Here is an update of the analysis presented above. In this new analysis, the ratio between the PSD of Hrec and the PSD of the PSTAB during the time of the injection is used for the projection instead of the transfert function. This is defined as:

PSD(Hrec(High_frequency_injection)) * (PSD(RIN_reference)/PSD(RIN_HIGH_FREQ_INJECTION)),

this ratio is plotted on the figure 1 and compared to the sensitivity curve during the time of the injection. Here we need to found a criteria to define which frequency keep for the projection. We tried two methods for the frequency selection.

Method 1:
Use the ratio between the PSD of the RIN during the injection and during the reference time. This is what is shown on the figure 2. The upper figure shows the PSD during the injection and during the reference time. The ratio of the two curves is plotted in the figure at the middle. For the projection, we keep only the frequency where the ratio is higher than three (the threshold is represented by the horizontal line). Finally, the projection with this method is shown on the bottom subfigure. We observe that several peak of the sensitivity curve are conserved during the projection.     

Method 2:

Use the ratio between the sensitivity curve during the injection and during the reference time. This is what is show on the figure 3. The upper subfigure show the two sensitivity curves, the one during the injection and the on during the reference time. As a criteria for the projection, we use the ratio between those two curves. The ratio is plotted on the middle graph of figure 3. The horizontal line show the position of a threshold equal to 2. Finally, The projection with this method is shown on the bottom subfigure of figure 3. With this methods, the peaks are no longer conserved.

The figure 4 show what the total projection with the second method look like. This projection is done, for the low frequency injection with the method presented in the previous logbook since no problem occur during the projection, and the high frequency projection is done with the method 2 presented in this logbook. We choose to keep the projection done with the method 2 because the peaks of the sensitivity curve are not conserved by the projection. Since the injection of noise keep the sensitivity curve below the peak, it seems reasonable to say that the method should not be able to show if the peaks are due to the PSTAB noise or not.

The results of noise injection analysis are presented above. The code used is a python equivalent of the matlab code used for the analysis described here (59512). The code is shared in the following folder: /virgoDev/Automation/scripts/LSC/Pstab_injections.py

Here are the descriptions of the different figures attached:

--> Figure 1 : RIN for the injections at low and high frequencies. As expected the RIN is higher over the frequency of the injection. The RIN has been derived as the square root of PSD(PSTAB_AC/11*12*PSTAB_DC).

--> Figure 2 : Show the coherence between the RIN and h(t) during the injection at low frequency and high frequency. The projection will be done only for the values with a coherence higher than 0.5.

--> Figure 3 : Show the sensitivity curve for the 3 data set : the one used as reference (230 secondes before the first injection), the one during the injection at low frequency and the one during the injection at high frequency. We observe as expected higher strain over the frequency with noise injection.

--> Figure 4 : Show the projection. The projection is done only for the point with a coherence higher than 0.5 (see figure 2). The projection for the low frequencies is well reconstructed, but it seems that a part of the high frequency is missing by comparison with the plateau observed on the RIN (figure 1), and on the sensitivity curve during the injection (figure 3). I didn't found explanation to understand why the coherence is not satisfying for the high frequency injection.

On Monday afternoon we tuned the amplitude of the automatized noise injections used for the noise budget which are in /virgoDev/Automation/scripts/LSC/inject_lsc.py

for the PSTAB_LF (gps of the injection 1414774610 for 2min, see plot 1):  PSTAB_noise_color_ON('PSTAB_LF', 5.0, 'PSTAB_flt_LF')

for the PSTAB_HF (gps of the injection 1414774833 for 2min, see plot 2):  PSTAB_noise_color_ON('PSTAB_HF', 3.0, 'PSTAB_flt_HF')

We also opened the PSTAB loop from gps 1414775495 for 30 s (plot 3).

We still have to analyze data.

During the recovery of the INJ/PSL we forgot to check the status of the noise eater of the ML. It appears that it was properly working, see the oscillation at 780 kHz on the fiigure 1.
WE switched it off ad on again and it improved the situation of the PSTAB (see figure 2).

To be noted that there is still an additional noise on the in loop photodiode (see PdA on the figure 3)
Also to be noted that this noise was not present jsut before the shut down (figure 4)

Comparing the mismatch towards the IMC in the two different configurations: purple is with max PMC transmission, blue with reduced power sent to PMC (current), comparable IMC_TRA power.

The IMC_REFL_DC channel is reported after offset substraction (V0 = 0.34V), divided by the EIB_Pout power and normalized to 1 when IMC is unlocked. IMC mismatch was around 16%, is now around 11%.

PSTAb behavior during last night confirmed that the situation has been improved after the reduction of input power in the AOM before the PMC. To be futher investigated.

This afternoon the rebalance of PSL power discarding has been performed by lowering the power sent to the PMC and increasing the transmission of IPC1 in order to keep the same power transmitted by IMC.

The power transmitted by the PMC has been reduced by roughly 18% by acting on the HWP/polarizers just upstream of the PSTAB AOM. We measured the beam profile coming out of the PMC before and after the power variation (results in a separate entry), and realigned the PMC by suing the steering mirrors after the AOM and before the PMC. This had the effect of improving the mismatch to the IMC cavity while delivering the same power from IMC.

Also, the reduced power sent into the AOM seemed to be beneficial to the PSTAB loop. We maanged to engage the PSTAB loop without seeing the glitches which wwere there during the past days - no gain tuning was deemed to be necessary at this moment.

Unfortunately, the drawback was that the IMC linear alignment works for the moment in drift control, a check of the working point and possibly the alignment matrix will be done as soon as possible. For the time being the engagement of full bandwidth linear alignment has been left commented in metatron.

The interferometer was locked up to carm_null_1f after the work, and left to the TCS measurements.

The power injected to the ITF is around 40W after the overall settling overnight. It also seems that the PMC throughput has slightly improved, but it is difficult to assess because noise on the recorded reflected power PMC_REFL is such to hide the expected increase.

In order to reach 40 W injected into the ITF, as agreed in the PSL/INJ meeting, we changed the correction offset voltage of the AOM of the power stabilization loop.

We proceeded as follows:

• we opened PSTAB loop
• we unplugged the corrections (named out corr) at 14:13:50 UTC --> the power at the output of the PMC increased by 5%
• the potentiometer for the corrections was set to 8.5. We set it 0 and plugged again the corrections at 14:15:28 UTC
• we increased the corrections voltage step by step from 14:17:00 UTC to 14:25:00 UTC (1st plot)
• we verified from a trend that the minimum correction range needed is 0.85 Vpp (2nd plot)
• we set the offset on the potentiometer to 2.5 (14:52:22 UTC), to be in the range where the voltage effect is known. The 3rd plot shows the corrections and the error signal before and after this adjustmnet.

Around 13:30 UTC we put down INJ_MAIN and NARM/WARM nodes, opened BPC (DC on) to allow the work in the laser lab by the Artemis crew.

Around 18:10 LT we went on the IB platform to examine the ground connection of the PSTAB box installed on SIB1.

We found the west flange with the fedd-through of the PSTAB cables as indicated by Jean-Pierre Coulon (see picture 1). The small black wire starting from the PSTAB connector is attached to the metallic structure by means of a crocodile connector (see picture 2,3). We measured the resistance by means of an ohmmeter between:

1) exposed metallic plug before the crocodile connection and the tower = 0.3 Ohm after a few seconds (starting at 1.4 Ohm)

2) attached chassis and the tower = 0.1 Ohm

3) the metallic cylinder holding the cables in place and the tower = 3.4 Ohm

(see picture 4 as a reference).

After we went back to the control room, we closed again BPC loop, resumed INJ_MAIN, sent IMC_RESTORED as a request and then were able to relock the arms with the IR by using NARM and WARM metatron nodes. We leave the ITF like this.

Two unusual features were spotted during last day:

1) a sudden drop of power (around 1%) occurred on Jan 13 at 18:33:50 UTC and did not recover (first plot). This drop was not visible at the level of the neovan amplifier, and it affected the isys chain starting from the PMC.
The IMC transmission normalized by the PMC trasnmission stays more or less constant shortly before and shortly after this drop, while the PMC transmission, normalized by the neovan transmission, shows a jump confirming that something happened at this level.
Looking at the fast channels, the only evidence of something weird is found in the PSTAB behavior (second plot), but it is diffcult to say whether the correlation between the glitchy periods is in one sense or the other as for the causation. Actually, zooming further in (third plot) it seems that the PSTAB corrections saturate at their minimum without a clear reason, but more investigations are needed to say something more firmly. And it does not appear that the DC correction of the PSTAB could be held responsible for the persisting steady state of the power at a level lower than before the drop.

2) from the first plot, it is also evident that the power transmitted by the IMC starts to decrease slowly after this event - even when normalized by the PMC transmission, while its reflection divided by the PMC transmission increases. This is something that could be explained by either a misalignment of the IMC or a slow change of the mode sent out by the PMC. But again, more investigations are needed to say anything more than hypotheses.

Following the mis-functioning of the rampeauto occurred during last days, with the help of Jean-Pierre we investigated the issue of no corrections and found a broken inverter (M29) in the rampeauto board. By replacing this IC, it was possible to close again the loop of the power stabilization. Now we have to tune the loop parameters in order to get the TF we had during O3 and have residual power noise at the nominal value (should be rms(PSTAB_PD2_AC_MONIT,10) < 0.05, now is about 0.2V ).

The rampeauto of the PSTAB seems to have some troubles.

After the relock of the RFC which was necessary to allow the lock of the north arm, we switched to put back in operation the power stabilization loop. First we checked that the position and power of the beam entering the PSTAB box were comparable with what we had during O3, which is the case (plot 1).

Then we had a look at the signal error, and compared again with what we had during O3 when we intentionally opened the Pstab to make noise injection (see 46660). The situation did not seem so different (see plots 2,3), so we tried to close the loop, but we did not get corrections to close it properly (noise did not change at all - plot 4 - when sending the PSTAB enable signal). So we started to investigate the issue. First we checked that the enable signal was indeed arriving at the rampeauto and it was the case. Then we made a series of tests, in particular we sent a sinusoidal signal about 0.8V pp in the AC IN input and monitored the corrections with a scope, changing the rampeauto gain to try and get significant corrections.

In picture 5, the signal is the undistorted sinusoid with roughly 0.8Vpp, the corrections are the almost-in-phase, distorted sinusoid displayed with 1mV/div scale. The outcome seems that the AC corrections sent from the rampeauto to the AOM chain are distorted and too small. This will be checked more accurately.

In order to look at the time dependence of the coupling from PSTAB to B1 I have looked at the height of the line at 2961Hz over time in both PSTAB and B1 over a few hours.

Figure 1, shows the time series for 20 minutes of data, the coulping fluctuates by up to a factor 3.

Figure 2, shows the spectrum of that coupling time series, the fluctuations are mostly below 200mHz, but there is no clear structure between 1mHz and 200mHz, maybe there is some higher lines at ~25mHz and ~40mHz, but it is hard to say if it is real or just a statistical fluctuation.

Figure 3, looking at two consecutive periods of 3 hours, the peak at 25mHz looks significant, but not the one at 40mHz.

users/mwas/detchar/PSTABcoupling_20200116/TFrms.m

During daily meeting I saw that the line was no longer in the data. At 14:31 UTC I switched off the function generator and unplugged the cable from perturb input of the PStab rampeauto.

Useful data with this injection lasted from 11:11 UTC to 13:00 UTC.

Figure 1. Comparing PSTAB_PD2_AC_MONIT and PSTAB_PD2_AC_MONIT_PRE, the PSTAB loop has a gain of ~1500 at 3kHz, this seems reasonable as the PSTAB loop gain at 3kHz in the AdV TDR is 1000.

Figure 2. The line injection allows to measure the PSTAB coupling to the dark fringe. In blue is data with the line injection and in purple data with the PSTAB loop open from December 16 2019. Surprisingly the coupling now is a factor 6 higher than back in December. But is is only a factor 1.6 higher than what was measured back in July 2019.

Figure 3 shows the projection of PSTAB noise (both sensing and sideband RAM contamination) using the July 2019 transfer function scaled to match the observed coupling at the 2961Hz line.

Figure 4 shows the same but using the PSTAB transfer function measured on Dec 16 (which was much lower, hence the low coherence between 100Hz and 1kHz). Above 100Hz the two projections agree, but at low frequency the Dec 16 transfer function show a much higher coupling. This might be just a symptom that scaling the transfer function using the line at 3kHz for a noise below 100Hz is the wrong thing to do. I.e. the changes in the PSTAB coupling at high frequency are not the same as changes in the coupling function at low frequency.

In any case, the PSTAB projection above 100Hz is about a factor 10 below the measured sensitivity.

Under request from commissioning coordinator, reporting monitoring needs coming from squeezing team, we moved the line on PStab to roughly 2468Hz. Amplitude has been left unchanged.

Upon request by Michal, yesterday we added a permanent line on the PSTAB loop, in order to monitor the coupling fluctucations between PSTAB and the dark fringe.

We injected a line at 2970 Hz generated by a function generator in the "perturb" input of the PSTAB rampeauto. Unfortunately this device was faulty, in fact the line disappeared after an hour (look at the spectrogram, 1st plot).

We realized it this morning and we replaced the generator. The line is now injected (plot 2) but it has still to be adjusted in amplitude and frequency (we will discuss about it with Michal). 

The BNS range drops of the last days seem to be indeed related with the PSTAB, as visible from their correlation obtained with NonNA. In the attached figure, the (anti-)correlation of Hrec_Range_BNS and PSTAB_PD1_AC_FS_rms immediately before the maintenace ofTuesday. The intervention seems to have fixed the problem, and less high frequency glitches and BNS range drops are present now.

In the last days many glitches seem to come from the PSTAB. We then performed two actions:

- PMC realignment, as already observed and suggested by F. Cleva (#47533).

- PSTAB gain adjustment. We injected a line on the "perturb" input of the PSTAB locking electronics and monitored at the scope the amplitude of the error signals before and after the perturbation. We then adjusted the frequency of the injected line in order to have them of the same amplitude: this frequency is a good estimation of the UGF. It was 120 kHz, and by lowering a bit the gain we brought it at ~ 80 kHz (the corresponding position of the potentiometer is 9 o'clock).  The first plot is a comparison of the current PSTAB err signal (blue) with the one we used to have before Oct 22nd (#47250), when we lowered the power on the photodiodes.

From plot 2 we can observe that the error signal is now much less glitchy, and also from the omicron scan the glitches at 1 kHz seem to have disappear (plot 3).

The attached plot shows a qualitative improvement over the frequency and ampitude of the glitches as seen by PSTAB_PD2 after the increase of power on the photodiodes. Investigation in progress.

Since Oct 23rd we observe some strange oscillations at 12 kHz on the power stabilization error signal (plots 1 and 2). Looking at the fast channels it seems triggered by a saturation of the corrections occurring at 3.5 V.

This started after having lowered the impinging power on the photodiodes and having increased the gain and it can be explained by some saturation occuring in a component of the locking electronics.

As a temporary solution we increased again the power and lowered the gain, as it used to be before Oct 22nd. The loop seems more robust, even if the rms on the error signal has increased (plot 3).

In the meanwhile J.P. Coulon will do some modifications on the spare locking electronics to overcome this issue.

In the projection reported in 46453 there was an error on the frequency axis, which was shifted by a factor 2. I attach here the corrected plot.

Today we performed a laser intensity noise projection on Hrec by opening the PSTAB loop at 1249742972 and closing it at 1249743115. We took as a calm period for the projection the gps 1249742750, during 60 s.

Compared to the last projection (46453) the level of PD1 (out of loop error signal) is very different: it is now a factor ~200 below the sensitivity curve. This might be explained by a different working point of the ITF (to be checked), since the last projection was done during the single B1 PD operation.