The tiltmeter transfer function has been measured by injecting noise on both the actuators couple, in order to verify that the response doesn't change from one couple to the other.

ACT2 = 2V --> GPS = 1393261900; D = 1200;

ACT2 = 0V --> GPS = 1393001500; D = 600;

In figure 1 the OLTF resulting from the two noise injections with the model superimposed.

In a quiet period the OL sensitivity of the tiltmeter has been measured (figure 2) and, dividing by (plant-1) the ground tilt is derived (red trace). Its ASD results to be about 2.5 nrad/sqrt(Hz) at 100 mHz, and is of the order of 1e-10 rad/sqrt(Hz) and below in the frequency band 1-20 Hz. This is compatible with what has been measured in the past months (figure 3), when the resonance frequency of the system was 23 mHz instead of 7 mHz.

What was noticed by Paolo is that, for a particular perturbation (maybe a nearby truck) the signal as read by the tiltmeter @2.6Hz has the same amplitude as read by the Optical Lever (figure 4).

Further analysis ongoing.

Since the correction has drifted and the actuation is currently sent to the other couple of actuators, a new noise injection was performed to measure the transfer function with these actuators.

Starting GPS = 1393001440, duration = about 15 minutes

This week some work was done to improve the tiltmeter permformance.

The readout interferometric system had a very poor SNR due to very low light power injected (about 300uW). After some investigation, it turned out that the light was very badly coupled to the optical fiber. Thus te coupler was better tuned and about 6.2 mW of power were coupled at the fiber output.

After this, the system was relocked with a slightly different filter (zeros were moved from 0.004 Hz to 0.03 Hz).

Before disconnecting the laser source, a noise injection was performed to have a way to better calibrate previous data (GPS=1392543613). Another noise injection was performed after the relock in order to measure the new OLTF (GPS=1392634074),

The factor to calibrate from Volts to radians is computed as follows:

alpha=lambda/(2pi*L) * 1/(V_max-V_min)

where lambda = 532e-9m (laser wavelength), L = 0.1m

before laser adjustment: V_max = 0.012V, V_min = 0.022V --> alpha = 8.5e-5 rad/V

after laser adjustment: V_max = 0.34V, V_min = 0.06V --> alpha = 3e-6 rad/V

Notice that the calibration factor can be further improved up to a factor 3 by making a better arm balancing.

Many analysis are ongoing to study the coherence of the tiltmeter signal with the closeby seismomenter and other sensors in different environmental conditions

On Tuesday we worked to put back in operation the tiltmeter.

First, we installed back the laser source which was taken away at the end of August.

Furthermore, we installed two thermometers inside the chamber: one is in contact with the chamber itself, the other is very close to one joint’s head. This is to monitor the temperature gradients and delays and possibly correlate the tiltmeter’s drift to temperature variations. They are acquired by the slow monitoring module ENV-NEB-W2, and the channel names are NEB_TILT_VAC_TE and NEB_TILT_JOINT_TE, respectively. In order to compare the tiltmeter signal to the seismometer one a seismometer was installed close to the tiltmeter chamber (details reported in 62717)

To reduce the noise introduced by actuators amplifiers at low frequency, we added a modulation of the correction signal in Acl code and a high pass filter in the path between the DAC and the actuators. However,

since the high-pass filter is hosted in the same box as photodiodes amplifiers, we noticed a cross-talk between the correction signal and PDs signals. We will need to separate the two setup.

Finally, we tuned the tiltmeter center of mass position as to be closer to the joints bending point, and we reached in air a resonance frequency of about 9 mHz. 

This morning, during the work to restart the tiltmeter, we installed a couple of auxiliary probes. In particular: a seismometer, placed very close to the tiltmeter (see photo), and a couple of temperature probes inside the tiltmeter vacuum chamber. The new seismometer channels are: ENV_NEB_SEISM_T_N, ENV_NEB_SEISM_T_W and ENV_NEB_SEISM_T_V, with the usual orientation, while the temperature channels are: ENV_NEB_TILT_VAC_TE (probe placed on the chamber wall) and ENV_NEB_TILT_JOINT_TE (probe placed close to one of the suspended arm's joints). In the plot the coherence between the two seismometers and a comparison of the spectra are shown. A good coherence is visible up to about 10 Hz.

Few points to clarify.

The channel NNC_NEB_TILT_ERROR  is given by:

ERROR = (SET - ITF/PICKOFF) * GAIN = ERROR_POST * GAIN

(channel prefix NNC_NEB_TILT omitted for sake of clarity)

where NNC_NEB_TILT_SET is the set point of the error signal, and is needed for the ERROR to oscillate around 0. 

The correct way to get the error signal calibrated in radiants is:

ERROR [V/V] * 3.8e-7 [rad/V] * mean(PICKOFF) [V] / GAIN

or, alternatively,

ERROR_PRE(or POST) * 3.8e-7 [rad/V] * mean(PICKOFF) [V]

where 3.8e-7 rad/V is the calibration factor.

The tiltmeter installed in the NEB has been operating in the last month (figure 1) and a first characterization has been performed.

The optical readout scheme is reported in figure 2. NNC_NEB_TILT_ITF is the signal as is read by the interferometer photodiode, while NNC_NEB_TILT_PICKOFF is a pickoff of the input laser power. This latter is used to perform a laser power noise subtraction from the ITF signal, and the normalized signal which accounts for this subtraction is called

NNC_NEB_TILT_ERROR = NNC_NEB_TILT_ITF /  NNC_NEB_TILT_PICKOFF.

(Notice that the *_ERROR channel has been added only after few weeks, to see data from earlier periods the signals ERR_PRE/POST can be used. These signals are different between them only when a noise injection is performed).

The tiltmeter signal has been calibrated and a first estimation of the sensitivity has been performed. The calibration factor for the ITF signal has been computed to be 3.8e-7 rad/V. This value can be considered correct within 10%, as it depends on the contrast defect of the readout interferometer signal and on the locking condition- which should be mid-fringe, but the signal could drift in time.  A new signal called "NNC_NEB_TILT_ITF_CALI" takes into account this calibration value.

To calibrate the error signal, instead, it is necessary to multiply it by the calibration factor and by the mean value of the pickoff in the considered data stretch.

Some noise injections were performed in order to measure the OLTF and therefore compute the out-of-loop sensitivity from the error signal. In figure 3 the OLTF model superimposed to two noise injection, one perfomed on the left couple of actuators, the second perfomed on the right couple. From this measurement we desume that the two actuators couples perform the same force on the arm. 

The OLTF can be modeled as follows (exploiting the "virgo2zpk" function in matlab format [function sys_zpk = virgoZPK(zero_freq,zero_Q,pole_freq,pole_Q,gain@gfreq,gfreq)])

fg = 0.12 %filter gain

H = virgo2zpk(0.01,0.5,[0 0.8],[0 0.707],1,0.1)*fg %filter model

G = virgo2zpk([],[],0.023,270,1,0) %plant model, where 23mHz is the resonance frequency of the system and 270 is the Q as measured in air by Paolo. This latter might be re-measured in vacuum

g = 850; % gain necessary to match the measured TF with the model

OLTF = G.*H*g

In figure 4, the best sensitivity of the tiltmeter during this period is shown. In figure 5, the sensitivity divided by the CLTF. 

The tiltmeter shows to be sensitive to different wind condition (speed and direction). However, it is also affected by low frequency noise when a higher correction is applied to the actuators to hold the lock. Figure 6. This aspect will be improved in the next future.

This morning the system for remote operation on the two tiltmeter motors was installed at NEB. At first I asked for the position of both motors:
ARUN : 0
PI : 5
After that I moved the ARUN motor (that acts on the screw) by +10 and after -10 steps, verifying the correct operation at each operation.

The system seems works properly. Meanwhile I connected two long network cables: one to the Raspberry PI and the other to "Ciabatta intelligente": now these two cables are on the floor.

This week we installed the tiltmeter in the North End Building. The tiltmeter arm is oriented along the main interferometer beam direction, i.e. toward North, and the vacuum chamber is positioned on the West side of the the tower. In figure 1 the position of the tiltmeter in the NEB is marked with a red star. In figure 2 a picture of the tiltmeter position with respect to the North End Tower is shown. 

The vacuum chamber was delivered to the Virgo site on Friday, and vacuum team took care of the transportation up to the NEB.

The main mechanical parts were dismounted in Naples to avoid damages, and were mounted back on the site. Everything went smooth. Attached few pictures of the setup (figure 3-6). The resonance frequency of the arm has been measured to be around 22 mHz. 

The electronics needed to amplify photodiodes was not working properly, so we dismounted it and debugged it together with the electronics crew. The problem was found and solved. 

The cables labeled with "Tiltmeter" (which were still connected to the distribution box since 2019) were connected to the ADC and DAC racks identified by LAPP people. In particular:

- the 3 ADC were connected to the VME ADC7674 board number 17, channels 12, 13 and 14

- the DAC are connected to the  DAQ box number 35, DAC mezzanine number 121, channel number 5,6 and 7

At present, everything is switched on and connected. What is needed is to activate the Acl process to acquire the signals and control the tiltmeter.

The two motors installed (a PI stepper motor to lock the arm and perform the center of mass regulation and an ARUN stepper motor to move a screw and rebalance the arm)  need to be controlled with a Raspberry Pi and Giulio is taking care of the configuration. It will be installed in the rack as soon as the configuration will be completed.

The chamber is not yet under vacuum, and Luca and Riccardo are completing the installation of pumps and controllers to do the vacuum.

The BS alignment control has been closed in full bandwidth. Only the BS_TY phase had to be slightly retuned. The change has been implemented in the Metatron configuration file.

In figure 1 the FFT of the BS error signals in drift control (pink trace) and full bandwidth (blue trace). The RMS improvement is visible at low frequency for the BS_TX and all over the bandwidth for BS_TY, although B1p is slightly higher, maybe for the different alignment settings.

From a better analysis of the photodiodes signals, the 155kHz line seem to be visible on B7 PD1 (figure 1), while no signal is visible on PD2. The same happens also for B8 PD1. This is quite surprising because the shutter in front of PD1 should be closed in dark fringe. Further investigation are needed to better clarify why the signal is not visible on B7_PD2.

We measured the modulation depth of the three sideband frequency both from the arm free swinging and from a RFC cavity scan. Notice that the modulation depth change during the lock acquisition process (in particular in LN1). 

Hereafter the value that we measured when the ITF was unlocked. After the arrow, the values in LN1. These latter have been derived by taking into account the variation as read from the channels INJ_LNFS_AMPL1,2,3. 

- m_6MHz = 0.29 -->  0.002

- m_8MHz = 0.18 --> 0.1

- m_56MHz = 0.11 --> 0.195

After the test of increasing the temperature of the NI using the heating belts, we still had some doubt that the line giving rise to the instability on Jan 7th and 8th was not the same as we observed to be moving during the test.

For this reason, a further analysis was performed: the frequency of the 155kHz line was followed during the unlocks taking place on Jan 7th and 8th (see figure 1). As the temperature of the four test masses is not constant over the two analyzed days, the frequency is also expected to change linearly with the temperature.

In figure 2, the frequency shift of the line is shown as a function of the four test masses temperatures. The behaviour is linear only for the NI and WI, pointing again to one of the INPUT. 

Finally, from the HB test, we observed that not only the NI, but also the NE test mass temperature was varying by about the same amount, so I superimposed the data from the instability events to the data from the heating belt experiment. The results are shown in figure 3. The data follow the same trend as the NI, and we also find the same frequency for the same temperature value. From these data we can say that the line giving rise to the instability belongs to the NI test mass. 

I gave a further look to the unlocks of Jan 7th, in particular to the once happening before LN3, which are slower and allow to appreciate the high order mode rising up on B1p camera.

The shape of the high order mode appearing right before the unlock evolves in time, passing from an HG01 on Jan 7th to an HG10 on Jan 8th (see figures 1 to 3). This latter is visible only in the last unlock before the normal operation of the interferometer is recovered.

Comparing the frequencies of the 155kHz line in the first and last case, they differ by almost 1Hz (figure 4). The temperature variation of the NI test mass is about 0.1K (figure 5), which would give a frequency shift of ≈1.7Hz, so it's difficult to understand if the line which gives rise to the two event is the same and is shifted by the temperature or maybe it's another mode which gets excited. 

Looking at the B7 and B8 signals sampled at 400MHz, it seems that on the event of Jan 8th a small peak at around 150kHz is also visible on B7, but it could also be some random noise. Still to understand why no clear signal is visible on B7.

Last night, the setpoint of the NI heating belt has been changed by 0.3 K in order to change the temperature of NI test mass only. This test had the aim to verify whether the 155.408kHz peak belongs to the NI. If yes, the frequency of this line should move by a known amount, as computed below.

In figure 1, the temperature variation for the four test masses is shown for the analyzed period (gps1 = 5/2/20 22h00 UTC --> gps2 = 6/2/20 10h38 UTC). Moreover, also the correction given to the two heating belts is plotted. What was not expected is that both the temperature of NI and NE changes by approximately the same amount, which is, respectively:

deltaT_NI = 0.3K

deltaT_NE = 0.27K

A much smaller temperature variation is seen instead on the West test masses, also thanks to the etalon loop acting on the WI tower and keeping the WI temperature almost constant. 

Taking into account the rate of frequency change reported in 42370, using the temperature variation for NI and NE, it is possible to compute the expected frequency change for the 155.4kHz line:

0.67Hz/K * (155.4kHz/7.8kHz) * deltaT_NE ≈ 3.6Hz

0.87Hz/K * (155.4kHz/7.8kHz) * deltaT_NI ≈ 5.2Hz

In the attached movie, a zoom around the frequency of interest of the B1 spectrum is shown. A plot is taken every 3500 seconds, and it is possible to follow the evolution in frequency of the 155kHz line.

In figure 2, the two superimposed plots correspond to gps1 and gps2, respectively.

The measured frequency shift is ≈ 4.8Hz, which is compatible with the value expected for the NI test mass.

I took a look to the signals sampled at 400MHz in correspondence of the 155kHz line ramping up and eventually unlocking the ITF (7th of January). Few things to notice:

- The 155kHz line is visible on all the central interferometer photodiodes and QPDs (see figures 1-3). Notice that the resolution is low because of the high sampling rate of the signals.

- A 155kHz modulation is visible around the sidebands (6MHz, 8MHz and 56MHz), but not in the same way on all the photodiodes. In particular, the modulation around the 56MHz is seen by all of the PDs (figure 4), the modulation around the 8MHz is seen by B2, B1p and B1, and the modulation around the 6MHz is only seen by B1 and B1p (figure 5). 

- The 155kHz line is much louder on B8 than on B7, where it doesn't really appear, at least in low resolution (figures 6-7). This could be related to a difference between their electronics, or could be a physical effect. If it is not an effect related to the PD electronics, this could give a hint on the arm where the instability starts. However, this point has to be further investigated with the help of PDs experts. 

Finally, I checked the same signals also for the similar events which happened on Jan 22nd-23rd. The behavior of the signals shown before is the same. In figures (8-9) B7 and B8 photodiodes are shown. On the two latter plots also B7_PD1 and B8_PD1, although no signal should be present on those PDs as their vBias is off and a shutter should prevent the light to reach them. To be better understood.  

After few hours of DF lock, we tuned again the B5_QD2_H(V)_56MHz demodulation phases by adding an angular lines to the PR, so we managed to close back BS_TX in full bandwidth. BS_TY has been left in drift control, as it needs a deeper study. We also retuned a bit the alignment setpoint of BS_TY, DIFFp_TY and PR_TY. 

After the reboot of the Daq box on SDB2, we had to retune all the demodulation phases of B1p and B5 photodiodes and quadrants.

As a bonus, also the SSFS phase had to be retuned (mostly related to the SPRB drift). 

However, during the morning, we experienced many unlocks due to different causes: 

- saturation of B1p Audio signals. Digging further, it seems that the signal where the noise (5.4Hz) was louder is LSC_PRCL_ERR. to be investigated

- 1 failure of the B1p QPD shutters opening

- in ACQUIRING_LN3, the engagement of BS_TX full bandwidth caused an unlock as the loop started oscillating. The BS_TX loop was gain has been increased from therefore left in drift control and the line commented in ITF_LOCK.py

- after this, we had a couple of fast unlocks

After the science mode has been achieved the AA has been tuned. It is worth to be noted that in the initial phase of the lock the optimal working point was quickly changing up to one point in which was stabilized (may be related with the IITF thermalization)

Yesterday, after the maintenance, an increase of the 155kHz line has been observed again during the first relock. Since the ITF was not in dark fringe, the B1 channel was not present, but it was possible to observe the peak on B1p starting from the "LOCKING_OMC1_B1s2_DC" state, so when the DARM offset is increased to lock the OMCs. However, there is no clear link between the peak amplitude increase and the DARM offset, since the peak could be already present but buried in the noise floor of the photodiode. The attached movie shows the process. It starts from GPS = 1263642318, and a frame is taken every 10 seconds.

Unfortunately, the order 1 mode that we noticed last time on the B1p camera (VIR-0018A-20, slide 11) is not as evident this time, probably because of the DARM offset hiding it, or because its amplitude was not so high.

It seems to be the same kind of phenomenon which happened on the 7th and 8th of January , although this time it stopped before giving rise to the instability, not yet clear what's the mechanism. Investigations are ongoing.

Moreover, analyzing the relocks during these last two days, I noticed that this phenomenon happens more than once, and in particular it seems to occur when the ITF is left cold for more than 20 minutes.

GPS of some of these events can be found in the attached figures.

After the maintenance on 10/12/2019, a free swinging measurement has been performed in order to evaluate the arm cavity Gouy phase and therefore extract the ETM Radius of Curvature. The reason for taking this measurement after the maintenance is that the mirrors had few hours for cooling down, so the effect of the Yag heating is reduced.

To perform these measurements, the modulation depth of the sidebands has been reduced:

- from 15dB to 6dB for the 6MHz (not possible to reduce it further because not to lose the RFC lock)

- from 15dB to 5dB for the 8MHz

- from 7dB to 3dB for the 56MHz

in order to ease the analysis of the carrier high order modes. Moreover, the cavity has been slightly misaligned (1.5urad) in order to highlight as many high order modes as possible, provide that the peak of the first order didn't become higher than the 00 mode.

The analysis has been performed in the same way as described in entries 42701 and 43276. In summary, the Gouy phase of the cavity has been extracted for the first 4 high order modes.

The results are pretty similar for the two arms, and are reported in the following table:

The error on each measurement is the standard deviation of the distribution computed over many FSR.

Considering the input mirror radius of curvature equal to 1425m, and knowing the cavity length (2999.8m) the radius of curvature of the end test mass has been extracted for the ETMs.

In figure 1 one FSR of the North cavity has been superimposed to the outcome of a Finesse simulation, where maps are not used and the ETM RoC is 1662m. However, if we also consider the maps in the Finesse simulation, we must add a radius of curvature of 1665m (figure 2), which means that the presence of the map shortens the radius of curvature "seen" by the modes. 

Few things have to be remarked.

1 - because of the high finesse and high cavity speed, each peak is affected by the ringing. This induces a shift towards right of the resonance peaks. However, as long as the speed doesn't vary too much within the same FSR, all the peaks are affected by the ringing in the same way, and therefore equally right-shifted. It was already found in the past analysis that the maximum allowable speed spread within the same FSR is about 0.5um/s (42701).

2 - to further check that the FSR x-axis is not bad scaled, the position of the sidebands has been taken as a reference, as they only depend on their frequency and on the cavity FSR.

In particular, the 8MHz USB is superimposed to the 3rd HOM, as we see from the slope of the PDH signal on B2_8MHz. (figure 3). Since we know that this sideband is at 0.3253 in fraction of FSR, this corresponds to a phase of 0.3253*pi = 1.022, which divided by 3 is 0.3407, which is the same gouy phase found for the HOM3. 

The positions of the 6MHz sidebands does not match exactly with the expected ones (figure 4): this could be an hint that there is a systematic error on the x-axis scaling. This has to be further investigated.

3 - For the ITM, the assumption is that the radius of curvature is 1425m. However, if this is not true, the value of radii of curvature derived for the ETMs will be different. 

Since this morning a couple of "jumps" of the B1p power (not related to the DARM offset reduction) have been noticed. They seem to be related to a slight change of position of the beam on the B1p camera. (figure 1)

We checked the signals for the vertical position of SR, and it seems that the 15 um of drift of F7 are compensated by the upper part of the suspension. At a first glance, it is not clear if anything is changing in the SDB2 alignment. To be better investigated by the experts.

This drift is also visible on the SDB2_B1p_QD2_GALVO_V correction.

We also checked in the same DARM offset conditions (1mW on B1) if there was any variation in the higher order modes content on B1, but there is no evident difference, besides a very small astigmatism of the beam (figures 2 and 3).