There is a typo in the table showing the timing of the whole operation. All was done from 12:28 UTC to 13:18 UTC, so the table should be

Moreover there are two other sensors that should be taken into account for analysis: INF_WEB_HEATER_PRES_IN and INF_WEB_HEATER_PRES_OUT: they are temporarily acquired at 10kHz

The goal of the shift was to finalize the DARM hand off and proceed in the reduction of the CARM offset.

After the issues with computing, we began the shift by attempting immediately the DARM hand off @50 mW, and injecting noise. We injected colored noise filtered with "MICH_noise" at UTC 18:28:30 + 2mins (ampl 1e1). This allowed us to prepare a filter for the DARM hand off given by ALS_Arm_notch*ALS_Arm_control3*Arm_mains*DARM_violin. This filter allowed us to increase the loop UGF up to 40 Hz, since we were able to simplify some of its structures. Fig.1 shows the outcome of the DARM hand off and filter swap (which happens about 10 seconds later, it's especially visible on LSC_DARM).

This achievement allowed us to push almost up to 100 mW in transmission of the arms (Fig.2), but we unlocked immediately after. In the figure one can see that the cause of this unlock was the trigger of B4_12MHz going to 0, so we forced it to 1 for the next attempts. We managed again to reach this threshold, and we unlocked again, this time for unclear reasons. During these offset reductions, we had to constantly monitor the gains of the DRMI loops in order to prevent huge increases of their UGFs, especially on PRCL side. In our attempts we consolidated a procedure that seemed to be working every time, and can possibly be automathized by setting gains changes with appropriate ramps, that can anticipate the growth of oscillations and UGFs. We also prepared some functions in the python script carm_darm_green_m.py. The following is the complete procedure with the used functions:

• set the ARMS_LOCK in LOCKED_CARM_OFFSET_HIGH (still doesn't work 100% of the times, might require manual copy pasting of commands from the .py scripts from the metatron nodes that get stuck)
• lock the DRMI on 3f (the procedure is smooth assuming good alignment)
• set ARMS_LOCK in LOCKED_CARM_OFFSET_MID
• execute function pre_carm_to_MC() from carm_darm_green_m.py. This function swaps the PDs on B7/B8 and sets the triggers of B1p and B4 DC to 1.
• execute function enable_carm_to_MC_test(carm_set = none). If there's any problem in reading the carm_set automatically, plotting LSC_B7_B8_SUM_SQRT and taking the rough average is enough.
• set LSC.CARM_MC_SET to 5.0 (~12 mW on the arms). We tested 20 s as ramptime.
• execute function darm_to_ir_rf(darm_set_raw = 0, darm_gain = 1.3e6). The darm_set is not used in the function at the moment, and the gain is already comprehensive of the consideration reported in previous shifts.
• execute function darm_to_ir_rf_filter() to implement the aforementioned filter.
• set LSC.CARM_MC_SET to 8.0 (~30 mW on the arms). We tested 20 s as ramptime.
• set LSC.CARM_MC_SET to 10.0 (~50 mW on the arms). We tested 20 s as ramptime. Be ready to decrease PRCL gain by about 33% (0.13 -> ~0.09). Change CARM filter to compensate structures at lower frequencies on its TF. There is no function for this at the moment, but there's a switch in the MC_Mir DSP (the one for the RFC and the SSFS), which needs to be set to 1 to engage the filter. gname: SW_CARM.
• set LSC.CARM_MC_SET to 12.0 (~70 mW on the arms). We tested 20 s as ramptime.  Get ready to reduce PRCL gain to ~0.06, MICH to ~1.8.
• set LSC.CARM_MC_SET to 14.0 (~98 mW on the arms). We tested 20 s as ramptime. This step is the most unreliable, and there is currently no strategy to survive further. Assuming the CITF has been kept under control up to this point, this state seems stable.

The gains modifications are recapped in Fig.3. The carm offset reduction will be continued in the following shifts.

We leave the ITF with arms locked on the IR

The first part of the shift was spent to recover the standard working condition after the Failure of some Virtul Machine hosted on the same HW Enclosure.

At around 16:46  UTC everything was properly running and the ISC activity could starte, still in progress.

In parallel activities for:

- FLT: IR - green matching;

- SVS: EQB1

both activities are still in progress.

DAQ
today, after the Failure of some Virtul Machine hosted on the same HW Enclosure , we lost the connection of some power supply Hameg 17. I restored the connection with the ones located in CB while; to be restored the connection with the ones located in the end buildings.

The python processes running on the Virtual Machine involved in the failure did not restart, in particular:

- olserver130: SusDAQBridge, SatServer, PduServer;

- olserver137: DMSserver.

We spent about two hours to understand that the processes were complaining about the leap-second file;  indeed inside /virgoData/LeapSeconds there was not a leap-seconds.list but only .old or _old files.

The problem was solved copyng the leap-seconds.list_old (the most recent one) into leap-seconds.list.

At around 16:46  UTC everything was restarted and back normal.

The aim of the shift was to continue the DARM handoff activities.

In order to perform the handoff, the usual procedure to reach a low CARM offset with the DRMI locked has been followed, but a few issues on the way of reaching the handoff point allowed to perform only an handoff attempt.

A few notes on the issues encountered:

• At the beggining of the shift the West ALS laser was slightly misaligned, but we deemed a realignment intervention was not necessary yet. However, this caused later a few unlocks, in particular due to the actuation on WE of the DIFFp loop when engaged, that worsened the alignment of the green laser in the cavity (see figure 1). Increasing the fast loop gain from 2.5 to 3.0 and increasing the ALS_WEB_PD_REFL_LP threshold used by metatron to check the lock, together with disabling teporarily the DIFFp loop allowed to keep the lock stable for the rest of the shift.
• At the beginning of the shift the North ALS beam was instead still in optimal conditions, but during the morning the alignment drifted up to a point that does not allow the lock to be stable enough.
• The demodulation phase of the ASC.B1p_QD2_H_56MHz signal used by the DIFFp loop required a substantial retuning (performed at 150Hz of offset), from 1.36 to 3.10 radians, while the vertical signal (ASC.B1p_QD2_V_56MHz) required a much smaller tuning (from 0.86 to 1.1 rad).
• Noise injections on DARM have been performed before attempting the handoff at 10.40 UTC with amplitude 1e5. Increasing the amplitude to 2e5 to achieve higher coherence caused an unlock due to saturation of the corrections. The injections confirmed that the overall gain estimated required to perform the handoff for the signal B1p_56MHz_I_ARM_NORM is ~1e4.
• A handoff attempt has been performed at 12.20 UTC (see figure 2), but failed due to too low gain. We later noticed that the gain set today and yesterday did not take in consideration the internal gain (6.7e-4) of the DARM loop.
• Relocking the DRMi has been proven difficoult at times, and was solved using the SR misalignment procedure:
  • misaligning the SR by +8urad in TY
  • lock the DRMI and engage AA in full bandwidth (sometimes might be necessary to lower the 12 and 112MHz trigger thresholds, but if the PR alignment is not too bad it is usually unncecessary)
  • ramp SR slowly (e.g. in 40 sec) to realign it.

The activity will be continued in the next shift, as soon as the computing issues will be solved.

Reboot needed to some Virtual Machines hosted in Chassis8 due to Failure. Please restart services on those machines.

Virtual machine envolved are:

• ctrl6
• ctrl22
• farmn1
• farmn13
• ctrl12
• farmn2
• farmn5
• farmn7
• olserver111
• olserver121
• olserver130
• stol04
• mysql2.ego-gw.it
• olserver116
• olserver139
• olserver140
• olserver142
• datagw
• olserver132 - Mysql
• dbserver01
• igwacc-ego.virgo-gw.eu
• ctrl16
• farmn10
• farmn14
• farmn15
• seiscomphost
• rdserver2
• cvmfs0
• cvmfsmgr
• pub14 
• swtest3
• condorcl-test
• olsvn
• seevogh01
• submit-test
• w6
• dataldr
• cvmfsclient
• swtest6
• olnode1-test
• olnode2-test
• olnode3-test
• olnode4-test
• lscgw
• cvmfs1
• swtest5
• win-vtf
• vCenter VMware

The planned ISC activity on "hand off" carried out by Valentini and Pinto went on wiyhout major problems for the whole shift.

Air Conditioning
At 06:45 UTC, PyHVAC DET Servo disabled under request of Maddalena Mantovani.
At 09:10 UTC, PyHVAC INJ Servo disabled under request of Maddalena Mantovani.

DAQ
09:39 UTC - WEB_ALS_fast process crashed / restarted via VPM.
12:02 UTC - FbTrend_10800 process crashed / restarted.

DET
06:17 UTC - I moved SDB1 in a safe position for the CITF locking.

On-line Computing & Storage
Since 12:50 UTC a Network problem occured. Login not working anymore for Logbook, TDS, VPM machines... Computing team solved it at 13:40 UTC.

This entry is meant to clarify the use of the Python loop for the DET and INJ labs HVAC, the data of the last days will be used for this scope, see Figure 1.

This python loop is simply a low pass filter of the corrections in order to cure the HVAC system auto-oscillations (https://logbook.virgo-gw.eu/virgo/?r=45411)

It is visible ideed that the accuracy on the temperature control is lower when the python servo is off, Figure 2, with respect to when the servo is on, Figure 3.

But this servo has to be used in standard and regime conditions, due to the low pass behavior, since it slowes down the reaction of the HVAC control system to modifications of the thermal load (people in the labs etc...).

This speed decrease can bring the system to not working properly, see Figure 4.

The Python servo should be switched on only when the system is @ regime and the error signal is @ 0, if this does not appen as for istance Figure 4, the system is too slow to compensate and will spoil the HVAC system performance.

For this reason all the Python servos should be switched off in this commissioning phase, and they will be switched on only when the system will be stationary and the low noise measurement will be done.

If needed a test to increase the LP cut-off frequency can be performed.

ITF State: cavities locked
Quick Summary: all suspension loops closed; all SBEs loops closed;
No activity ongoing.

The goal of the shift was to further increase the CARM offset reduction and attempt the hand-off of the DARM error signal to B1p_56MHz_I normalized with B7+B8.

For the duration of the shift, assuming a good alignment of the central ITF, locking the arms, the green lasers and the DRMI was a smooth procedure that very rarely required any intervention on our side.

We were able to proceed further than the well known 150 Hz offset many times. During these attempts, we noticed that, after swapping the photodiodes on B7/B8 from PD2 to PD1, a glitch on the arms transmission at the moment of closing the shutters caused an unlock of the IMC every time.
We avoided this inconvenience by swapping the PDs before feeding CARM to the IMC.

This allowed us to proceed to 50 mW on the arms multiple times during the shift, at UTC: 20:28:00 we had our longest lock, that lasted more than 15 minutes.
During these attempts, we noticed that the drift control on Diff+ struggled to maintain a good recombination of the beams on the BS, once the power in the arms grows. This is arguably due to its error signal disappearing into noise around 0 (Fig.1).
Of the multiple unlocks of the IMC that we caused, one in particular caught our eye: around UTC 19:14:00 the transmission of the IMC decreasing while the power on B7 and B8 increased in the span of about 5 ms, before the unlock of the IMC (Fig.2). At the moment we have no explanation for this weird behaviour.

The other unlocks of the IMC were arguably due to a hand-off of the DARM error signal with a bad gain, or the DoFs in the DRMI (PRCL and MICH, as usual) oscillating to the point were their gain could no longer be increased without causing un unlock by itself.
This seems to be triggered by the sidebands in the DRMI behaving significantly different between each other when reducing the CARM offset, a problem in which an imperfect alignment of the SR might be playing an important role (Fig.3).

About the hand-off of the DARM error signal, we tried it using 1.2e4 as gain, and phi = 4.74 rad as demodulation phase. While the phase should be correct, the gain was too small and the loop failed to close.

About the lock procedure, the following was quite robust and repeatable: using the Automation, one can lock the arms up to CARM_SET  = -3 kHz, than the DRMI can be locked as well in an automated way up to the 3f signals. Then, after moving to CARM_SET = -150 Hz, B1p_DC and B4_DC triggers are forced to 1 (maybe we should do that also for the B4_112MHz_MAG given the SBs'  behaviour), B7 and B8 PDs are swapped, the DIFFp is engaged and finally CARM can be moved to the IMC. At this point, we could just increase the CARM_MC_SET offset from the usual ~ 0.4 initial value up to 10 in 60s. During this process, often the PRCL gain needs to be first increased from 0.12 to 0.15 in order to damp a bit the 38 Hz noise, then reduced to 0.07 to avoid loop saturation, especially after CARM_MC_SET ~ 5. All other loop gains were left untouched, however the MICH loop could have had a part in some of the unlocks.

We leave the ITF with the arms locked on the IR.

The shift was dedicated to the planned ISC activity, CARM offset reduction and handoff of DARM, carried out by Bersanetti and Boldrini. The work proceeded without any major problem, activity still in progress.
Other activities carried out during the shift:

• FLT: FCEM Optical Lever recovery (see report #52173);
• SGD: FLT IR GR Superposition (see report #52174).

Air Conditioning
At 14:12 UTC, due to the human presence inside the Detection Lab, Soldani opened the PyHVAC DET Servo (see entry #52169). At 20:39 UTC, after the SQZ activity was completed, I closed the loop again.

DET
At 18:12 UTC I moved SDB1 in a safe position for the CITF locking.

SUSP
From 15:01 UTC to 15:06 UTC, due to the sudden increase of Sa_MC_F0_COIL_H1_50Hz (see report #52164), Boschi decreased the value using the related motors. For this operation, with the support of Chiummo, we opened the RFC loops as a precaution.

The goal of the shift is send the IR beam retroreflected by the filter cavity on EQB1. 

After the recovery of the optical lever we realigned the filter cavity with the green beam:

1. We misaligned by 1000 urad in TX the input mirror and we centered the direct beam in the FCEB_GR_CAM
2. We religned the input mirror and we realigned the end mirror by 1000 urad in TX
3. We superposed the incoming beam and the retroreflected beam on EQB1 looking after the green FI
4. We realigned the end mirror and we sligthly moved the two filter cavity mirrors in order to maximize the flashes (max value found ~2V) see Fig 1.

When the cavity was aligned with the green beam we started to align the IR retroreflection up to EQB1

1. Acting on Picomotors on SQB1 M21 and M22 we steered the IR retroreflection in order to have it transmitted by the FI
2. When the retroreflection was transmitted by the FI we moved SQB1_M13 and SQB1_M14 in order to send the retroreflection on EQB1
3. We have to move the diaphragm after SQB1_M14 because the IR beam was not centered into its hole.

After this operations we had the retroreflection on EQB1 and we started to align it into the homodyne cameras.

When we was doing the camera alignment we realized that the retroreflection was slightly clipped by the FI thus we moved M21 in order to unclip it. We measured the power of retroreflected IR beam after the FI and we had 1.208 mW compared by the 1.25mW of incoming beam.

After the realignment the retroreflected beam was too low on SQB1_M12 thus in the next shifts maybe could be useful to increase the height of the beam by acting both on M21 and M22

We tried to see IR flashes in transmission by the filter cavity without success. We didn't check the alignment of IR sensors and we didn't move the FLT Mirrors during this attempt.

FCEM marionette has been properly oriented looking at the back reflection from the open flange. We did the adjustment with open loop.

• TX has been adjusted moving the marionette TX motor (number 5)
• TY has been adjusted adding a DC on the coils (0.15 V)

Then we centered oplev PSD and closed the loop. We needed to further adjust the loop set-points to find the flashes. (TX = -100, TY= -500)

We have also realigned PSD2 (TZ), PSD3 (MIR_TY, MIR_TX) and PSD 4 (MIR_Z) 

Config file has been updated with the new values.

Today at 17h37 utc I changed the gain of the OMC Peltier, in order to restore the gain normally used under vacuum. This concerns the following line in the SDB1_OMC configuration file:

ACL_DAC_CH    dac1955_SDB1_OMC_ch02    1    DAC1955_FREQ    OMC_L_Peltier_dac    0.0    -1.0        "Butterworth_4_20"    "ad1955-10v"    "" #0.2

The gain was set to -1.0 (it had been temporary set to -0.3 during the intervention on SDB1).

Profiting of the microtower in air for alignment purposes, we installed the missing baffle also on the end mirror:

• we removed both the dome and the intermediate ring of the tower to have access to the payload. In order to remove the intermediate ring, we had to dismantle the windows to avoid interference with the optical lever breadboards and to separate the cavity pipe;
• we blocked the inverted pendulum and we installed the frame on the front side of the cage;
• we installed the two parts of the baffle on the frame;
• we closed the tank, reinstalling all the windows and reconnecting the pipe;
• we released the inverted pendulum, we recentered it and we proeprly closed the position loops;
• we realigned the optical levers, we adjusted the marionette working points and we closed also the angular loops.

On Wednesday 16/06/21 we started slow-down tests on the hot water pump of WEB. We have proceded at steps of -5Hz starting from the standard 50Hz supply and switching between the left and right pump in order to check for different responses.

Before the test the following 4 accelerometers have been added:

Before the beginning of operations the right water pump was active, with 50Hz supply

During the whole process a regular series of glitches from air compressors occurred, every 3 minutes.

Task of the shift was to continue the activity on CARM offset reduction and the handoff of DARM to the B1p_56_MHz_I_ARM_NORM signal.

The shift started with a preliminary alignment of the Green beam: the West was quickly realigned modifying the setpoints of the BPC of TILT and SHIFT on the X direction. 

New setpoints are:

TILT X= -0.40;

SHIFT x = -0.88.

The ALS West trigger and lock threshold was also updated accordingly to the power. 

For the alignment of the North Green beam we called the ALS on-call, who took care of it (see entry 52163).

After that, at 9.01 UTC we were able to proceed with the scheduled activity by locking the CITF and reducing the CARM offset first to -3000 Hz. and then to -150 Hz through the new nodes on Metatron (GRD.ARMS_LOCK). At 9.05 UTC we performed the handoff to MC.

We then proceeded recovering the DIFFP loop by reopening the B1p_QD2 quadrant (now closed in default 52155) and checking its demodulation phase against misalignments of WE mirror. The phase of B1p_QD2_H_56_MHz was changed from 4 to 4.5 rad and then DIFFp was engaged. However, this caused a sudden drift (see figure 1) due to an incorrect sign in the loops. 

We used the remaining time of the shift to perform some preliminary tests on the new automated arm-prealignment tool (autosnail, more info on it will be posted soon), while the DARM handoff procedure will be continued in the afternoon shift. 

This morning was dedicated to ISC activity; Manuel and Michele worked all the shift without any major problem, the ISC work will go on in the afternoon...


other activities of this morning:

- SQZ: FCEM baffle installation and filter cavity alignment recovery;
- EDB cabling from 7:30UTC to 10:00UTC;
- ENV: west end environmental sensor installation, done at around 10:30UTC;

SUSP
one of the MC F0 coil is going to saturation (see plot).

TCS
at 13:00UTC we lost the data from the TCS chiller guardian (expert informed).

With the recent tuning of the working point for the 3 QNR lasers, the temperature of SC laser crystal was decreased by only about 1 deg while the main and CC lasers were moved by about 1.5 deg. Thus the sign of the offset frequency of the SC laser from main laser was changed, and the PLL was locked at 1060 MHz instead of the nominal 1260 MHz. In order to engage the SC AA loop, which requires a frequency offset of 1260 MHz, on Friday 11/06 I changed the SC laser temperature from 30.26 deg to 29.41 deg, and I managed to engage the PLL at 1260 frequency offset. I could not reach the needed temperature with the control voltage from the DAC, and I observed that the control voltage saturates below DAC count = 22000 and above DAC count = 40000; the linear range of the control voltage is only from -1.6 V to +1.1 V. Thus I set the control voltage to DAC count = 32000 and I moved the laser temperature with the manual knob on the driver.

I engaged the SC AA loop, however the loop did not maximize the SC-CC beat. On 15/06 I verified that this was due to some clipping of the SC beam on the Faraday isolator, then I manually adjusted the aligment, I engaged the AA loop, and I verified that the beat note amplidute was maximized.

On Monday 14/06 the squeezer was switched to squeezing mode around 6:30 a.m. UTC. On first we did not find the 4 MHz beat note at the homodyne detector, and we realized that the delay line was misaligned. We recovered the coarse alignment using the subcarrier beam, and we performed a fine tuning by observing the CC beam on the HD cameras, then by engaging the HD AA loop.

After the recent tuning of the working point for the QNR lasers, we scanned the OPA and observed the transmission of the CC laser on the SQZ_EQB1_IR_PD_MONI photodiode to check the single mode operation as already done in the past. We do not see evidence for multimode operation, see Fig 1.

We measured the OPA parametric gain by observing the magnitude of the 4 MHz LO-CC beat at the homodyne detector while scanning the LO optical path length, see attached note CC_scan.pdf. We repeated the measurement for three different values of the green pump power, with the following MZ offset and timing:

12:10 + 2min - MZ Offset = 0V

12:15 + 2min - MZ Offset = 0.25V

12:20 + 2min - MZ Offset = 0.5V

Fig 2 shows the 4 MHz magnitude for MZ offset = 0V (magenta) and MZ offset = 0.5V (blue). The measured parametric gain from the data, i.e. the ratio of maximum and minimum magnitude of the CC-LO beat,  is g=1.9 for MZ offset = 0V and g=2.3 for MZ offset = 0.5V. We also observed the effect of pump power tuning on the amplitude of the pump-phase error signal, see Fig 3 with MZ offset = 0V (magenta) and MZ offset = 0.5V (blue). By comparison with the SQZ_MachZ_PD_DC signal we confirm that both the amplitude of the pump-phase error signal and the parametric gain measured from HD RF channel scale as the square root of the green pump power, as expected.

With the lower pump power setting (MZ offset = 0V) we engaged the HD CC loop with DSP filter 1 at gain 1500, and we changed the CC phase  between 0.9 rad, corresponding to maximum anti-squeezing, to 0.9 rad corresponding to squeezing, then we open the CC loop and closed the LO shutter. Fig 4 shows the rms of HD audio channel vs time: the three successive plateau values correspond to anti-squeezing, squeezing, and shot noise respectively. The measured anti-squeezing is about 2.3 dB and the squeezing is about 0.7 dB, while the expected produced squeezing from the measured parametric gain should be around 3.2 dB, suggesting the presence of extra losses and large phase noise.

While extra losses can be attributed to possible clipping along the EQB1 delay line limiting the accuracy of the AA loop, we investigate the phase noise issue by performing noise injections on the CC loop. Data were taken first with the fast CC loop alone using DPS filter 1 at gain 1500.

• Add Report

- A stream of clean data without injected noise injection but with the angular dither lines on (0.005 mV amplitude) starts at 2021-06-14-23h35m03-UTC.

- Dither lines were switched off at 2021-06-15-05h35m45-UTC

- White sensing noise injection with 50 uV amplitude at 2021-06-15-06h43m39-UTC   

- Coarse CC loop added with gain = 1 at 2021-06-15-06h51m11-UTC

- Coarse CC loop gain increased to 3 at 2021-06-15-06h53m19-UTC

- Coarse CC loop gain increased to 10 at 2021-06-15-06h53m54-UTC

Fig 5 shows the measured TF with the fast CC loop only; the structure around 230 Hz prevents to increase the loop gain.

Fig 6 show the TF with the addition of the coarse CC loop at gain = 10.

I made a factor 2 mistake in the sampling frequency of the channels, this made the 50Hz harmonics appear as 25 Hz harmonics in the figures.

Figure 1 and 2 show respectively the amplitude spectrum density (in V/rtHz) and the amplitude spectrum (in V) for a FFT length of 10s (so a 0.1Hz frequency resolution) once this factor 2 error is corrected.