The lock acquisition was not so straight forward for the whole morning.
-det
this morning I had to call the det on-call because of the shutters automaton get stuck in a UNSAFE_B1p state preventing to go on with the lock (#43584)
-vpm
HRec restarted at 12:32utc because of crashed
The electronic noise of the standard coil driver board (P4, installed at WE) has been estimated again, looking at DARM in low noise 1, but this time the measurement has been repeated with the board in use for locking. The result is almost the same, just a small increase (~10 %) which could even be due to the (home maid) calibration uncertainty.
yesterday we added in SIB2 clean area:
- one horizontal accelerometer attached to the plate of the water cooled beam dump at the W side of the SIB2 tank, named ENV_SIB2_ACC_Z
- (temporary) one loudspeaker wrapped in chellophane foil for cleanliness, presently driven by the ENV_NOISE_CEB_EEroom DAC channel
Figure 1 shows the B1 beam during O2, while figure 2 show the B1 beam in these days. The shape seems to be no more Gaussian. Figure 3 and figure 4 show the same images plotted in log scale. It is visible that, in these days, B1 has some HOM components, clearly visible increasing the integration time and the gain of the camera (figure 5).
Checking B1s2 (figure 6 and figure 7), it can be noticed that the beam reflected by the second OMC is changed wrt the past, as expected considering the substitution of the first OMC. In B1s2 the same HOM components present in B1 are clearly visible.
Yesterday the unplugging test at WE has been repeated, in the new configuration with 13kohm series resistor instead of 4.5 kohm. No visible reduction wrt clean data (fig 1). Unfortunately yesterday the ITF was noisier than the day before (fig 2), so the accuracy in detecting some common noise from the board was again poor. Let's say that we have no new diagnostics about WE P4 board.
After several attempts the logic safety properly kicked in and moved DET_MAIN to the UNSAFE_B1p dead end state.
To get out of this dead end state DET_MAIN needs to be reloaded, added a notification in DET_MAIN so it tells that the node needs to be reloaded to get out of this state.
A solution proposed a month ago, is for ITF_LOCK to request DET_MAIN to go to ENABLE_B1p instead of SHUTTER_CLOSED whenever it is locking the arms or recombined.
I have added corresponding lines in LOCKING_ARMS, LOCKED_ARMS et LOCKING_RECOMBINED and and reloaded ITF_LOCK around 09:15 UTC. Hopefully they will fix this issue and not cause other issues.
Searching for the time with lowest level of noise using band rms in the 2.0-2.1 kHz, here is the spectrum for 1 second of data (blue), compared to a time when the squeezed beam was not injected (purple line). On this very specific 1 second of data (lowest level of noise floor), there is maybe a 10% reduction of the shot noise level.
ITF_LOCK steps are overloaded, in each of the ITF_LOCK state several transitions are made.
LOCKING_OMC1_B1s2_DC corresponds to adding a dark fringe offset, adjusting ITF alignment loops, opening the slow shutter, locking OMC1, changing DARM error signal from RF to DC
Please also look at the state of other metatron nodes. DET_MAIN metatron node should have been displaying a flashing yellow message:
"Too many steps done and shutter not open, waiting for manual action"
As a consequence the OMC never received the request to lock. OMC_LOCK was not in the ACQUIRE_OMC1 state, but in TEMP_CONTROLLED.
Figure 1 shows the slow shutter opening, the figure in the lower right pane is a monitor of the current applied to the slow shutter.
Normally it should be in steps like in the second half of the time series, as metatron sends request for motion, with gaps in between them so we can count them.
In the first half there is a continuous motion of the slow shutter, which probably means the metatron cycles were too fast (less than 0.6 seconds?), and half of the motion requests were not received by the slow shutter rotator because send when the motion from the previous request was still happening. This would explain why the automation arrived at the safety limit of the maximum number of steps, as have of the steps requested were not done.
I have increased this morning the maximum number of step safety limit from 40000 to 45000.
We coarsely tuned the loop parameters (gain and offset of error signal) to make the loop stable without measuring the TF. Then we tuned the demodulation phase in order to maximize the effect of anti-squeezing on DARM noise. Detected anti-squeezing is now aroung 5.5 dB, higher than yesterday because of the reduced phase noise. We then changed the phase by 90 deg to inject phase squeezing.
Fig 1 shows the DARM and Hrec spectra in three different conditions: with the SQZ fast shutter closed (purple), with injected anti-squeezing (orange) and with injection of squeezing (blue). Differently from the effect of anti-squeezing, the effect of squeezing is barely visible, probably due to a combination of the large residual phase noise and of possible extra misalignment losses.
We also characterized the 7 MHz PLL which locks the B1 demodulation frequency to the local oscillator for the SQZ PLL.
Fig 2 shows
- the CC error signal (SQZ_B1_PD1_7MHz_I),
- the derivative of the phase extracted from SQZ_PLL7 monitoring (SQZ_PLL_7MHz_raw_dFreq), which is used as error signal for the 7MHz phase loop
- the coherence between these 2 signals
in the following conditions
- Blue: CC error signal gain equal to 1 and 7MHz phase loop closed
- purple: CC error signal gain equal to 0.5 and 7MHz phase loop opened
One can see that
- with a gain of 0.5 the bump above 100 Hz decreases
- with 7MHz loop open there is no major modification of coherence and the coherence remains at the same high level in the 7-30Hz region : to be investigate
We spent the remaining time to characterize the alignment error signals from B1p quadrants. We closed the SQZ fast shutter at 4:32. ITF is left in LN3, and the SDB1 Faraday shutter is left open to avoid unlocking.
Searching for the time with lowest level of noise using band rms in the 2.0-2.1 kHz, here is the spectrum for 1 second of data (blue), compared to a time when the squeezed beam was not injected (purple line). On this very specific 1 second of data (lowest level of noise floor), there is maybe a 10% reduction of the shot noise level.
The first part of the afternoon has ben dedicated to the suspension tuning, at about 17:05 UTC started the second part of the shift dedicated to the squeezer optimization.
17:27 UTC - OMC - we had again troubles to lock the OMC1, therefore we contacted expert oncall that fixed the problem.
18:20 UTC - ITF locked in LowNoise3.
21:45 UTC - SUSP - DMS showed red flag for the NE electronics (picture attached), I contacted the oncall expert who noticed a board switched off (probably by itself)... he re-started it remotely with the ITF locked without major problems.
22:00 UTC - Commissioning Mode stopped, activity still in progress (ITF left locked in LowNoise 3).
All the subsystems worked properly, no particular problems or events to report.
ITF_LOCK steps are overloaded, in each of the ITF_LOCK state several transitions are made.
LOCKING_OMC1_B1s2_DC corresponds to adding a dark fringe offset, adjusting ITF alignment loops, opening the slow shutter, locking OMC1, changing DARM error signal from RF to DC
Please also look at the state of other metatron nodes. DET_MAIN metatron node should have been displaying a flashing yellow message:
"Too many steps done and shutter not open, waiting for manual action"
As a consequence the OMC never received the request to lock. OMC_LOCK was not in the ACQUIRE_OMC1 state, but in TEMP_CONTROLLED.
Figure 1 shows the slow shutter opening, the figure in the lower right pane is a monitor of the current applied to the slow shutter.
Normally it should be in steps like in the second half of the time series, as metatron sends request for motion, with gaps in between them so we can count them.
In the first half there is a continuous motion of the slow shutter, which probably means the metatron cycles were too fast (less than 0.6 seconds?), and half of the motion requests were not received by the slow shutter rotator because send when the motion from the previous request was still happening. This would explain why the automation arrived at the safety limit of the maximum number of steps, as have of the steps requested were not done.
I have increased this morning the maximum number of step safety limit from 40000 to 45000.
Figure 1 shows a violin mode in B1 (OMCs transmission) and in B1s1 (OMCs reflection) before (purple) and after (blue) an alignment adjustment that improved the ITF optical gain by ~4%.
B1s1 sees ~7% of the reflected light and B1 PD1 50% of the transmitted light. So the heights of the lines corresponds to
33e-6/7e-2/0.00142/2 = 16.5% of reflected carrier light before alignment improvement
26e-6/7e-2/0.00142/2 = 13% of reflected carrier light after alignment improvement
If this is really true it would imply there is lots of light rejected by the OMCs for a different reason than alignment. A possible explanation is that the mode matching tuning done on the single bounce beam is not representative of the ITF beam size when the ITF is hot. To be further investigated.
This morning, there were some troubles in the lock acquisition. While reaching the SSFS boost step the IMC was unlocking. This must be due to the conversion of the additionnal frequency noise into power noise of the PMC that we were investigating yesterday (entry 43562 )
We went into the laser lab to realign the PMC and set the modulation frequency to readjust its working point: from 14.56 MHz to 14.53 MHz. The transmission of the PMC went from 0.84 V to 0.91V.
Then we retuned the offest on the IMC_REFL galvos. The new values are given below:
ACL_CONST_CH NF_H_offset "V/V" 1 LOOP_FREQ 0.05
ACL_CONST_CH NF_V_offset "V/V" 1 LOOP_FREQ 0.02
ACL_CONST_CH FF_H_offset "V/V" 1 LOOP_FREQ 0.35
ACL_CONST_CH FF_V_offset "V/V" 1 LOOP_FREQ -0.05
We finally remeasured the matrix of the IMC AA:
AA. ACL_MATRIX_BEGIN sensing_matrix 1 MCT_hNORM FF_ACh NF_ACh MCT_vNORM FF_ACv NF_ACv
ACL_MATRIX_CH MC_ty_err "" 12.57 0.00 -1.72 0 0 0
ACL_MATRIX_CH IB_ty_err "" 20.56 0.00 -22.09 0 0 0
ACL_MATRIX_CH MC_tx_err "" 0 0 0 6.29 17.81 -8.12
ACL_MATRIX_CH IB_tx_err "" 0 0 0 34.30 19.06 -8.69
ACL_MATRIX_CH IB_tz_err "" 0 0 0 10.45 2.44 29.62
ACL_MATRIX_END sensing_matrix
The noise injections to define the matrix have been done with the SSFS locked instead of the RFC locked as it was done before. Thus the AA is less sensitive while reaching this step of the lock acquisition. The lock has been relocked in low noise 3. There are still a lot of coupling of the frequency noise to the IMC AA error signal. This would deserve more work.
-det
during the relock attempts we had systematic unlocks while locking the first OMC, I called the det on-call to fix the problem.
During yesterday evening shift (Nov 13), we worked on 2 topics: we checked the DIFFp TY signal symmetry with the OMC now better aligned (see results posted in entry 43571), and we worked on the OMC lock acquisition on single bounce beam.
Here is the list of measurements performed:
LOW NOISE 3:
- At 16h34m46 UTC we put an offset of +10 um in the SDB1_B5_QD2_H error signal which displaced the bench in TY by about +1.5 urad. The transition to this new angular position was achieved towards 16:37:05 UTC. Starting from this time we collected about 2 Min of data with an on-going scan of the DIFFp TY degree of freedom.
- At 16:41:34 UTC the SDB1_B5_QD2_H offset was put back to 0, and another scan of the DIFFp was started. However the ITF unlocked about ~1 min after the beginning of this measurement.
- Low noise 3 was finally recovered and around 20h42 UTC we took another ~2 min of data with the scan on the DIFFp, and SDB1_B5_QD2_H offset to 0. ITF unlocked again at the end of the measurement.
SINGLE BOUNCE BEAM:
Michal fixed the OMC lock acquisition on the single bounce beam, and we could test it once successfully.
Then a scan of OMC1 was performed at 21:37:40 (2 min) in order to recheck the misalignment defect and mode mismatch (see attached Figure). Peak power was 11.8 mW on TEM00, 0.12 mW on 1st order mode (only 0.06 mW after subtraction of the tail of the TEM00), and about 0.1 mW on the 2nd order mode. This corresponds to about 0.5% of misalignment defect, and a bit less than 1% of mode mismatch between the single bounce beam and the OMC1.
I made a mistake in the evaluation of P4 differential noise, because I was assuming the impedance in LN1 to be 228 ohm as for P6, but in fact it is 128 ohm. In fig 1 the new evaluation of the differential noise (in agreement with the lab measurement).
The conclusion about the limiting noise before unplugging the coils does not change: differential noise can explain the excess noise. This is shown in fig 2: adding in quadrature the differential noise to the residual (unplugged) noise, we obtain the'plugged' noise.
The increased series resistor should put the differential noise at a negligible level.
Yesterday we repeated the measurements of last week on the dependency of the DIFFp TY error signal on the SDB1 alignment, after Michal realigned the OMC last weekend (#43516 ).
IN this condition the asymmetry is not present anymore. IN the nominal position the error signal is very symmetric (see Figure 1) and when misaligning the bench on the "wrong side" it is still very symmetric (see Figure 2). Notice that last week, on the same misaligned position, the ratio between both halves of the parabola was 0.04 and now is 0.7, so the improvement has been significant.
So now the problem can be considered as solve, but it is interesting to know that we the DIFFp TY error signal is very sensitive to OMC misalignments.
The power on B1s2 measures the amount of 56MHz TEM01 leaking through OMC1, the OMC alignment should be to maximize the carrier TEM00 transmission.
A good measure of OMC alignment on the carrier is the ITF optical gain divided by the power stored in the arms.
Figure 1, shows that during the changes in BS TX set point, there was no correlated change in ITF optical gain, only a gradual decrease in gain by 1% due to the decrease in power stored in the arms by 1%.
The changes in BS TX set point, are compensated by tilting the SDB1 bench, as visible in the SDB1 TX correction, which keeps the beam aligned on the OMC.
Figure 2, shows that the changes in BS TX set point have a clear impact on the overlap between carrier and 56MHz. So it seems that the BS TX set points misaligns the 56MHz with regard to the carrier, which could be causing more 56MHz TEM01 through OMC1.
It is likely that several actuator set points (alignment and maybe DAS) need to be changed together to find a working point optimal for CMRF, B1p power fluctuations and 56MHz passing through OMC1 (B1s2 power).
Last night attempts at exploring the impact of SDB1 (and OMC) alignment were hampered by ITF lock stability
Figure 3, shows the single change we had the time to make. Changing the SDB1 alignment set point, caused the optical gain to drop by 1%.
Note that the optical gain measurement has a large delay, as it is computed using a median of calibration lines height over several minutes.
A faster monitor of OMC carrier throughput is being investigated to help with these measurements.
It is likely that the seeder versus the neoVAN deserves a re-alignment after the events 42865and 42256 (misalignement accident and investigations on pump fibers).
It maybe be that such misalignement leads to a higher ratio of HOM in the beam aside from a power reduction it may impact on the overall performances (RIN, Freq?).
Up to now such action has been exclude because of the associated risk (contamnition) and the commissioning higher priorities.
> We stopped at 5:00 and tried to relock the ITF without success: north arm flashes are missing. We leave the IFT in DOWN state.
It was possible to get back the north arm flashes by having the NI and NE Automation nodes cycling to MISALIGNED -> ALIGNED in order to restore the positions recorded during the last lock period.
Locking attempts ongoing.
To speed up the search we scanned the SQZ laser frequency and kept the OMC temperature stable. At the end we scanned the OMC temperature to check the mode matching, see attachments. Misalignment is now reduced to 1%, mode mismatch is at the same level.
We should check again the mode matching after some hours to check stability of the alignment over time. By the way, some of the PZT actuators are now close to the end of dynamic range.
We stopped at 5:00 and tried to relock the ITF without success: north arm flashes are missing. We leave the IFT in DOWN state.
> We stopped at 5:00 and tried to relock the ITF without success: north arm flashes are missing. We leave the IFT in DOWN state.
It was possible to get back the north arm flashes by having the NI and NE Automation nodes cycling to MISALIGNED -> ALIGNED in order to restore the positions recorded during the last lock period.
Locking attempts ongoing.
At 20:30 UTC we could reach LOW_NOISE_3 and started the activity on OMC alignment and mode matching.
At 20:50 UTC the ITF unlocked and DET people from remote started working with only NI mirror aligned, activity concluded at 21:40 UTC.
After that started the activity of the SQUEEZING with the same configuration,
This Afternoon we have increased the 4k4 (2 x 2k2) total resistor to 13k6 (2 x 6k8), thus allowing some studies on the noise of the P4 board actually installed instead of the standard P6 with shaping filter.
To have back the signal in the DAQ (and then in the DMS) we have:
- replaced the application for TCSChillerWI from /virgoApp/PyArtic/v1r1/scripts/PyArtic.py to /virgoDev/Chiller/v3r4/Linux-x86_64-SL6/ChillerApp.exe in the VPM;
- renamed the old configuration file from TCSChillerAux.cfg to TCSChillerWI.cfg;
- renamed the old channels name with the current ones (from TE_Aux_Chiller to TE_WI_Chiller, from TE_Aux_Chiller_Set to TE_Aux_Chiller_Set, from Aux_Chiller_Alarm to Aux_Chiller_Alarm);
- configured the serial bridge with the proper settings: the device has been connected on port 3 of ethadvtcs2 with 9600 Baudrate;
- properly switched off the TCSChillerWI server;
- restart the process from VPM with the new configurations.
We report here a summary of the activities performed on the injection system during this morning shift:
- IMC locking point. We turned on the FmodErr line at 1111 Hz and we adjusted the offset on the locking point by looking at the transmission and by minimizing this line. We set the offset in ISYS Acl from -0.1 to +0.1.
- Rampeauto PSTAB. We tried to lock the PSTAB with the rampeauto (PSL-ELEC-1803-01), which we want to keep as a spare. We weren't able to lock in 1/f since it seems that the input signal has the wrong sign. We measured in fact the gain between the photodiode input and the correction by sending a sinusoid of 200 mV amplitude @ 400 Hz, finding a gain of +15 (while in the rampeauto which is currently working this gain is -10).
- PMC locking point. As already mentioned in the entry 43494, in order to minimize the frequency into power noise conversion, we tried to tune the offset of the locking point of the PMC. We already did it by looking at the error signal offset as a figure of merit and setting it to zero by adjusting the modulation frequency. Today we tried in another way: we sent in the perturb input a sinusoid at 1900 Hz of ampiltude 10 mV and we tried to minimize it in the PMC trasmission. We weren't able to do this since the PMC unlocked due to the fact that we were also changing the phase. We need to think to another way to adjust the locking point, knowing that for the moment the perturb input is used as a security in case the EIB starts to oscillate.
In order to check the PMC alignment we scanned it @ 4 Hz for a minute. You can find the scan starting from the GPS 1226142528. As you can see in the attached plot the misalignment mode is quite high and we need to quickly realign the cavity.
- weight added to the EIB and rebalancing. Since we are planning to replace EIB_M8 mirror with a larger one (3 inches) which weights 300 g more, we added an equivalent weight next to the mirror we want to change, we removed an unused mirror (the one was used for the optical lever), and moved a wehght on the EIB in order to rebalance it. We could enable the actuators and the PID, and finally B. Boom moved the stepper motors to bring back the actuation signals to less than 2V. We are ready to replace this mirror whenever it is possible.
It is likely that the seeder versus the neoVAN deserves a re-alignment after the events 42865and 42256 (misalignement accident and investigations on pump fibers).
It maybe be that such misalignement leads to a higher ratio of HOM in the beam aside from a power reduction it may impact on the overall performances (RIN, Freq?).
Up to now such action has been exclude because of the associated risk (contamnition) and the commissioning higher priorities.
This morning the DMS reported a failure of the laser chiller that triggered the laser over temperature protection. Upon investigation, we found that the chiller indeed needs servicing, since it seems it is not cooling the water anymore. Note that it does not show any failure error.
Since the spare chiller is still being serviced (will be back next week), we managed to re-install an old Virgo+ laser chiller, intentionally left in the intercapedine for emergency.
The old server has been restarted, the correspondign process is running on the VPM and there is a working flag on the DMS.
The CO2 laser also has been restarted and seems to be working fine.
To have back the signal in the DAQ (and then in the DMS) we have:
- replaced the application for TCSChillerWI from /virgoApp/PyArtic/v1r1/scripts/PyArtic.py to /virgoDev/Chiller/v3r4/Linux-x86_64-SL6/ChillerApp.exe in the VPM;
- renamed the old configuration file from TCSChillerAux.cfg to TCSChillerWI.cfg;
- renamed the old channels name with the current ones (from TE_Aux_Chiller to TE_WI_Chiller, from TE_Aux_Chiller_Set to TE_Aux_Chiller_Set, from Aux_Chiller_Alarm to Aux_Chiller_Alarm);
- configured the serial bridge with the proper settings: the device has been connected on port 3 of ethadvtcs2 with 9600 Baudrate;
- properly switched off the TCSChillerWI server;
- restart the process from VPM with the new configurations.