much, and actually some noise increase in visible.
• Figure 4, 5, 6: two narrow peaks at 50Hz +-6Hz (Figure 4) and slowly oscillating in amplitude (Figure 5), appear in the Voltage monitors: these peaks were present also before but just in the IPS monitors (Figure 6)

Note: UPS Currents Probes need to be replaced with more sensitive ones.

Yesterday evening, for no apparent reason, the SL switched off.

This morning we went to EER to investigate. The safety chain had been triggered; after resetting it, the SL switched on without any issue. It then took some time for the system (SL, AMP, PMC) to thermalize, and some power was initially missing at the PMC and IMC outputs.

While attempting to speed up the recovery by increasing the power with IPC1, we noticed that we could not reach the nominal 18 W in IMC transmission. We suspected we were limited by the IPC1 transmission power, especially considering the PMC transmission decrease observed in December (#68408). We therefore went to the laser lab to recover the power discarded by IPC0.

We managed to increase the PMC transmission by about 10%, although this came with a 20% increase in reflection. As a result, the IMC transmission rose slightly above 19 W.

We then attempted lock acquisition again and realized that the limitation was not due to IPC1 transmission power. Instead, the IPC1 rotator server was not responding. We went to Piscina, rebooted the rotusb1 server, restarted the EIBAgilisRot server, and IPC1 resumed normal operation.

During lunch, we continued with lock acquisition, but INJ was quite unstable, most likely due to the PMC loop not being retuned after the power increase.

Since IPC1 power was not the issue, we decided to (approximately) revert the power previously recovered from IPC0. While there, we realigned the beam inside the PMC. After realignment, we measured the higher-order modes:

• TEM20 ≈ 20/800 mV/mV (~2.5%)

• TEM30 ≈ 30/800 mV/mV (~3.7%), which is unexpectedly high.

At this point, triggered by this TEM30 and noticing that PMC transmission, IMC error signal and PSTAB were quite noisy (approximately a factor 3 increase on PMC TRA, figure 1), we decided to try to improve the situation by adjusting the Neovan pumping diode currents.

We changed the Neovan pumping diode currents from (5.2 A – 5.0 A) to (5.0 A – 4.9 A). This produced a significant improvement in overall system performance (figure 2)

• AMP output power decreased from 0.375 V to 0.36 V

• PMC transmission increased from 0.775 V to 0.782 V

• PMC reflection decreased from 0.17 V to 0.14 V

• Normalized PMC transmission increased by ~3%

• Normalized PMC reflection decreased by ~15%

• Additionally, the Neovan head temperature decreased by 0.6 °C

Overall, this adjustment seems to have a clearly positive impact on system stability and noise.
We keep those values and went on with the planned activity which was the RFC realignment (dedicated entry)

After replacing the communication bridge to the PZT driver (#68586), we attempted to realign the RFC using the two PZTs directly from the VPM.

Both drivers are now in online mode and can be controlled directly via the VPM. This is done through the PiezoApp server in the INJ section of the VPM.

The channels used to move the PZTs are:

M15_axis_x
M15_axis_y
M16_axis_x
M16_axis_y

The status can be monitored through channels such as INJ_M15_axis_x_pos, INJ_M15_axis_x_Reqpos, etc.

We explored both axes of both PZTs to maximize the coupling. Giving the drift of the past months, we hoped not to be limited by the 120 V range limit of M15_y, as has historically been the case; however, this limitation still applies.

Nevertheless, we managed to improve the RFC TRA from 2.7 V to 3.4 V.

ITF found in Commissioning mode and locked in LOW_NOISE_ALIGNED.

Under request of the commissioning team at 14:54 UTC I launched the script inject_asc_test.pyat 15:17 UTC the script inject_lsc.py.

After that the ITF was manually realigned and relocked in LN2 for ITF characterization.

Under request of Julia at 19:30 UTC I manually unlocked to relock the ITF in LN3_Aligned.

Unfurtunately at 19:50 UTC, during the lock acquisition the Slave Laser switched off.  I contacted the INJ on-call ; I went to the EE room and, through a video call, we verified that there were no evident damages or potentially dangerous situations. It was decided to leave the system in this state; the INJ experts will investigate tomorrow morning.

ITF left in DOWN and in Troubleshooting mode.

Since the recovery was already finished by the morning, we started with some characterization of the interferometer. The operator (Berni) launched one set of angular noise injections and one set of longitudinal noise injections at the beginning of the shift. Then Piernicola went to turn on and off the power supply for EDB. After his action we unlocked and relocked in LN2.

Then we started with the SRCL SET scan. We put the BS in full bandwidth and we opened the SRCL_SET loop. Then we starte a scan towards lower SRCL offsets. We waited 20minutes between each point.

• Reference time: 1454862497 – SRCL_SET_TOT  = 40.7 – DCP = 424
• 1st point (-5, START 16:34:32 UTC): 1454864091 - SRCL_SET_TOT = 35.8 – DCP = 432
• 2nd point (-10, START 16:57:10 UTC): 1454865528 – SRCL_SET_TOT = 31.2 – DCP = 422
• 3rd point (-15, START 17:23:03 UTC): 1454867052 – SRCL_SET_TOT = 25.5 – DCP = 423
• 4th point (-20, START 17:47:32 UTC): 1454868509 – SRCL_SET_TOT = 20.5 – DCP = 414.5
• 5th point (-25, START 18:13:07 UTC): 1454870228 – SRCL_SET_TOT = 15.5 – DCP = 410

Then I put the offset to 0 with a ramptime of 20 minutes.

After Piernicola restarted the EDB pico-motor driver I have realigned the EDB OMC between 17:30 UTC and 18:00 UTC. The power of the 56MHz USB increased from 0.2mW to 0.6mW on EDB B1t. It back to the same value it was two weeks ago, and during a scan I couldn't see clearly the amplitude of the order 1 mode of the 56MHz USB, which means it has at least a factor 10 less power than the TEM00. I have turned off the pico-motor process on VPM.

I have left the EDB OMC scanning with the standard parameters freq 0.0005Hz, ampl 0.5, offset 23.27 

Today we realized that the Hameg power supply PowerUnit14 of SDB2 is sending a current to the picomotor drivers. It appears that the picomotors drivers are powered since January 27th, 15h09 utc. At that time there was an attempt to start the picomotor process but it failed after powering on the picomotor drivers.

We decide to leave the situation as it is for the moment, to avoid creating another thermal transient.

When we will want to fix this issue, we will also need to recover the communication with PowerUnit11 which is currently out of reach.

Same plots including early data after the second diaphragm installation from yesterday ad today (yellow dots). The effect of smaller diaphragm is quite evident.

ITF found relocking up to CARM_NULL_1F in COMMISSIONING mode.
The planned NI Cam reframing activity (carried out by Boldrini) went on without problems.

Lock acquisition is working fine up to LN3_ALIGNED.

Michal noticed that the B5 beam was getting close to the edge of the B5_QD2 quadrant when the SDB1 bench was aligned with the B1_DARM dithering signal. This is a consequence of having half of the B5 beam clipped by the SR diaphragm.

In order to solve this issue we locked the CITF and then actuated on the B5_M4 TY picomotor (by +3450 steps) to zero the B5_QD2_H signals. This action was completed at 13h22 utc.

In addition, Michal has introduced an offset in the horizontal and vertical corrections applied on the SDB1 marionetta, to reduce the shaking of the bench that occured during the transition from low to high bandwidth control (which is due to the resetting of the filters). These offsets are +4.6 V on SDB1_MAR_TX and +1.3 V on SDB1_MAR_TY. They are implemented since ~12h30 utc.

Today around 12h30 utc the DET_MAIN node could not reach the SHUTTER_CLOSED state due to a too large positive offset on B1_PD3 (about 0.14 uW). We reduced the value of the offset to bring the B1_PD3 signal close to 0, which solved the problem encountered by DET_MAIN. It is likely that this offset is still going to evolve in the next hours and that we will have to perform again this type of correction.

SDB2_dbox_bench configuration reloaded at 12h50 utc.

Comparing the sensitivity with aligned SR in different conditions might not show clear outcomes, as the noise depends on the double cavity pole frequency wich is not always stable. However, comparing data with the same DCP frequency may also give unclear results, as the sensitivity depends on the optical gain too. 

In the attached pictures I plotted some data with aligned and misaligned SR, both before and after the first SR diaphragm installation. Different colors refer to:

- red points: October 2÷4 2025, GPS time range 1443422300 - 1443622200

- green points: December 3÷5 2025, GPS time range 1448779100 - 1448979000

- black points: January 8 2026, GPS time range 1451905220 - 1452105120

- blue points: January 14÷19 2026, GPS time range 1452456018 - 1452955918

- brown points: January 23÷28 2026, GPS time range 1453147218 - 1453647118

Each point is an average over 100 s of data. All sets contain data with aligned and misaligned SRM, except those from Jan 14÷19 when SRM was always aligned. 

The first plot shows the Hrec BRMS noise in the 110÷133 Hz band vs the DCP frequency. Noise with misaligned SRM is similar in the various data sets, while noise with aligned SRM has a large spread; the difference between different data sets is much clearer when plotting the noise vs optical gain, see second picture.  

Assuming the broadband noise scales inversely with the optical response, the last two plots show the net broadband noise in relative power units, computed from the Hrec BRMS in the 110÷133 Hz band by:

a) quadratically subtracting the quantum noise (computed from measured optical gain, double cavity pole frequency, and B1_DC power) and Paola's estimate of thermal noise;

b) multiplying by the optical response  (computed from measured optical gain and double cavity pole frequency)

c) dividing by the sqrt of the band;

From the plots in relative power units it appears that after the first diaphragm installation, SRM misalignment reduces noise by a smaller factor after SRM diaphragm installation in  January then before.

During the evacuation of SR tower, a test on a vertical control of SR base ring has been carried on.

The base ring is the suspension stage placed below the inverted pendulum legs. The stage is equipped with three vertical position sensors (LVDTs) and three actuators (piezos); in principle, it can be use to provide an active pre-isolation of the suspension for three dofs (the vertical one, and the two rotations along the horizontal axes). For the control of the rotation, the inertial rotation sensor is missing, but this is not the case for the vertical, because the suspension itself is a very good inertial sensor.

SR is the only tower where also the driving board of the piezos is installed; a few years ago the sensing and the driving system were prepared for the use and some preliminary loop was closed, just to check the correct functionality of the system. In the past days the work has been finalized and a first result in terms of platform inertial control has been obtained.

The working principle is based on the use of the vertical LVDT, measuring the internal motion of Filter0 attenuation stage, as an accelerometer. A simple reconstruction/calibration, taking into account the mechanical response of the suspension, is required. Then, the usual strategy of blending the accelerometer with a ground-based position sensor (the base ring vertical position, measured through its own LVDTs) can be applied and a loop can be closed on the blended signal. The expected effect is that the top stage internal motion will be suppressed by moving the base ring in the opposite direction of the ground, providing in that way an inertial platform.

Fig 1 shows the three signals when no vertical control is applied:

- blue curve: the ground motion (displacement), obtained from the velocimeter installed in the central building.

- red curve: the motion of the base ring with respect to the ground. The signal is very low because the stage is quite rigid (first resonance 26 Hz).

- yellow curve: the ground motion provided by the reconstructed F0 accelerometer described above. The compensation of the mechanics is good enough, even if not perfect. A remarcable difference between the seismometer and the SUSP accelerometer is visible below 0.1 Hz: SUSP ACC is a very heavy and soft accelerometer, perfect for having a very low noise at low frequency.

The forth curve is taken from the LVDT installed in the second vertical stage of the suspension. It remains out of loop when the others are controlled and can be used for the evaluation of the control performance. In open loop condition, one can see the 6 resonances of the suspension and a steep attenuation starting from about 1 Hz, which puts soon the residual motion at higher frequency well below the sensor noise.

Fig 2 shows the same signals when the control of the base ring is closed. The blending between position and acceleration is about 40 mHz; the UGF of the loop is about 3 Hz. The base ring spectrum copies perfectly the seismic noise. The in-loop signal (yellow) is suppressed up to a factor of 40 and it is low also in the critical region around 0.1 Hz, where normally it is more difficult to have a good performance. The out-of-loop signal (purple) shows small and narrow peaks as residual motion above 0.1 Hz.

Fig 3 shows the same signals when the standard control of the topo stage is closed. The performance visible in the purple signal is much worse at 0.15 Hz. The peak corresponde to a notch in the plant of the control from the top stage, where the perfomance is expected to be lower. The mechanism of noise reintroduction needs to be better studied and understood. This is a lack of accuracy normally present in the vertical control of all the suspensions, but on SR is particularly bad. No attempt of loop optimization has never been performed and maybe it would be better to do something, but after having seen the performance of the base ring control, I think it would be worth to go ahead in that direction.

In order to finalize the development of the new strategy, its impact on the horizontal control needs to be better observed. The actuation of the base ring has been balanced as better as possible, in order to avoid any rotational disturbance, which would become fake acceleration affecting the IP inertial control. The performance needs to be checked when SRCL is locked, because this is the most sensitive condition to observe any additional noise coming from the controls. We need also to check the control behaviour in heavy conditions in terms of seismic noise, because the dynamyc of BR actuation is much more limited than the actuation on F0. Very likely, an earthquake would saturate the control, causing the unlock. An early worning strategy should be applied in order to move in time the control back to the more robust one.

today the actuation on the SR payload has been balanced. in fig.1 and 2 the trend data and the correlation wrt to SRty opLev signal and BNS fluctuation before and after the action (blue and red, respectively). A slight improvement is visible.

Today around 9:50 utc, I restarted NCalmoni to update the hInj values of the NCal according to the WEN-NEN and WWN-NWN exchanges done on 3/2/2026 (https://logbook.virgo-gw.eu/virgo/?r=68617).

Note that between 3/2/2026 and thi restart, NCalmoni used wrong hInj values for WEN, NEN, WWN and NWN

I have reframed and refocused NI_Cam,

I only adjusted the last articulation of the camera's mount, leaving the yellow stand that also hosts the thermo camera immobile to the best of my ability.

The image looks fine now.

This morning around 6:00 UTC I have reload the SDB2 dbox configuration to change the B1 PD3 offset, which is now close to zero. I have also undone the changes in SDB1 B5 QD2 offsets done last night, as there were many unlocks during at the end of CARM NULL 1F during the night, and at the end of the CARM NULL 1F SDB1 switches to high bandwidth control which kicks the bench and may be moving the beam outside of the quadrant.