While we were retuning some step of the lock acquisition, the handoff of DARM to the IR was constantly failing due to an oscillation around ~70 Hz. We realized that this could be due to the missing structure at 68-72 Hz (fig.1), that we compensated in the DARM/CARM control filter while on the beating. Maybe its absence during the handoff was causing some stability issue.

We modified the controller in order to remove the resonant gain. After such modification, we succesfully performed a smooth transition to B1p 56MHz error signal.

We sligthly adjusted the beating control filter in order to recover the bandwidth and stability margins of yesterday, in order to cope with the 68-72 Hz structure, which was eating a lot of phase reducing the available margins. After the modification we recovered the situation of yesterday (performance wise). In fig.2 the comparison of CARM and DARM signal and correction spectra of yesterday (blue) and today (red) while locked on the beating.

We still don't know what is the structure of fig.1, but we know that we need that

:)

The goal of this commissioning task was to get the right configuration - arms resonant on IR beam with MICH in half fringe (HF), to get the PC reference images. This state, LOCKED_MICH_HF_DC, has been automated in the virgo system (virgoDev - Automation - userapps). It derives from the arms locked on IR (metatron state 40). Once this is achieved, the only d.o.f. to be controlled is the MICH. The LOCKED_MICH_HF_DC which is the desired state and its locking processes lives in the DRMI_LOCK.py and initialization of the locking process (gains, filter, weights, etc.) lives in DRMI_LOCK.ini.

To control the MICH d.o.f. we need the right error signal which can be used to lock when in the optimal configuration (half-fringe). The error signal used for MICH is the combination of the signals acquired at B4 and B1p. These signals are given by:

1. SPRB_B4_DC
2. SDB2_B1p_DC

which are in a.u. and something we don't care about. Since we know the min and max of these signals we can try to optimize the error signal for our configuration using dedicated weights. The error signal for the MICH d.o.f. is given by the combination of the afforementioned signals:

LSC_B1p_SUM_NORM = SDB2_B1p_DC/(SDB2_B1p_DC + w*(SPRB_B4_DC))

where w is the weight for the B4 DC signal. This error signal should range between 0-1 which triggers the lock on the MICH d.o.f. when it is in the desired range. Figure below (pc_calib3) shows the error signal channels and the signals which it is derived from. To get the perfct half-fringe this should be 0.5 (given it ranges between 0-1). In the current algorithm, the locking of the d.o.f. is triggered when the error signal ranges from 0.4-0.6 (range). Once the error signals were defined and the process automated, we were able to achieve a relatively stable lock for the configuration. One of these locked state of this configuration along with the MICH trigger and MICH correction is shown in figure below (mich_hf_lock4). The trigger being 1 means the target d.o.f. is locked. Also in the plot we can see the correction signals, of the MICH d.o.f. which acts on the BS. We use the BS to control this dof. The range of the correction is +/-10 V. As we can see that there is a lot of correction that is being applied to keep the dof locked. The error signal also fluctuates a lot between 0.45-0.55 which means we cannot narrow down our range of the error signal as it will loose lock once the threshold is crossed. Apart from these signals also plotted are the B7 and B8 DC signals representing the power transmitted through the arm cavities. The SPRB B4 and SDB2 B1p PDs are the powers (a.u.) measured in bright and dark port of the interferometer. The lock acquisition of the MICH dof shows a ramp (figure: mich_hf_lockstep3) which is not clearly understood why. The trigger should work as relay switch (0 or 1). 

In the next days we will also elaborate on the reults from noise injection on MICH control loop used in this configuration. 

ITF found in LOCKED_ARMS_IR and COMMISSIONING mode

06:00 UTC ADC7674_SN13 in the TCS room unreachable : problem solved by changing SPF connector on TCS ADC7674_13 (Masserot, Zaza, Kraja)

07:30 UTC Recombined configuration (Arms@IR, MICH half-fringe) (Bothra, Bersanetti)

08:57 UTC earthquake in Marina di Pisa, suspended benches recovered

13:30 UTC CARM/DARM handoff to IR (Bersanetti)

Quick recap of the past activity done on NI and WI payloads:

Over the past days, after the NI and WI payloads installation, some activity concerning the optimization of the driving has been performed. We focused on reducing the coupling among the longitudinal and the angular corrections at the level of the marionette.

North Input payload:
Concerning yaw (ty) and roll (tz), the usual procedure to estimate the driving coefficients has been performed. Iterative DC corr steps in the longitudinal DoF have been applied in order to estimate the correct driving coefficients which minimized the couplings among the different DoFs.

For what it concerns the coupling between length and pitch (tx), other than DC constant coefficient (same as for ty and tz), above about 100 mHz there is an additional non negligible frequency dependent component which is mainly related to the mechanics of the payload, which couples the pitch rotation of the system for a given longitudinal force applied to the marionette.

This mechanical component is usually implemented as a filter filtering part of the Z correction going in the tx correction branch (zC-txC).
In order to estimate the new zC-txC filter we injected noise separately in the Z driving and TX driving, in different working conditions in order to estimate differences within the angular spectra for different driving implementation. Namely, relevant GPS are the following:

gpsZc=utc2gps(2026,08,05,14,05,00); durZc=2400; % noise zC in WI and NI with old compensator zC_txC
gpsZcWI=utc2gps(2026,08,05,14,48,00);durZcWI=1000;  % noise zC in WI with NO compensator zC_txC
gpsZcNI=utc2gps(2026,08,05,14,48,00);durZcNI=2400;  % noise zC in NI with NO compensator zC_txC
gpsNItxc=utc2gps(2026,08,05,15,31,30);durNItxc=2400; % noise TXc in NI
gpsWItxc=utc2gps(2026,08,05,16,40,00);durWItxc=1000; % noise TXc in WI
gpsZcNI2=utc2gps(2026,08,11,14,35,30);durZcNI2=900;  % noise zC in NI with NEW compensator zC_txC

The zC-txC filter is extracted by computing the ratio between the two measured transfer functions of TX opLev response over Zcorr and TX opLev response over TXcorr during the several noise injections.
Results of the measurements are reported in fig.1, where in blue is reported the overall zC-txC transfer function, from which the DC flat coefficient has been obtained, while in red the consequent residual frequency dependent filter, to be fitted and implemented, is reported. In fig.2 the fit of the compensator is shown.
In fig.3 are reported the comparison of the NI TX opLev signal in different time stamps, during Z corr injections, i.e. the initial condition with the older driving compensator (blue), the same signal with the compensator switched off (red), response with the new implemented global driving (both tx, ty and tz), with only the DC coefficient for txC applied (yellow). Final response with the final driving configuration with also the zC-txC frequency dependent compensation implemented, is reported in the purple trace.

Both in the yellow and in the purple plots, the coupling reductions is evident with respect to the initial conditions, apart from mainly the region of the pendulum resonances, namely 0.4 and 0.7 Hz.

For reference, in fig.4 and 5, response of NI TY and TZ local control signals at the same gps time stamps are reported.

West Input payload:
Driving optimization has been performed as well for the WI pay. Result of the activity can be appreciated by looking at the clean data, while locked on the arms, in particular the TFs between the angular corrections and the longitudinal one before (blue) and after (red) the activity, see fig.6.
Concerning zC-txC frequency dependent compensator, same measurement done for NI has been done also for WI, by obtaining similar response to be fitted (fig.7). This compensator has not been implemented.

In the next future, we will explore the option to investigate further the length 2 pitch coupling, in particular in the region of the resonances, if there is a way to improve it.

We went to investigate in TCS room and DAQ room and found the following:

TSC room ADC7674_SN13 link0 : TX light red instead of green

DAQ room , Rack 15, MxDx SN12 link21 : RX arrow (DOWN arrow) ligh off instead of on

Since we already were in DAQ room, we first unplugged/replugged the fiber and tried to change the SPF connector, with no results. Moving on to TCS room, we changed the SPF connector with a spare one, the TX light went back to green and the data were restored shortly after.

An earthquake (3.7 mag) coming from Marina di Pisa unlocked the INJ system this morning at 8.57 UTC as EIB started to shake a lot. After some time the system recovered by itself. 

Some settlement earthquakes could disturb the system soon.

Cecilia fixed the issue by making the investigations and by replacing the fauly optical transceiver at the ADC7674 SN23 level  .

Operation performed at  2026-08-12-07h28m38-UTC

Since 026-08-12-01h12m40-UTC the ADC7674_SN13(TCS_ADC_Moni0), located in the TCS room, is unreachable via the Tolm network .

As consequence, the follwing servers became blue in the VPM interface: SUSP_SBE_Tpro, SUSP_Fb, TCS_CEB  and PyTCS 

Below the report  of the SUSP_Fb Tolmframebuilder, collecting the data sent by the TSC_ADC_Moni0 :

2026-08-12-01h12m37-UTC>WARNING-[TolmFrameBuilder::ControlMerging] frame: 1470532375.600000000: merging is triggered on timeout (internal 1470532375.962515050 > max 1470532375.960000000)
2026-08-12-01h12m37-UTC>WARNING-[TfbSubFrame::Fill] GPS: 1470532375.600000000 - Source TCS_ADC_Moni0_0: 452/2000 missing packet(s), 0 data break(s) (missing data are at the end of the vector)
2026-08-12-01h12m38-UTC>INFO...-[TfbSourceManager::Clean] source TCS_ADC_Moni0_0 (1595:v05) has stopped -> waiting for data...
2026-08-12-01h12m38-UTC>INFO...-[TfbSourceManager::Clean] source TCS_ADC_Moni0_Monitoring (1601:v01) has stopped -> waiting for data...
2026-08-12-01h12m38-UTC>INFO...-[TfbSourceManager::Reporting] Sources reporting: / nSrc: 57, missing(2): TCS_ADC_Moni0_0, TCS_ADC_Moni0_Monitoring
2026-08-12-01h12m38-UTC>INFO...-CfgReachState> Active(Active) Ok

The faulty ADC7674 SN13 is located in a VME 21slots crate with the 3 others TCS ADC7674 boards.

The 3 others ADC7674 data are correctly received, so someone should check first the optical transceivers at boths sides

• TSC room ADC7674_SN13 link0
• DAQ room , Rack 15, MxDx SN12 link21 

ITF found in LOCKING_ARMS_IR State and COMMISSIONING Mode.
All times are UTC.
07:50 - 08:20 INJ: IMC OLTF (Melo, #69545). 
08:31 - 08:51 DET: OMC Lock in SINGLE_BOUNCE_NI, locked from 08:44 to 08:49 (operator, see Fig 1).
08:52 - 09:13 DET: OMC Scan in SINGLE_BOUNCE_NI (operator, see Fig 2).
09:13 - 09:17 INJ: Chiller refill (Melo, #69549).
09:27 - 09:42 DAQ: Solving timing issues on SQZ DaqBoxes (Masserot, Pacaoud, #69548).
09:37 - 10:12 TCS: Thermal Camera References (operator).
09:20 - 09:40 INJ: RFC OLTF (Melo, Ruggi, #69546).
09:50 - 10:11 SQZ: Block SQZ beams towards SQB1 (Sorrentino, 69550).

The work on CARM/DARM handoff to IR went on during the afternoon. 
TCS: Power Checks skipped (requested by experts).

ITF left in COMMISSIONING Mode and LOCKED_MICH_HF_DC State with work on recombined configuration (Arms@IR, MICH half-fringe) (Bothra, Bersanetti) ongoing.

After the recent Free Space implementation, today we tested new control filters for CARM and DARM while locked on the beating signals. The idea was to make more robust the lock on the green by increasing the bandwidth of the loops.
In fig.1 is reported the comparison of the OLTF of DARM (for reference) with the older controllers (blue), the extracted model (red) and the new open loop TF with the new controller. We were able to increase the accuracy pushing up the badwidth to ~25 Hz (G.M. of a factor 2), by adjusting the roll-off in order to not renintroduce too much correction noise. To be noted that there's still margin to increase the bandwidth without occurring towards corr saturations, in case it is needed (see fig.3).

In fig.2 is reported the plant measurements. Both in DARM and CARM a notch strutcture at around 70 Hz was present. After some digging in Acl we found that in CEB_ALS process, filters for CARM and DARM (double structure at 68 and 72 Hz) are implemented in order to compensate some INJ structure (?). During the controller design we compensated this structure with a resonant gain at the same frequencies and Qs, so to gain more phase at that region.

In fig.3 are reported the first comparison of DARM and CARM signals and corrections spectra before and after the implementation of the new control filters. In order to reduce the bump visible at ~50 Hz and to reach the declared factor 2 of gain margin, we slightly adjusted the loop gains accordingly.

N.B.

modification performed (automation side) are the following:

• DARM/CARM control filter: ALS_Arm_control4 -> ALS_Arm_control5;
• carm_gain_boost_final: 0.006 -> 0.0138;
• darm_gain_boost_final: 0.018 -> 0.0138;
• darm_line_ampl: 0.07 -> 0.03;

I noticed that the pixel size of the camera SDB2_B1s2_Cam set in the configuration of DET_Img_CEB is wrong. It is currently defined as 5.2 um (the pixel size of the Smartek CMOS  GC1281XM-S90-NoCG). However this camera is not a CMOS one. Instead this camera is a Smarteck CCD GC1392M-A90-NoCG, for which the pixel size is 6.45 um.

The consequence of this mistake is that the beam size extracted from the camera images is underestimated by a factor 5.2/6.45 = 0.806 (almost 20%).

This mistake is probably present since 09/07/2019 when the B1s2 camera was replaced ( https://logbook.virgo-gw.eu/virgo/?r=46309 ).

We will fix this mistake next week when we will have the opportunity to restart the process DET_Img_CEB.

This morning around 10 UTC I put two dumpers in front of the output windows of the EQB1 bench. This will prevent any beam to reach SQB1.

There was the following DBoxes with the timing_error not at zero , this appeared around 08h50m-UTC for most of them:

• EDB_DBOX_OMC (SN075) and EDB_DBOX_SQZ(SN120) managed by the EDB_QD_dbox_rack server
  • Related to EDB_DBOX_SQZ(SN120), the demodulation board SN38 (mezzanine3) appeared as not available for the DaqBox server, but was weel reachable from the DBox ARM CPU
    • from the DBox ARM CPU; the command /home/daqbox/virgo/App/DaqBox/v8r3/scripts/daqbox_main_fpga_reset.sh  was executed and after the the faultymezzanine was recovered
  • The reconfiguration of these DBoxes allowed to recover the synchronization with the Timing system
• SQB1_DBOX_Up (SN103) and SQB1_DBOX_Up(SN104)  managed by the SQB1_dbox_bench server
  • the LC and SBE SQB1 controls were openned 
  • The reconfiguration of these DBoxes alloweed to recover the synchronization with the Timing system
  • the LC and SBE SQB1 controls were successfully closed after these operations
• EQB1_DBOX_03 (SN091) managed by the SQB1_dbox_rack server and the SQZ_DBOX_O2 managed by the  SQZ_DBOX_DET_EERoom server
  • The reconfiguration of these DBoxes allowed to recover the synchronization with the Timing system

After these operations, all the faulty DBOxes are synchronized  as well the FAST_DAC mezzanines . The attached plot show the trend of the issues 

Today, we verified the OLTF of the RFC, which was 200 Hz and 0.5 rad of phase margin (See Fig. 1). However, this loop should have a unit gain at 180 Hz ideally. So we changed the RFC gain on the DSP card from -0.25 to -0.22. The new OLTF is shown in Fig. 2:  the unit gain is now at 178 Hz and the phase margin is now at 0.62 rad. The gain margin is now 0.5 and it was 0.6 before.

Ps. to be noted that if the DPS card is reloaded the RFC gain will be reverted to -0.25 again.

Today I measured the OLTF of the IMC. Currently, we have 110 kHz of bandwidth and 20 degrees of phase margin.

ITF founf in LOCKED_ARMS_IR in COMMISSIONING mode.

planned activities

- Improve CARM/DARM handoff to IR (TBC)
- Implement new filters for ALS (TBC)
- Block SQZ beams towards SQB1 (Sorrentino, TBC)

During the shift P. Ruggi noticed a strange behavior of NI Etalon (plot1). After a check, Dattilo and Cavalieri decided to replace (at 08:45 UTC) the old KERT with a new one.
ITF unlocked due to an earthquake in Colombia (fig attached).  SIB2, SNEB, SWEB, SDB2, SPRB SBE loops opened. Loops closed. Relocking in progress...

TCS
14:00 UTC - Post heating loop switched ON, on AUX cooling system (Dattilo, Cavalieri).

This afternoon also the TCS WI AUX laser has been turned on again, and left as the other ones ON for monitoring and thermalization.

The CO2 laser status is as follows:

• NI bench:  
  • AUX LASER:  ON
  • MAIN LASER  ON
• WI bench:
  • AUX LASER:  ON
  • MAIN LASER: ON

All laser beams are blocked inside the benches.
Do not open the acoustic enclosures.
If access is required, please contact Marco and Ilaria.

Another quick look to the acoustic noise peak around 490Hz:

this one is not associated to the INJ HVAC, but switching it off (tests of July 28) caused the frequency of the peak to drift up: see Figure 1

the LB and EIB benches detect a correlated vibration, which is louder in the EIB (vertical accelerometer) than in the LB (horizontal acc.): Figure 2

Browsing back the VIM, the peak seems present since long time, but recently it has changed frequency and became more intense:

• it was around 550Hz and fainter up to Feb 28: Figure 3
• after the March black-out it moved to 400Hz and became more intense: Figure 4
• on May 4 it drifted up to 500 Hz: Figure 5

   Hints may suggest for something located inside the INJ lab, probably on EIB, which is getting more noisy... maybe because of ageing?

ITF found in LOCKED_ARMS_IR in COMMISSIONING mode.

planned activities communicated to the ctrl room and performed:
- Camera installation on MC building (Fabozzi, external company) concluded at 10:20 UTC.
- OMC lock in NI single bounce (Gouaty, Romero) concluded at 10:00 UTC.
- DRMI recovery (w/o TCS) concluded.
- TCS - CO2 WI laser recovery (Nardecchia, Cavalieri, Gherardini) concluded at 10:00 UTC. TCS in standard state!
- CARM/DARM noise injections while on ALS (Bersanetti, Pinto) concluded at 13:30 UTC.

The activity on  OMC lock in NI single bounce started at 07:12 UTC. Gouaty asked me to set ITF in SINGLE_BOUNCE_NI and misalign PR about 20 urad ty-. PR realigned upon concluded the activity.

ITF left in LOCKED_ARMS_IR.

We profited of some machine time to make noise injections on CARM and DARM, to be used to develop new control filters after the free space implementation.

Timestamps:

• 12:33:00 UTC: clean, 180 s;
• 12:40:00 UTC: DARM noise injection, DARM_noise shape, amplitude 1e5, 180 s;
• 12:58:00 UTC: CARM noise injection, DARM_noise shape, amplitude 1e5, 180 s;

This morning, Fabio and Roberto performed additional checks on the electronics of the WI main CO2 laser. They found that the laser sync was intermittent: touching the sync board caused the signal to fluctuate. The board was therefore replaced with the spare one.

After the replacement, the WI main CO2 laser was switched on again at 09:49 UTC. The restart was successful: the laser reached its nominal output power and has remained stable since then (at least up to the time of writing), as shown in Fig. 1.

In the afternoon, Nicola reopened the cooling valve of the WI auxiliary laser, which was then switched on at 12:15 UTC.

The TCS is now back in its standard operating configuration: all CO2 lasers are on with all beams blocked on their respective benches.

Finally, together with Nicola and Diego, we updated the DMS. If any flip mirror on either CO2 bench is opened, allowing a laser beam to propagate toward the interferometer, the corresponding DMS status flag automatically turns red. 

The CO2 laser status is as follows:

• NI bench:  
  • AUX LASER:  ON
  • MAIN LASER  ON
• WI bench:
  • AUX LASER:  OFF
  • MAIN LASER: ON

All laser beams are blocked inside the benches.
Do not open the acoustic enclosures.
If access is required, please contact Marco and me.

This morning we performed some preparatory actions before the intervention on SDB1/SDB2 scheduled in 2 weeks. One of the goal was to test the DET python scripts (used for OMC lock and picomotor actuation on SDB1) to make sure they were still operational. We also took this opportunity to take an alignment reference in NI single bounce with the OMC locked, we adjusted the demodulation phase of the B1_PD3 photodiode, and we optimized the OMC alignment with 0 offsets on B5_QD2. 

07h15 utc : Thanks to Nicola, the NI single bounce is set. SDB1 bench is in drift control.

At 07h07m10 utc, we open the OMC shutter.

Noticing that the correction in TX applied to the marionetta was around -5V, we acted on the motorized weight of SDB1 (BENCH_TX) by -25000 steps, to lower the correction up to 0.4 V.

At 7h48m25 utc we remove the offsets of the B5_QD2 quadrants (the ones used for ITF lock). This gives us an alignment reference for the single bounce (with B5 QD2 offsets at 0), while B1p is well centered on the camera (within +/-50 um from the center).

We observe that the initial OMC temperature was 22.55 deg.

Between 7h50 and 8h40 we tried several time to lock the OMC but we faced the following problems:

• Initially the B1_PD3 shutter was closed (although the state of the shutter indicated in VPM was open).
• While we were switching off/on the Vbias and closing/opening the shutter of B1_PD3, this triggered the closing of the OMC shutter.
• Then around 8h24 utc we made another attempt of OMC by the lock was not triggered when passing around the resonance. We realized that DET_MAIN probably reseted LSC_DARM to 0 when the OMC shutter automatically closed. So we reset LSC_DARM to 1.
• We also observed that the demodulation phase of B1_PD3 was mistuned. We adjusted it by -1.55 urad at 7h59m09 utc. The new demodulation phase entered in the configuration is 1.85 rad. We also changed the B1 demodulation phase by -1.6 (new value in the configuration file is -0.6 rad) although we did not test the B1 phase.

OMC was locked at 08h42m50 utc (See attached figure). Power on B1_PD3 is initially equal to 12 uW.

We adjusted the OMC alignment by acting on the picomotors, and reached a power on B1_PD3 equal to ~15.7 uW.

At 9h11 utc we start an OMC scan to check the quality of the alignment.

Order 1 = 0.4 uW > This gives an alignment defect of 2.5%

Order 2 = 0.3 uW > This gives a mode mismatch of ~2%

We restored the dark fringe offset for B5_QD2, closed the OMC shutter and concluded our activity.