From a message received from our external collaborator Renato Romero:
Friday June 5th at 17:30 UTC the 1 Hz comb disturbance stopped completely: Figure 1
Interestingly, in the subsequent minutes are again well visible the disturbances that we attribute to trains: Figure 2
The problem of the PDU server was that the TANGO PDU server was also up and running and was interfering with the current server. Tango instance has been stopped, problem resolved.
ITF DOWN in UPGRADING mode.
Palnned activities:
The problem of the PDU server was that the TANGO PDU server was also up and running and was interfering with the current server. Tango instance has been stopped, problem resolved.
This morning at around 1AM UTC the external magnetometer detected a decrease (but not disappearence) of the comb noise intensity.
We investigated further the 1Hz comb noise, which is still present.
Last Friday (May 29) we performed some measurements with one Bartington 3D Magnetometer and the portable spectrum analyzer CoCo80x.
First we found that the noise was present and quite intense close to the vacuum pipe of the W arm. Figure 1 shows the probe position in the first measurement: its tip was approximately 1 cm from the chamber at 900W and so approximately 0.6m from the tube center, the X axis was oriented along the tube, Y was vertical and Z was radial to the tube. In this position, the 1Hz comb field is oriented mainly along Y, being consistent with the field radiated by a current flowing along the tube. This is shown in Figure 2. The intensity is similar to what measured by similar sensors positioned close to vac chambers in CEB: that is a few nT/sqrt for the 1 Hz peak - compare with Figure 3 which uses the same 20s fft window.
The same Figure 2 shows recordings when the same probe was positioned by doubling the distance from the tube center: the field orientation is as well along Y and the intensity approximately halves. This looks consistent with a field being emitted by the beam tube. We then positioned the probe on the tunnel floor (which is the 3rd plot in Figure 2): the intensity is rougly the came but now the B field vector is oriented at 45 deg in the plane perpendicular to the tube, which also looks consistent with a field emitted by the tube. Yet, when positioning the probe close to the tunnel roof a large signal is noticed (Figure 4). No measured amplification at the ethernet switch box and cables running along the tunnel wall.
We then repeated the measurement close to the tube at 1800W and 2700W: the emitted comb is there with similar intensity and orientation characteristics (Figure 5). This seems to contraddict the hypothesis that the noisy current is generated in CEB and decays flowing outwards along the tubes. While at 1800W we also sniffed the vacuum pump station as well as cables but found no amplification. At 2700W we also took again a measurement close to the tunnel roof, this time being outside. Interestingly the noise disappeared from the probe when positioning the probe in the middle of the nearby road (green track in Figure 5).
We finally took some measurements in the central external area: Figure 6. The comb is present! The signal is particularly intense close to the methan gas pipe tube nearby the guardiania. Il looks that one such pipe runs inside the EGO site along the South fence. We also moved in the external road and towards the location (approx. 100m before the EGO gate) where the SNAM company performed heavy works between February 5th untill the first week of April (Thank you Maria fro recording the dates). The noise level is more or less the same when moving the probe on a tripod at 1.5m above the ground (Figure 6).
Our suspect is that the source of the noise might be related to galvanic anticorrosion currents in the methan pipes, which in the past produced a non stationary 5Hz comb: https://logbook.virgo-gw.eu/virgo/?r=55542. This time instead the noise is very stationary: the comb is extremely narrow and consistent with a GPS synch device.
A more stringent proof could come from measuring a few km away from EGO, in the locations explored in the past.
The FbTrend100s server configuration has been updated to use the 1TB disk available :
ITF DOWN in UPGRADING mode.
Activities communicated in control room:
SBE
SBE expert working from remote to restore and close the loop for EIB.
Software
BACnetServer found frozen: it was alive but it was providing data. To restart it I had to manually kill it and then restart it from VPM.
This morning (08:00 - 10:00 UTC) we performed the cleaning of NI tower with CO2 system.
this morning we replaced the two ZnSe disks recently found to be defective, elog_69077 . The spare disks had been inspected by Suzanne elog_69088. The two new viewports will have then to be He-tested for leaks when back under vacuum. Note: the larger ZnSe disk is the one numbered #03 , the smaller size disk is the number #3.
A 18 kohm resistor has been inserted in series on each signal driving the piezo (X, Y), with the goal of insterting a low-pass filter (cut-off frequency = 3Hz) on the correction signal sent to the piezo limiting the bandwidth and filtering out the amplifier noise.
Today we inspected the small NI viewport that was removed. During visual inspection of the ZnSe viewport after removal from the NI tower, a large number of reflective particulate-like features were observed on the optical surface exposed to the vacuum side while installed. The particles appeared as bright metallic/glitter-like flakes distributed across the central aperture and surrounding areas of the window. The distribution of the flakes was concentrated over a significant fraction of the clear aperture. Similar particulate debris was recovered from the surrounding handling surface. These observations are indicative of coating degradation and delamination from the ZnSe substrate (Figs. 1 and 2). Fig. 3 shows the air side of the viewport, which did not underwent the delamination process. The particles attached to this side seen on the photo are coming from the tissue that was contaminated from the vacuum side where the viewport was handled.
The spare viewports that are going to be installed were also inspected by eye and both of them seems perfectly fine. Figs. 4 and 5 show both sides of the small viewport and Fig. 6 and 7 show both sides of the big viewport.
The "tranfer function" measured in the previous logbook entry is the tranfer function from a noise injected before the driving matrix and after the filter, to the normalized signal of the quadrants. So when we inject a noise on one of the TILT/SHIFT signal, this noise is only visible on the corresponding quadrant signal. In the example given in the logbook entry, the noise is injected on TILT_X signal, and arrives on the far-field quadrant horizontally. The next image is a scheme of the BPL loop with the injected signal :
Another remark: EIB was not suspeded when preforming all the measurements. So they have to be repeated with the EIB suspended, which may affect the measured sensing matrix, and so the corresponding driving matrix.
The sensing and driving matrices measured in the precedent post 69150 were wrong, due to an error on the threshold on the phase that separates the positive from the negative responses between the piezo and the quadrants. The new matrices were computed:
sensing_matrix =
-0.6230 -0.0278 0.0033 0.0012
0.0233 -0.3838 -0.0010 -0.0059
-0.5035 -0.0063 -0.4332 -0.0151
0.0154 0.3421 0.0166 -0.2886
driving_matrix =
-1.5909 -0.1001 1.8540 -0.0972
0.1081 -2.5520 0.0166 -3.0190
-0.0128 0.0071 -2.2893 -0.1237
-0.0080 0.0517 0.1270 -3.3971
Using the new driving matrix, we managed to lock the BPL loop in drift control (f < 5 Hz). And then in broadband.
When the BPL is locked in broadband, a 122 Hz signal appeard on the QPD signal spectrums.
In order to measure the open loop transfer function, the BPL loop was open and a colored noise signal with a 0.001 V amplitude was injected on the channels:
To see the transfer function at low frequency, a 0.03 V signal at frequency < 20 Hz was injected on the same channels at the times:
The shape of the open loop transfer functions from the CORR signal to the corresponding of the quadrant signals are similar. The following plot is a transfer function from TILT_X_CORR signal to QF_h_norm during a low frequency injection:
The 122 Hz noise was due to a saturation of the control loop, so the loop filter was changed. The QPDs signal spectrums in closed loop, with the old and with the new filter are shown in the following plot:
The 122 Hz signal disapeared.
ITF DOWN in UPGRADING mode.
Operations carried out throughout the day:
Motor interfaces were not working since the relative TANGO servers were down. I recentererd the IP using the centering script. Now F0 positions should be fine.
When Cecilia tried to close the controls through the operator interface, the setpoints both of TOP stage and BOTTOM stage changed with respect to the used values, see fig.1. This issue has already been observed in 68701.
F0_X_SET: from -2000 to -450
MAR_TX_SET: from -104 to -20.5
MAR_TY_SET: from 163.5 to 144.
Maybe the setpoints have been changed without saving the values in DSP card and have been overwritten to older values.
Correct setpoints have still to be 'recovered'.
On june 2nd at 14.52 UTC the control of MC suspensions opened. Now the top stage controls can't be closed again since the suspension is far away from the setpoints, it should be recovered maybe through the motors, but the motor UI is not working. Experts should address the issue.
When Cecilia tried to close the controls through the operator interface, the setpoints both of TOP stage and BOTTOM stage changed with respect to the used values, see fig.1. This issue has already been observed in 68701.
F0_X_SET: from -2000 to -450
MAR_TX_SET: from -104 to -20.5
MAR_TY_SET: from 163.5 to 144.
Maybe the setpoints have been changed without saving the values in DSP card and have been overwritten to older values.
Correct setpoints have still to be 'recovered'.
Motor interfaces were not working since the relative TANGO servers were down. I recentererd the IP using the centering script. Now F0 positions should be fine.
To measure the sensing and driving matrix of BPL, a sinusoidal signal at 2 Hz, with 0.2 V amplitude was injected on channels:
Directly on the piezo correction signal. The BPL loop was open while injecting noise.
The signals were injected respectively at GPS times [1464363022, 1464363121, 1464363296, 1464363397] , each injection lasted for 60 s.
The channels QF_h_norm, QF_v_norm, QN_h_norm and QN_v_norm were monitored to see the effect of each piezo on each quadrant.
Piernicola's matlab script "Sensing_matrix_BPL.m" was used to compute the matrices:
sensing_matrix =
-0.6230 -0.0278 0.0033 0.0012
0.0233 -0.3838 0.0010 -0.0059
-0.5035 0.0063 0.4332 0.0151
0.0154 0.3421 0.0166 0.2886
driving_matrix =
-1.6099 -0.1078 -1.8808 0.3216
0.1242 -2.6459 0.0738 3.1261
0.0118 0.0090 2.3271 -0.1449
0.0085 -0.0545 -0.1126 3.5359
However, the loop does not close. We tried to adjust the setpoint of the piezo (TILT/SHIFT_X/Y_CORR_SET) so as the norm signal of the QPDs come close to 0, and tried several values of the DRIFT_FILTER gain but we did not succeed.
Paolo got a similar driving matrix, the values of the set point has to be studied more deeply.
The sensing and driving matrices measured in the precedent post 69150 were wrong, due to an error on the threshold on the phase that separates the positive from the negative responses between the piezo and the quadrants. The new matrices were computed:
sensing_matrix =
-0.6230 -0.0278 0.0033 0.0012
0.0233 -0.3838 -0.0010 -0.0059
-0.5035 -0.0063 -0.4332 -0.0151
0.0154 0.3421 0.0166 -0.2886
driving_matrix =
-1.5909 -0.1001 1.8540 -0.0972
0.1081 -2.5520 0.0166 -3.0190
-0.0128 0.0071 -2.2893 -0.1237
-0.0080 0.0517 0.1270 -3.3971
Using the new driving matrix, we managed to lock the BPL loop in drift control (f < 5 Hz). And then in broadband.
When the BPL is locked in broadband, a 122 Hz signal appeard on the QPD signal spectrums.
In order to measure the open loop transfer function, the BPL loop was open and a colored noise signal with a 0.001 V amplitude was injected on the channels:
To see the transfer function at low frequency, a 0.03 V signal at frequency < 20 Hz was injected on the same channels at the times:
The shape of the open loop transfer functions from the CORR signal to the corresponding of the quadrant signals are similar. The following plot is a transfer function from TILT_X_CORR signal to QF_h_norm during a low frequency injection:
The 122 Hz noise was due to a saturation of the control loop, so the loop filter was changed. The QPDs signal spectrums in closed loop, with the old and with the new filter are shown in the following plot:
The 122 Hz signal disapeared.
The "tranfer function" measured in the previous logbook entry is the tranfer function from a noise injected before the driving matrix and after the filter, to the normalized signal of the quadrants. So when we inject a noise on one of the TILT/SHIFT signal, this noise is only visible on the corresponding quadrant signal. In the example given in the logbook entry, the noise is injected on TILT_X signal, and arrives on the far-field quadrant horizontally. The next image is a scheme of the BPL loop with the injected signal :
Another remark: EIB was not suspeded when preforming all the measurements. So they have to be repeated with the EIB suspended, which may affect the measured sensing matrix, and so the corresponding driving matrix.
.jpg)
.jpg)
.jpg)
.jpg)
.jpg){kind=link}
.jpg){kind=link}
.jpg){kind=link}
.jpg){kind=link}
