This morning the interferometer was unlocking systematically in CARM NULL 3F. It seemed that the DIFFp TY loop was starting to oscillate as soon as we increased the gain (while still in drift control). We have commented this gain increase and the lock went well. In any case once the full bandwidth is increased the gain is adjusted automatically so this has no further impact on the DIFFp UGF.

While studying this we noticed that the flags that indicate the Full Bandwidth engagement were stucked at 2 since yesterday at noon. But this could not be real because while the flag has a value 2 the local controls are disengaged and we would have noticed. Diego noticed that the bridge between the DSPs and the DAQ that collects the data from the Gnames was not working. He restarted the SusDAQBridge process, as it was hanging again as a few weeks ago.

Some recent unlocks of the ITF in LN3 look like Fig.1, most recently this morning at UTC 07:25:16.

The cause seems to be very low coherence of the DIFFp_TY UGF monitor line, which disables the servo acting on the loop gain: eventually the loop enters its high frequency instability regime and starts oscillating at ~4 Hz, saturating B1 PDs. Shortly after, the ITF unlocks.

I have tested a substantial increase in the monitor line's amplitude from 0.75e-3 to 1.8e-3. The test happened online, the automation has not been changed. The coherence of the line rised to ~0.7 and the servo started operating again.

Fig.2, Fig.3 show the effects on DARM and on B1 PDs, I cannot spot any significant negative consequence on them of this adjustment.

If this is confirmed, the new line amplitude can be introduced in the automation and it will help prevent a few unlocks in the future.

we post the angular noise budget computed after a set of measurements performed this morning.

working conditions of the ITF: decentered position of the beam on the WE (12mm offset on vertical).

DIFFp TX (blue curve) shows higher coupling.

Today we investigated the issue experienced yesterday with the ID of the WE.

We performed a measument in order to recompute the calibration for the accelerometeres. We found out that ACC_H1 was the one miscalibrated, but the correct compensator was the one used before the last update. One possibility is that the accelerometer switches between two states. The reason of this behaviour is not clear. Some investigation is needed.

In the meantime, we reloaded the older compensation filter, and we modified the DSP card adding one switch for the two compensators in order to have the possibility to move between one 'state' and the other.

Today we started the recovery of the interferometer by working on the West Arm.

We could see from the beginning two (one and its reflection?) white spots on the WE camera, which we could not remember (see Figure, WARM unlocked but flashing).

We looked for flashes on both B8 and, after misaligning NI, B4, without finding any, even with the "snail". We then checked the old position of WE MAR LCs and they were quite different with respect to the current ones. We recovered those and the snail could find flashes quite soon, first on the camera markers signals, then on the photodiodes.

We could easily relock the West Arm (no change in demodulation phase or anything), but the locks did not last much, and we could see a very big oscillation on the MAR correction at about 80 mHz. Fabio pointed out that the Inertial Damping of the WE was not working as expected, probably since quite a while.

Valerio had a look at the ID loop, but in the end we kept it open (Position Control) and somebody will look at it tomorrow morning.

With the West Arm locked we checked the drift control on SWEB and slowly discharged its memory; also, we re-engaged the BPC loop for the West ALS. Then also the ALS could be locked without issue.

We also did the initial tuning of FmodErr, which was quite off, without problems.

At this point the automation was put back together so that later in the evening we could do some tests of the full lock acquisition. The reverted changes are the same ones of last time, with a few minor differences:

• SQZ_MAIN is still standalone, due to the ongoing activity on squeezing;
• the automatic 6 mm setpoint for WE_Y in CARM_NULL_1F has been disabled;
• the minor changes done to ALS_NARM for the ongoing OptChar measurements have been reverted, and they will be put back again next week when we'll continue such work;
• the bypass of the DOWN condition for the Etalon loops has been removed from NARM_LOCK and put back in ITF_LOCK, since now it is in charge again.

yesterday, due to the test I was doing during the shift, I disangaged the automatic BStx fbw transition in the ITF_LOCK metatron node. Once I finished the shift, I forgot to mention that the node should have been reloaded after one unlock. Indeed after the unlock of yesterday night, today the ITF was running with BStx in drift control.

I engaged the fbw by hand at around 14.48 UTC (adjusting mode for 1 minute).

Please at the next unlock reload ITF_LOCK metatron node.

today we performed some test for BS blending strategy in order to find one proper combination of parameters (blending filters and gain of the low frequency brench), to see if the interaction with PR tx loop could be reduced on the basis of what we understood in the past months.

None of the actions seemed to have a big impact on the interaction at 1.2 Hz. the most significant results are reported in fig.1 and fig.2 where the error signal spectra of PRtx and the correction spectra of BStx are shown respectively.

the legend speaks for itself:

• standard blending, higher LF gain (1 instead of 0.67). this didn't cause any improvement or major modification, besides an expected worsening of the correction (blue);
• the blending with higher crossing frequency (higher order with poles at 3.22 Hz), with higher LF gain, where the 1.2 Hz got damped, but an instability at around 2 Hz started to arise (red), together with a major worsening of the correction spectra. The situation was recovered when the gain was restored to its nominal value (yellow).
• the major effect on the loop was obtained switching to the older blending (first order, poles 1 Hz), where the 1.2 Hz oscillation improved a bit (purple), together with other region of the spectra. On the other side, the 400 mHz oscillation got excited. Indeed in the first place we moved to the 2Hz blending in order to get rid of this instability. Changing the gain didn't help as well but caused only the increase of the 1.2Hz.

in conclusion, working only on the blending didn't help to solve the problem.  the only knob left is to work on the control filters of the two loops.

we left the itf locked in LN3 with the standard working configuration of BStx.

fig.1 LSC budget march 8th vs DARM;

fig.2 LSC budget march 8th vs Hrec;

fig.3 LSC budget march 10th vs DARM;

fig.4 LSC budget march 10th vs Hrec;

clean data of march 8th were characterized by a noise bump up to 25 Hz.

march 10th measurements were more significative. CARM slow shows poor coherence wrt to hrec.

yesterday we increased the LF gain only of about a factor 1.3, which allowed to damp a bit the 1.2 Hz  interaction with PR.

from recent lines injections on PR and BS driving we noticed that the optimal ratio between the calibration weigths of the two BS branches (B1p and B4) might be wrong of a factor 3.

we could try to increase up to this value the gain of the LF branch to see if we can improve better the fbw response of the loop and reduce further the interaction.

It is also true that yesterday we were in the middle of a thermal transient, and this could have impacted a lot the optical gain of the 50 MHz signal (which depends on both the 6 and 56MHz optical gains), but anyhow this gain change is worth to be tried.

Profiting of a long lock (gps start 1424152564 for almost 36 hours) I ve compared the camera centering signals with the dithering signals and the B7 and B8, for the north and west signals respectively, QPD DC h/v error signals.

The TF and the coherence are shown in the attached plots, highligthing a good coherence below ~0.1Hz (apart from the WIx).

This study will be repeated once the signal from the cameras will be obtained with the new subregion (which detect only the beam).

An example of ASD between camera image signal and dithering signal for NE Y and WI Y is shown in figure 9, it is visible that the noise floor in much lower in the case of camera signal.

it is worth to be noted how the arm cavity are controlled

this morning we found the ITF locked with BStx in fbw, BSty in drift control. The BNS range was above 53 Mpc. We didn't want to affect the science data taking phase, so we left the ITF like this and we just uncommentend in ITF_LOCK.py the engagement of BS AA loop.

at the next unlock please reload ITF_LOCK metatron node.

looking at the injection of yesterday we could fine tune the gain of the loop by increasing it of about a factor 1.2 in order to have the right UGF and the right gain at LF. 

the engagement went fine.

fig.1: time domain response comparison of older filter VS new filter.

fig.2: spectra comparison. As expected, the filter gains more at the region of 2 and 3 Hz. In that region, also the phase margin has been a little been improved. There is a predicted worsening in some LF region, but the overall accuracy has not been touched.

we leave this control on line by default. we will evaluate the response also in better weather conditions, with less wind and sea activity.

Today I carried on the tests for the BS_TX loop using EDB_B1s_QD1_50MHz_Y_I that started a while back (entry 66104).

Giving a quick look at the spectra of the in-loop signal and the new one, back when the last test was done and today, no big difference can be spotted (Figures 1 and 2 respectively).

I first used the line at 3.1 Hz to verify the last measurement: line amplitude 2e-3 since 15:53:00 UTC onwards, while in LOW_NOISE_3. After finding the weight (with ASC_BS_TX_INPUT, not ASC_BS_TX otherwise the loop gain will cause a miscalibration) I attempted the handoff, failing twice (LOW_NOISE_3 and _2 respectively). The measurement is also not obvious given that the DIFFp_TX line nearby (3.3 Hz) muddles the reading.

A good attempt was done injecting the line (amplitude 5e-3) from 17:42:12 UTC to 17:56:40 UTC, while in LOW_NOISE_2; I found a ratio between the weighted in-loop ASC_BS_TX_INPUT and B1s_QD1_50MHz_Y_I equal to 0.015 (Figure 3).

At 18:01:33 UTC I made a successful handoff, while in LOW_NOISE_2, from the blending B1p_QD1_50MHz_V_I_LP/B4_QD1_6MHz_V_HP to B1s_QD1_50MHz_Y_I. Clean data from 18:01:40 UTC to 18:13:15 UTC.

At 18:13:17 UTC I injected again the same line, with the same amplitude; the line was turned off at 18:23:30 UTC, clean data afterwards.

In Figure 4 the in-loop signals with the usual one (in purple) and the B1s one (in blue).

At 19:01:53 UTC I moved on towards LOW_NOISE_3 with this configuration; clean data in LOW_NOISE_3 starting from 19:11:59 UTC.

I leave the interferometer in this configuration, which is not macroscopically detrimental to the sensitivity (Figure 5); at the next unlock the lock acquisition will bring back the usual loop configuration.

Today we had an unlock, at 12:52:13 UTC, as SRCL oscillated at ~2.25 Hz (Figure) causing then the saturation of B1 and, eventually, the unlock.

During the afternoon shift we put online a new filter, SRCL_ctrl5c, which should address this problem by removing some unnecessary structures and gaining more at low frequency. This is intended to be used instead of SRCL_ctrl5b in LOW_NOISE_3.

We haven't tested it yet not to mix things up with the ongoing shift, but it will be tested soon.

injections of last Feb 10th.

Fig.1-2: projections of SFFS I and Q sig to DARM and Hrec respectively.

Fig.3-4: TFs during the injections of SSFS vs DARM and Hrec respectively.

No major changes wrt to the previous set of measurements.

During an analysis of ITF angular stability, it has been observed that PR_TX low frequency residual rotation seen by AA signal can be well explained assuming that the source of noise is the actual rotation of NI. The AA loop imposes PR to follow NI rotation, which is large enough to be seen by NI and PR optical levers: the coherence between the two is almost 1 at 40 mHz and the TF is flat (fig 1). There is also coherence with all the other TMs of the arms: all together seem to have an excess noise in COMMpTX dof. This is evident comparing the OptLev signals in DIFFpTX and COMMpTX combinations (fig 2). The actual DIFFpTX rotation is forced to be very small (about 1 nRad) by the loop on B1p_QD, so the corresponding combination of the optical levers gives the level of noise of those sensors: we can see in figure that COMMpTX motion is much larger than the sensor noise.

The component of noise below 0.1 Hz is very likely due to the centering loop on the dither signal. The one used for COMMpTX is NI_MIR_Y_AA; in fig 3 we can see that it is noiser than the other three. We can also see a different shape below 40 mHz: it should be the suppression due to the COMMp slow loop. I would say from this plot that the gain is too high respect to what is really needed.

In fig 4 the coherence between two witnesses of this noise is shown: PR_TX is coherent with the west arm centering signals in the combination given by a rotation of the axis, while the combination given by the axis translation (soft mode) is not coherent. This is another confirmation that the excess noise affects COMMpTX.

The first thing to do is to reduce significantly the gain of COMMpTX slow loop and see what happens. I would do it both for TX and TY, even if a similar analysis for TY has not been done yet.  The there should be the possibility to reduce a bit the noise using a different error signal and excluding the noisy one. The COMMp combination of NE, WI and WE seems to be the cleaner option.

The first part of the shift was dedicated to check the different loops connected to the quadrants: galvo centering (using the 112MHz), the one that adjusts the phase by minimizing the DC component and the phase noise subtraction. The first two can be seen in Figures 1 and 2 for QD1 and QD2 respectively. The latter one works well after some gain tuning, it can be seen in Figures 3 and 4. Alain made available the corresponding buttons on the VPM to turn them on and off.

Then we unlocked in order to get to LN2, but the guardian of the NE opened. Looking a bit closer, we realized that the north green arm was not locking but Metatron thought it was and it proceeded with the lock, kicking the north suspensions. It looked like the north arm was misaligned, but we called the ALS expert, since we already had some problems with the north bench earlier in the week. He realigned the green beam but it was too late to continue the shift so we ended it here.

preliminary entry. detailed report will be posted tomorrow.

given the models that we recently designed, today we worked on the blending strategy for BStx loop in full bandwidth, in order to investigate the coupling with the PRtx loop at 1.2 Hz.

• we tested a new blending HP-LP (higher order with poles at 3.11 Hz). We tried it in LN2, by hand. The full bandwidth transition worked fine, but the 1.2 Hz oscillation is still there.
• by looking at the data we think that the calibration weights between the LF and HF signals might be wrong of about a factor 3. We performed some lines injections at 1.6 Hz and 5.8 Hz both on BS and PR driving. We will use this measurements to better characterize the loops.
• with the new blending we explored new weights for B4 branch, by reducing it of about a factor 10 (situation similar to have only a Low-passed B1p error signal in loop). In this case we didn't noticed any major change in the signals, and also on the amplitude of the 1.2Hz. We also increased the weight of almost a factor 2 with respect to the nominal one, with the result of slightly get worse with the 1.2 Hz instability (note that in the past, the same test with the former blending at 2 Hz, caused a higher increase of the oscillation). 

data have to be better analyzed but we think we can improve the situation working also on the main controller for BStx.

at the end of the activity we reversed everything and we put the nominal weigths and the usual belnding filters.

Some issue concerning the stability of the BStx AA has been observed. Often, when the loop is closed in full bandwidth, a coupling with the PRtx loop is observed, which lead to the arise of an unstable loop oscillation at ~1.1/1.2 Hz and eventually to the unlock.
We remember that the full bandwidth strategy for the BStx loop is based on a blending between the error signals coming from two quadrants: b1p for the low frequency, and b4 for the high frequency. The blending filters are a 1st order low-pass and high-pass with pole (and blending frequency) at 2 Hz (moved few times ago from the original value of 1 Hz).
The coupling between the two loops is a well known issue which comes into play when b4 is put in the loop, sensor which read both the motions of BS and PR. This coupling causes a modification of the closed loop response (modification of the PR openloop with a drop of stable phase at 1.1 Hz) and so the hypothesis of SISO control strategy for the two loops, the one which is based the control design, becomes less reliable.
In order to understand better and to address the problem, we've been trying to design a simple MIMO model describing this coupling, to see if it was possible to reproduce the unstable behavior experienced in the data.

Considering the relationship y=Gx where, y is the output vector of an opto-mechanical system, x a sensing input vector and G the MIMO transfer matrix, we can describe the system (only for bs and pr for now) by writing:

(ybs; ypr)=[Gbs, Gbs, 0; 0, c*Gbs, Gpr]*(b1p; b4; b2)

Where the term c is an arbitrary frequency indipendent coupling term, which if put to 0, leads back to the SISO modeling of the system.

The closed loop response can be obtained rewriting the eq in terms of open loop outputs, including in the model also the control terms u:

(TFbs(s); TFpr(s))=[Gbs(s)*LP(s), Gbs(s)*HP(s), 0; 0, c*Gbs(s)*HP(s), Gpr(s)]*(ubs(s); ubs(s); upr(s))

Without getting into too much details, the interesting results are the following in fig.1 and fig.2: in fig.1 an unstable PRtx spectra has been used to 'tune' the coupling term in order to have the same unstable frequency between model and measurement, while in fig.2 are reported different closed loop response for different blending filters (blue and red) and the stable closed loop spectra with no 1.1 oscillation (yellow)

One can see that actually, if we want to believe to this model, the decision of moving ahead the blending frequency from 1 Hz to 2 Hz was not the best, but caused a slight worsening of the stability of the system (this is a bit counterintuitive).
However, based on this modeling we can decide to reshape better the HP-LP blending filters in order to damp/get rid of the instability, by trying to push up the loop phase in that region.

---‐---‐-------------

Additionally, an other model to describe this coupling interaction has been made with Paolo through OCTOPUS in which the interaction between the PR and BS payloads system has been designed with properly tuned elastic links in terms of stiffness.
In this model the results were quite the same and allowed to obtain the closed loop response of the system, tuning the unstable oscillation modifying the control terms within the model.
Although, if I remember correctly, with OCTOPUS the comparison between the two blendings (1 and 2 Hz) were giving different outcomes (improving instead of worsening).

The conclusion of this study is that in order to get rid of the unstable coupling, one knob is to modify more the blending filters than the control filters.

One last step is to allign the two models in order to have exactly the same results with two different approaches.

After the long lock of the weekend, on Monday 23/12 the interferometer started to struggle, not in the lock acquisition, but in keeping LOW_NOISE_3, where we consistently unlocked upon arrival or shortly after, with a ~2.3-2.5 Hz oscillation.

We tried to figure out what, at the end of the lock acquisition, could go wrong or leave the ITF in a unstable state:

• we disabled the BS full bandwidth, to no gain;
• we disabled the new PR/BS reallocation (which is still off), as we could also see some glitches in the top-stage corrections;
• we checked the MICH/SRCL demodulation phase: although the one used in CARM Null was quite good, the second one we put just before opening the slow shutter at the end of the transient was not, so that one was changed;
• in general all DRMI UGF servos were struggling as, maybe because of the bad weather, they had low coherence so the servo was not effective all the time; the lock which survived up to this morning was characterized by setting by hand a SRCL gain quite higher to the one that the servo put (before it lost coherence) (0.26 vs 0.2), which makes sense considering that the last thing we do in the lock acquisition is to weaken the SRC.

I put higher (~50-80 %) lines for MICH and PRCL in ACQUIRE_LOW_NOISE_3, and more than +100% for the SRCL one; today after the Maintenance the ITF locked at first attempt, with the servos behaving better (but also the weather is better today) (see Figure); I will adjust offline a couple of amplitudes for the next locks (PRCL looks overkill now), and consider whether to raise the SRCL UGF setpoint a bit from 0.2 (arbitrary units) as the current configuration doesn't have that much margin. I already put a very minor increase to 0.21, to be monitored.

Given the time coincidence of the excavator moving near the central building there's a possibility of correlation with the unlocks/ mulfonctioning of the reallocation strategy at the central suspensions.

The reallocation engagement will be enabled if at the next unlock it will be reloaded ITF_LOCK metatron node.

The first unlock, which was caused by a drift from zero of BS TX, was preceeded by a little 'glitch' (?) on the local control of the BS TX, see fig.1. In that stage we are not in full bandwidth and this is a problem that sometimes occurs when we are still in drift control with the loop (https://logbook.virgo-gw.eu/virgo/?r=65234).

The few times we were able to go to LN3 we were sistematically unlocking during the engagement of the BS TX  in full bandwidth.

After one last intervention, its engagement, instead of being performed after the SRty misalignment, was being done at the beginning of LN3 acquiring.

During this stage, a 1.2 Hz oscillation started to arise which increased fluctuations on B1s DC (already quite high, around 20 mW) and B1PD1 Audio channel up to saturation and to unlock.
This oscillation was visible also on the High frequency error signal of the BS tx loop which looked like a divergenge caused by some loop instability. This was indeed cured recently by moving the blending frequency of the LF and HF branches above this point, to 2 Hz. The inloop signal itself indeed didn't seem affected by this oscillation, but it was visible also on PR TX and other signals.

The 'correct' engagement point has been restored after the misalignment of SRty, when B1s fluctuations are smaller, see fig.1 and fig.2. The mechanics of interplay between the arise of this oscillation on the BS V HF signal and the witnessing on B1s DC and other AA signals is not totally clear. 

The 1.2 oscillation is still there, but less harmful. To mitigate further this instability, some other shaping of the blending filters can be done. I will work on that.

BNS range now is still below 50, but around 47 Mpc. With respect to a better range in the past (november 10th), we still have (besides the violin modes to be yet damped) some excess of noise between 15 and 100 Hz, fig.3.

SRCL filter and UGF line modification performed few days ago are still there, I didn't reverse them, so we are still running with the older filter for SRCL.