Abstract:

I tried to reduce the Type-A suspension kick when lockloss happens.
Since it seems difficult to set proper ISC WD threshould and the speed of suspension kick is not to fast, I modified ASC_LOCK guardian to turn off ASC feedback to MN stage as early as possible when lockloss happens.

Detail:

To set proper threshould for ISC WD by using ASC signals, I first checked the signals of K1:VIS-ITMY_ISCWD_WD_AC_Y_RMSMON during the OBSERVATION state and when lockloss happened.
Figure 1 shows the largest signals of the trigger channel during the silent run yesterday.
Figure 2 shows the smallest signals of the trigger channel when the lockloss happened.
Though the signals in fig2 is slighty larger than the signals in fig1, it doesn't seem good way to set the threshould between them because the difference is small.
So, I gave up to set ISC WD threshould.

Then, I checked the ASC feedback signals when lockloss happened (fig3).
Since feedback signal change is about 1 cnt even after 300ms (about 5 smples with 16Hz) after lockloss, it would be enough slow to stop feedback by guardians.
So, I modified the DOWN state of the ASC_LOCK guardian as shown in fig4 and fig5.

Since the guardian doesn't change OBSERVATION state, this modification oesn't affect to the IFO during OBSERVATION.

Motivation

The problem of a long time to reach steady state of some BPC for PRM YAW was pointed out in klog34073. After that, to improve the situation, the gains of ITMY BPC PIT and YAW were updated by Ushiba-san in klog34197 on 6/13.
As a follow-up analysis, I chceked the trend data of recent lock with long lock duration.

Result

Figures 1~4 show the timeseries of several BPC controls. In terms of K1:BPC-YAW_PRM_OUTF_OUTPUT, it reached the equilibrium in 12~28 minutes. It is relatively shorter than previous.

In the previous work (klog34073), it was reported that it took 30 minutes to reach an equilibrium of BPC for ITMX YAW. But, this problem does not appear to be happening now.

On the other hand, I found that K1:BPC-YAW_ETMY_OUTF_OUTPUT and K1:BPC-YAW_ITMY_OUTF_OUTPUT reached the equilibrium in ~40minutes after OBSERVATION state. Figures 5~7 show the examples (same lock as Figure 1~4. T-cursors focus on the ITMY and ETMY).

We were in the mine from around 10:40 to 11:45 JST.

Abstract:

I increased the gain of ITMY BPC PIT and YAW by a factor of 3 and 12, respectively.
To achieve this state at OBSERVATION state, I modified the DOWN and SET_BPC_FREQ_FOR_OBSERVATION of ASC_LOCK guardian.

Detail:

 As reported in klog34073, current BPC for ITMY is very slow, so I increased the gains of BPC.
Since PIT and YAW loop takes 30 minutes and 2 hours for reaching steady state, I increased the gain by a factor of 3 and 12 for PIT and YAW loop, respectively.

To implement this modification into the guardian, I modified the DOWN state (fig1) to revert K1:BPC-{PIT,YAW}_ITMY_INF_GAIN and SET_BPC_FREQ_FOR_OBSERVATION state to implement new gains (fig2).
After modification, I confirmed the new guardian works well.

[Ushiba, Hirose]

Abstract

REFLPZT3,4 are the DC centering loop actuator for QPD3,4 of WFSf3. As reported in klog34133, it was possible that PZT3 PIT was failed. We went into the mine to check if the PZT was working. We found PZT3 PIT was shorted. We also placed a dumper to dump the WFSf3 QPD3 optical axis.

Detail

As reported in klog29939, the PIT and YAW signals for PZT3 are connected to the two output ports of thorlab's 3-port PZT controller. I checked the YAW control of PZT3 to see if the PZT controller itself could be broken. The YAW control on PZT3 is behaving as if the error signal is going to zero in response to the feedback signal. The PIT on the PZT3 is not act to the control signal, as reported in klog34133. (FIG1) I checked if the QPD3 signal saturates on the digital system when at 10W, but it does not be saturated. (FIG2)

In the mine, we checked the PZT controller monitor. It should have been coming 75V, but it was showing about 3.0V.
After turning off the offset in the digital system, we unplugged the cable at the output of the PZT controller. When we turned on the offset in the digital system again, it displayed 75V. (FIG3)
So it is most likely that the PZT element is shorted. The PZT mirror mount and cable needs to be replaced.
Since PZT3 is BS as shown in the REFL table layout, replacing the mirror mount will take a little time including alignment adjustment. At this time, the dumper was placed downstream of the QPD3 PICO without replacing the mirror mount. (FIG4)
Currently, the cable of the QPD3PIT output of the PZT controller is unplugged. 
And, the ASC guardian's PREPARE_PZT state automatically sets the PZT offset for ASC's DCcentering loop. To avoid inadvertent Guardian changes, we left with the offset 75V on the degital system in this case.
SDF changed: klog34190

As responce of klog34102, I investigated the related channels of ASC, K1:ASC-REFL-PZT1_YAW, K1:ASC-REFL-PZT3_PIT, and K1:ASC-REFL-PZT4_PIT

Currently, the related PZT channles are used for the DC contorls for REFL QPD1 YAW, REFL QPD3 PIT, and REFL QPD4 PIT, respectively. And the controls seem to be saturated when the LSC_LOCK guardian is in the OBSERVATION state (Fig.1). So we need to offload them with picomotors to solve this situation.

To solve the issue, I made following changes in OBSERVATION state.

1. Set offset of 0.097 at K1:BPC-YAW_ETMY_INF_OFFSET.
2. Increase gain of BPC by a factor of 10 by changing the gain at K1:BPC-{PIT,YAW}_ETMY_INF_GAIN from 1 to 10.
3. Change demod phase of ETMY YAW BPC for swapping the sign of the signals.

To achieve above modification without affecting the other states, I modified several states in the ASC_LOCK guardian as follows:

DOWN state:

Turn off OFFSET switch of K1:BPC-PIT_ETMY_INF filter bank.
Also, set the gain of 1 at K1:BPC-{PIT,YAW}_ETMY_INF_GAIN.

SET_BPC_FREQ_FOR_OBSERVATION state:

Change setting demod phase at K1:BPC-PIT_ETMY_DEMOD_PHASE from 47 to -133.
Turn on offset switch of K1:BPC-YAW_ETMY_INF filter bank.
Change gains at BPC-{PIT,YAW}_ETMY_INF_GAIN from 1 to 10.

I checked the alignment during the longest lock last night and found that large drift can be seen in Y arm (fig1).
Since there seems no effect on X arm, the reason of the drift is very likely to be ETMY BPC, which controls Y arm SOFT mode.

According to the error signals of ETMY BPC (bottom two signals in fig2), following two things might be a problem:

1. ETMY PIT BPC doesn't seem to go zero. The sig of the loop might be wrong.
2. Large offset happened in ETMY YAW BPC error signas, which was reported in klog33989.

First, I will start from flipping the sign of ETMY PIT BPC.

[Komori, Ushiba]

Abstract:

We aligned the IMMT1 and PR2 angle with respect to the OpLev values when the best sensitivity was achieved (2025/4/29 21:30 UTC).
Then, we tweaked INP2 and PRC2 offsets and found that the current offset value seems still working well in terms of the peak height at 116 Hz.
Also, we found that BPC seems slowly drifting and degrade the alignment, which causes the lockloss, so BPC should be retuning not only for better sensitivity but also for the stability improvement.

Detail:

Since IMMT1 and PR2 OpLev value was far from those when the best sensitivity was achieved, we aligned these mirrors with respect to the old OpLev values.
Figure 1 shows the time-series data during the alignment.
There seems no significant effect on PRG and arm alignment.

Then, we once turned off INP2 and PRC2 ASC offsets to see what happens.
Figure 2 shows the comparison of the DARM spectrum when offsets were zeros (blue) and offsets were nominal values.
The peak at 116 Hz seems reduced, so the nominal offsets value seems better than at least zeros.

During the work, we found PRM is drifting in +yaw direction with several tens of minutes after reaching OBSERVATION state (fig3).
This drift seems degrades the IFO alignment and causes the lockloss, so the longest lock during commissioning today is about 30 minutes.
So, this issue should be fixed.

[Guo chin, Yuzu]
I made the histogram to count the phenomena before the lockloss between 2025/02/10 and 04/21.|
Recently, the number of the lockloss related to the IMC oscillation is the dominant phenomena. After 4/3, the lockloss related to the oscillation (saturation) of OMC drastically was mitigated.

This time, I focused on the lockloss from OBSERVATION_WITHOUT_LINES. Note that I counted the number manually. It's possible to shift the number a bit.

Important date for commissioning

• 2/19 : 10 W operation started (klog#32748) (klog#32756)
• 4/3: engaging whitening filters for GAS to mitigate the 20 Hz resonance kick by DAC noise (klog#33197)

After the maintainance, I tried to recover the IFO and found that DSOFT_Y loop started to oscilate (0.5 Hz) at INCREASING_LAS_POWER.
So, I reduced the DSOFT_Y gain by a factor of 2 by turning of FM9, and then oscillation was stopped.
Figure 1 shows the ASC signals around the time when I turned off the FM9 at DSOFT_Y.

I'm not so sure if the UFG becomes too low but I keep this configuration today.

Increasing the ASC gain can help improve the stability of the interferometer, reduce the overall RMS fluctuations of the arm transmission and AS_RF17_Q, and potentially enhance the DARM sensitivity.
As a first step, I increased the DHARD_{PIT,YAW} gains by a factor of two.

Although the 2-Hz fluctuation in the arm transmission slightly increased, the ASC loop appears to remain stable.
I will leave this increased gain in place overnight to monitor its long-term stability.

[Tanaka, Komori]

Abstract:

We attempted to optimize the IX pitch BPC offset.
A negative offset may be preferable, but further confirmation of its reproducibility is necessary.

Details:

This analysis is based on measurements conducted last Wednesday.
One potential approach to reducing input jitter noise is adjusting the beam spot position on the input test masses.
As a first step, we focused on the IX pitch BPC, as its miscentering is significantly large.

We applied an offset of +0.04 and -0.03 to K1:BPC-PIT_ITMX_INF_OFFSET, as shown in the top-right panel of Fig. 1.
In both cases, lock loss occurred when the actual offset reached around ±0.01 (top-left panel), after a clear reduction in arm transmission and reflection power.
This indicates that the interferometer approached a critically coupled state.
Further investigation is needed to understand the cause of the lock loss to maintain the offset at this level.

Figure 2 compares the DARM spectra for negative (brown), zero (blue), and positive (red) BPC offsets.
The 53-Hz peak and the input jitter noise around the 100-Hz region appear to improve significantly and slightly, respectively, with the negative offset.
However, the reproducibility of these improvements must be verified.
This negative offset is desirable as it reduces the IX pitch miscentering.

In the near future, we plan to examine the sensitivity with offsets applied to other BPC degrees of freedom, including IX yaw and IY.

Detail:

TFs from CHARD_{P,Y}_SM_EXC to TM{P,Y} of each Type-A OpLev were meaasured with compensation filters for ETMY.
Figure 1 and 2 show the ratio and relative phase between ETMX OpLev and the others.

The coherence of yaw measurements seems high around UGF of ASC yaw loop (0.6 Hz), so I calculate the relative actuator gain.
Since the coerence for pitch measurements seems low around UGF of ASC pitch loop (0.2), I performed the same measurement with swept sine excitation (fig3).
According from these results, I calculated the actuator matrix for ASC as shown in fig4 (pitch) and fig5 (yaw).

To check the actuator balance of yaw, I measured yaw TF with swept sine excitation after actuator balancing (fig6).
Thanks to the compensation filter, gain difference is less than +/-0.5dB and phase difference isless than +/-5 degrees, which is much smaller than without compensation filter.

Then, I measured TF of CHARD_Y loop with new balanced actuators to check if the small phase margin around 0.3 Hz was improved or not.
Figure 7 shows the OLTF of CHARD_Y loop.
Though phase delay around 0.1-0.2 Hz seems slightly better than before, phase delay around 0.3 Hz is still somehow large.
Since I only performed actuator balance of TM stage, it would be better to check the actuator balance of MN again.

Abstract:

Though I adjusted relative gain of TM stage of each Type-A suspension, CHARD TF didn't change significantly.
Next step is to check MN actuator efficiency.

Detail will be posted later.

Abstract:

I made compensation filter to adjust ETMY suspension plant to ETMX one.
The filter was made at FM4 of K1:VIS-ETMY_TM_LOCK_Y filter bank.

Detail:

To achieve the better actuator decoupling of ASC, I made compensation filter for ETMY yaw plant.
Figure 1 shows the TF from TMY excitation to TMY motion of each test mass.
ETMY has a slightly larger phase difference around 0.3 Hz and 0.8 Hz compared with the other suspensions.
So, I designed compensation filter for ETMY susension plant.

Figure 2 and 3 show the suspension model of ETMX and ETMY I designed, respectively.
Though high frequency gain and phase are slightly different between models and measurement, gain and phase below 1 Hz seems well matched.
Since the current UGF of yaw ASC is 0.6Hz, it should be fine at this moment.

According from the zero-pole vales of the model, I made a compensation filter (fig4).
I will test this filter when I willhave time.

Abstract:

I started to fine tuning of ASC after VIS local damping control update.
For DHARD_P and CHARD_P, tuning was finished but I noticed DHARD_Y and CHARD_Y plant is very different.
I suspected the actuator balance of ASC.

Detail:

After finishing the VIS local control tuning, I started to tune the ASC controls.
{C,D}HARD_P loop was tuned (fig1, fig2).
I just changed the overall gains and notch filters for these loops.

Then, I tuned {C,D}HARD_Y loop (fig3, fig4) but notied that phase margin around 0.1-0.3 Hz of CHARD_Y loop is too small compared with DHARD_Y loop and hard to proceed further tuning.
So, I suspected that this difference comes from actuator balance between suspensions, which makes the notches n the TFs due to some phase mismatch and causes the coupling between diferent DoFs.
So, I measured the TF from K1:ASC-CHARD_P_SM_EXC to each TM OpLevs (fig5).

Since the difference of the resonant frequency, TF of ETMY has a slightly larger difference compared with the others.
So, it would be necessary for compasation filter for the difference of ETMY resonances.

Note:

Data are stored at /users/Commissioning/data/VIS/TypeA_general/2025/0217/.

Although very late, I summarized the cross-check of the December measurements and the comparison of the September(klog31128) and December(klog32115) measurements.

Did the cross-check for this measumerement.

As the cross-check of this measurement, I compared the plot from the Ushiba-san's script (klog31047) with the plot from my script.
These plots were plotted only for results what I phase and Qphase coherence both above 0.7.
At this time, the resulting plots were almost identical. So the data plottings seem to be correct. (FIG1, FIG2)

Compared the measurements between September and Decemeber.

I summurised the comparement between the september(klog31128) and this time(klog32115).
As for what has changed when compared to September and December, QPD3_RF11 phase was only rotated demodulation phase -95deg in klog31210.
Both measurements were plotted only for results what both I-phase coherence and Q-phase coherence were greater than 0.7.

The degrees of freedom plotted in both measurements had almost the same phase relationship between the degrees of freedom. 
The phase of each degree of freedom with high coherence was within ±35deg (≒arccos(0.8)).（FIG3, FIG4）
It is difficult to consider changes in the magnitude of each degree of freedom. The degree of freedom with the largest difference in measured values was 3.141 times. (klog32259.pdf)

This post summarizes the work conducted last Friday.

The improvement achieved by setting the offset on ASC-INP2_P_OFFSET has shown high reproducibility during each IFO lock.
Therefore, I have left the offset at +1.0 in the filter bank.
This value will be re-tuned when the optimal input alignment for the IFO changes in the future.
I also attempted to optimize the offset of INP2 yaw but did not observe any significant changes in the DARM sensitivity, so the yaw offset remains at zero.

At this point, we are uncertain whether the improvement is due to better input alignment for the IFO or improved alignment specifically to the CARM PD.
To distinguish between these two effects, it will be useful to adjust the alignment of the mirror right before the CARM PD directly by hand while the IFO is locked, as an initial trial.
If the improvement is found to be related to the alignment of the CARM PD, we can consider installing a pico-motor on that mirror for remote control adjustments.

Komori, Tanaka

We tried to decrease the sensing noise of the CARM PD which is caused by the jiter to the PD by adjusting the beam centering on CARM PD. We adjusted the offset of IMMT2 PIT ASC (ASC-INP2_P_OFFSET) for this purpose. Fig. 1 show the result. We found that the bumps decreased when the positive offset was applied. So this indicates that the bumps from 300 Hz to 380 Hz in DARM sensitivity may be reduced by CARM PD centering

>On the other hand, according to Ushiba-san, there are a high pass filter in the IP_BLEND_ACC{L,T,Y} filter after ACCBLEND_FLDACC{L.T,Y}_OUT so the large DC value in those OUT should be cause of the saturation of IP controls. However, the cutoff frequency (80mHz) of high pass filters seems to be faster that the one (0.1 mHz) of the integrater so the DC value seems to remain in IP_BLEND_ACC{L,T,Y}_IN1. So if the guardian go to LOCK_ACQUISITION even though the DC value in IP_BLEND_ACC{L,T,Y}_IN1 still large, the saturation of IP controls will occur.

Even though IP_BLEND_ACC{L,T,Y}_IN1 is large at DC, it should be fine because there are high-pass filters at IP_BLEND_ACC{L,T,Y} filter banks and DC values are cut.
However, since high-pass filter cutoff frequency is 50/80mHz and not so fast, if we requested LOCK_ACQUISITION soon after the suspension is tripped, IP_BLEND_ACC{L,T,Y}_OUT values can be still high, which causes large feedback signals to IP when switching IP caontrol to inertial damping.
So, we need to wait engaging inertial damping control of IP until IP_BLEND_ACC{L,T,Y}_OUT values become small enough (enough longer than the time constant of high-pass filter).
Now, new function was implemented into the guardian (klog32491), this problem can be avoidable.

Komori, Tanaka

## What we did

• We measured the TF from the TM actuator to TM oplev by injecting a white noise from CHARD_{P,Y}_SM_EXC. At that time, in order not to excite the MN stage, we turned off MN_LOCK_OUTSW_{P,Y} in advance.
• This time, we adjusted the OUTMTRX element value for ITMX, ITMY, and ETMY, respectively from the ratios between ETMX and the others at 0.1 Hz in the P case or at 1 Hz in the Y case which frequencies are near each UGF so that each DC gain get the same.
• Then, we measued the TF from the MN actuator to TM oplev with current LOCK filters by injecting a white noise from CHARD_{P,Y}_SM_EXC. At that time, in order not to excite the TM stage, we turned off TM_LOCK_OUTSW_{P,Y} in advance.
• We adjusted the overall gain of MN_LOCK filters to make the crossover frequency 0.1 Hz. Fig. 1 (PIT) and Fig.2 (YAW) show the results. The cross over seems to be 0.1 Hz, the DC gain seems to be almost the same. We use these actuators to try to lock ASC.

## Note

I made a mistake to apply too large excitation (~6000 cnts) to the MN stage in performing the above TF measurement. And then, ITMX and ITMY got tripped because the NBDAMP_Y4 with TM oplev signal kicked BF YAW due to TM oplev got out of range by exciting MN largly and reached the BF RMS to the WD threshold.

After this RMS calmed down, I requested the LOCK_ACQUISITION to ITMX and ITMY guardians but they stopped in the CALM_DOWN state maybe because the IP control seems to be saturated. At once, we set ITMY and ITMX to SAFE state

ACCBLEND_FLDACC{L.T,Y}_OUT and IP_FLDACCINF_H{1,2,3}_OUT seems to be stored much larger value than usual due to this trip. I and Yokozawa-san tried to clear their integrator history and then we could restore ITMX and ITMY to LOCK_ACQUISITION.

On the other hand, according to Ushiba-san, there are a high pass filter in the IP_BLEND_ACC{L,T,Y} filter after ACCBLEND_FLDACC{L.T,Y}_OUT so the large DC value in those OUT should be cause of the saturation of IP controls. However, the cutoff frequency (80mHz) of high pass filters seems to be faster that the one (0.1 mHz) of the integrater so the DC value seems to remain in IP_BLEND_ACC{L,T,Y}_IN1. So if the guardian go to LOCK_ACQUISITION even though the DC value in IP_BLEND_ACC{L,T,Y}_IN1 still large, the saturation of IP controls will occur. So we ask Yamamoto-san to implement the waiting state to calm down the input value and Yamamoto-san implemented it as reported in klog32491.

we performed the actuator balance for HARD/SOFT mode.

We excited CHARD P at 13.125 Hz and obtained the ratio between the ETMX oplev value and the others at 13.125 Hz. Then we input the obtained ratio to the ITMX, ITMY, and ETMY elements in OUTMTRX_P for CHARD_P, respectively. After that, we measured the ratio again and confirmed the ratios were 1 (fig.1).

Similarly, we performed the same for Y. In this time, we excited CHARD Y at 6.125 Hz. Fig.2 shows the result. The current balance seems to be fine for both P and Y.

I will put the same value in OUTMTRX for other DoFs (DHARD, C/DSOFT) later. 

Ushiba, Miyoki, Aso

On the new year's day, the IFO failed to switch to MICH 3f.
An obvious problem was that the POP90 was too low, suggesting bad alignment.
So we aligned the interferometer. Then we were able to lock PRFPMI with RF.

We still had several oscillation problems.

1. ETMX Pitch oscillates around 1.1Hz, especially when ALS DARM is engaged. After transitioning to the IR DARM, this oscillation tamed down.

2. ITMY Yaw oscillates around 1.1Hz. Ushiba-kun fixed it by enabling the damping loop for the recoil mass Yaw mode, which used to be disabled because this mode was not excited. Actuator imbalance caused by the cooling may have started the excitation of this mode (seems to have started about a day ago). We should check if this damping loop contaminates the sensitivity or not. If so, we should adjust the actuator balance and disable this damping loop again.

3. ETMX Yaw rings up around 1.65Hz when the arm ASC was engaged. We were able to tame down the oscillation by reducing the ASC gains.
Specifically, we inserted the gain of 0.5 in the FB9 of the following loops:
K1:ASC-DHARD_Y
K1:ASC-CHARD_Y
K1:ASC-DSOFT_Y

Our current theory is that a feedback instability is induced by the couplings between these loops. Probably the inter-loop couplings increased because of the actuator imbalance caused by the cooling. We need to re-optimize the actuator balances once the suspensions are fully cooled down.

For the moment, we leave the IFO in the RF locked state with the above mentioned temporal gain change. We need to remove these filters once we re-optimize the actuators.

Remember to do

• Disable the ITMY recoil mass Yaw damping loop
• Revert the ASC gains