By the way, did you add and turn on several circuits in the EXV area? If so, the temp increase by ~ 0.6 °C can be explained. Anyway, Takahashi-san performed the off load for F0 and F1 because it drifted a lot. 

At present, no "small" cooler in the EXV area for the temp compensation.

[Kenta, Yuzurihara]

For the lockloss (2025-04-06 20:26:15 UTC), the lock duration was longer than other recent locks (50335 s = 13.98 hours)!! So, we checked the stability of the lock.
There are two isses: (1)​​ t​​​he BPC for EY pitch was drifting away from 0. (2) the unknown excess in 500~2000 Hz started in the middle of the lock

Details

Figure 1 shows the BPC error and feedback signals during the lock. We found that the BPC for EY pitch had drifted so that the error signal (K1:BPC-PIT_ETMY_OUTF_INMON) was away from 0. See the y-corsor located at y=0. (In my understanding, there was no BPC control for IY pitch.)

Figure 2 shows the normalized the spectrogram of the strain signal. From -5 hours before the lockloss (~15:00 UTC), the excess in 500~2000Hz had appear. The BPC frequency for EY pitch and yaw was 860 Hz and 855 Hz, respectively . So, this excess could disturbed the BPC control.
Note that, during this lock. the laser power and AS17 were stable (Figure 3).

Figure 4 shows the non-Gaussianity monitor (Gauch) of the corresponding time. I can't find any coincident change (also in other time and channels).It's possible that the excess was not far from Gaussian noise or that the time resolution should be much more large to catch this excess.

Figure 5 and Figure 6 show the behavior around 860 Hz and 855 Hz of the strain channel. I picked up them from line monitor (developed by Takahiro S Yamamoto-san). The amplitude of the injected line for BPC was changing in time, as seen Figure 1.
In addition,the magnitude of the spectrum was different before and after 15:00 UTC. This is corresponding with the finding in Figure 2.

Figure 7 shows the channels related to the PMC. Once the PMC temperature control was saturated, it recovered from the saturation.

[Kenta, Yuzurihara]
We​​​​​​ performed the lockloss investigation for the recent lockloss from the OBSERVATION state of the LSC_LOCK guardian, between 2025-04-05 03:54:16 UTC and 2025-04-07 05:54:54 UTC. The previous lockloss investigation was posted in klog33153.
During this period, there were 10 lockloss. The plots are summarized in wiki.

PMC HV Saturation

The 1 lockloss (2025-04-05 03:54:16 UTC) was occured because the PMC HV was saturated. Figure 1 shows the screenshot that K1:PSL-PMC_PZT_HV_MON_OUT_DQ touched the limit value (300).
During this period (4/4~4/7), the most of lockloss was occured under the the saturation of K1:PSL-PMC_HEATER_INMON (15000). Except for one lockloss, the direct lockloss cause was not related to the PMC saturation.

1Hz seismic motion made the IMC lockloss

For 2 lockloss, we saw the excess of the seismic motion with 1~10 Hz, which made the oscillation of the IMC length control and made the IMC lockloss.

• 2025-04-06 06:02:46 UTC (Fig)
• 2025-04-07 00:53:24 UTC (Fig)

Oscillation of OMC DC PD

This phenomena was observed many times in the lockloss in March. In this period, this phenomena was observed in 1 lockloss. (2025-04-06 20:26:15 UTC), as seen in Fig 4. The cause of the oscillation was unclear.

Oscillation of OMC DCPD over 5~10 seconds

• 2025-04-06 02:53:10 UTC
• 2025-04-06 03:09:56 UTC

This new phenomena were observed in 2 lockloss. It took 5~10 secods for the OMC DCPD to oscillate and saturate, as seen Fig5 and Fig6. The oscillation frequency was 44 and 43 Hz, respectively.
As shown the figures, the OMMT2 TRANS was not increased just before the lockloss. It indicates that the laser power arriving the AS port was not increased. But, the laser power of OMC DCPD increased with 5 seconds... We need more samples....
When the OMC DCPD started to oscilalte, the power of IR trans dropped coincidently.
There was no strange behavior on AS17 (=DC OFFSET value).

Others

For 1 lockloss (2025-04-07 05:54:54 UTC), there was no hint, except for the saturation of K1:PSL-PMC_HEATER_INMON. The OMMT2 TRANS was stable.

For 3 lockloss, the excitation was underway. I will let these lockloss be.

Abstract

The front-end calibration was updated with today's measurement results.
There is no large change in the actuator efficiency of TM from previous ones which was measured on Feb. 19th klog#32745.
Because we had large changes in the mirror temperature in Mar. it was not an enough evidence that actuator efficiency had been stable during Mar.
On the other hand, an assumption in recent OLTF measurement on Apr. 2nd (klog#33198) and updated on Apr. 4th (klog#33260 that 1.5dB change in OLTF came from optical gain (klog#33213) seemed to be almost correct. (Mirror temperature was ~80K in Feb. and Apr., but it changed as 80K -> 40K -> 80K in Mar.)
A change in the actuator efficiency of IM is slightly large but it's not so serious for sensitivity evaluation now. So it will be checked later.

Update of the front-end calibration was done at 18:58:26 JST.
So it's valid from an IFO lock at 19:08:59 JST.
 

Changes

Actuator efficiency of TM (Fig.1-2 shows measured TF to DARM)
old: 3.7791e-14 * (10Hz/f)^2
new: 3.7947e-14 * (10Hz/f)^2 (+0.4%)

Actuator efficiency of IM (Fig.3-4 shows measured TF to DARM)
old: 1.39319e-14 * (10Hz/f)^4
new: 1.62888e-14 * (10Hz/f)^4 (+16.9%)

Optical gain (Fig.5-6 shows measured OLTF)
old: 248.094dB
new: 247.993dB (-1.2%)

MN_LOCK_L (Fig.7 shows total filter shape of all FMs)
added 40Hz band stop filter at FM6

IM_LOCK_L (Fig.8 shows total filter shape of all FMs)
added 40Hz band stop filter at FM1

LSC_OMC_DC
removed +2dB and -2.08dB gains at FM2 and FM3, respectively.

The most direct way to investigate the mystery of the power discrepancy would be to measure the power right after the secondary lens, where the IR beam radius would be sufficiently shrinked, when the relevant chambers will be opened.

Abstract:

I tried to lock DARM with newly designed hierarchical actuators but failed.
Though the reason is not clear, DARM loop oscilated at 2.1Hz.
According to the phase behaviour around 2.1Hz, there seems to be negative zero around there.
So, decoupling of the actuator might solve the problem.

Detail:

First, I measured transfer function from ISC_INF_EXC to DARM1 with only turning on one of MN, IM, and TM actuator switches with ALS_CARM_LOCKED state (fig1:right column).
Pink, cyan, and black lines show the TF when only TM, I, and MN actuators were opened, respectively.
I slightly tuned the notch filter frequency for MN filters but basically the designed filter seemed to be as expected.

Then, TF from ISC_INF_EXC to DARM1 with turning on all actuator switches with ALS_CARM_LOCKED state (fig2: left column).
Red and blue lines show the TFs with nominal filters and newly designed filters, respectively.
With the initial design, hollow around 1.9 Hz is too deep, so I tuned the notch filter of TM _LOCK_L to mitigate it.

After that, I tried to lock the ALD_DARM with new hierarchical actuators but started to oscillate at 2.1Hz (fig2).
2.1Hz gain should be enough high after engaging ALS_DARM, so it doesn't seem gain peaking.
According to the TF in fig2, the phase was rotated by 360 degrees between 1.7Hz to 2.5Hz, which implies that there was a negative zero around that frequency.
If there is a negative zero around 2Hz, actuator sign at 2.15 Hz resonance would be flipped, so oscillation might be explained.
I'm not so sure why there was a negative zero but one possibility is actuator coupling from the other DoFs, so actuator decoupling should be performed.

[M. Honjo, Michimura]

We have reduced the PD gains for TMSX and TMSY polarization monitor PDs to avoid saturation in 10 W PRFPMI operation.
Power at TMS PDs seems to have half the power than estimation for some reason (as was the case in 2018 and 2022).

Summary of PD whitening situations:
  X_IR_PDA1: 0 dB, no whitening gain, no whitening filters
  X_IRSPOL_PDA1: 0 dB (was 10 dB), no whitening gain, 2-stage whitening filter
  X_IRPPOL_PDA1: 10 dB (was 30 dB), no whitening gain, 2-stage whitening filter
  HWP angle: 20 deg
  Y_IR_PDA1: 0 dB, no whitening gain, no whitening filters
  Y_IRSPOL_PDA1: 0 dB (was 10 dB), no whitening gain, 2-stage whitening filter
  Y_IRPPOL_PDA1: 20 dB (was 30 dB), no whitening gain, 2-stage whitening filter
  HWP angle: 80 deg
  See klog #30902 for previous settings. Note that X and Y have different settings now.

Calibrated polarization spectra:
  - We calibrated measured spectra into polarization rotation angle using the factor described in klog #30885. But with following changes

if ARM=='X':
    PDp2Rad=1/(4*(300.7-16.3)*np.sin(4*np.deg2rad(20-10.1)))*10**(20./10)  # Calibration from klog #30113, but calibration was done in 30 dB, but now in 10 dB
    PDs2Rad=1/(4*(297.0-19.7)*np.sin(4*np.deg2rad(20-54.7)))*10**(30./20)  # Calibration from klog #30113, but calibration was done in 30 dB, but now in 0 dB
    PDp2Rad/=130*10.8/1.1/2  # 130 for PRFPMI instead of single arm, 10.8/1.1 for input power difference, 2 for installation of additional BS in TMS (klog #32183)
    PDs2Rad/=130*10.8/1.1/2
elif ARM=='Y':
    PDp2Rad=1/(4*(358.0-24.1)*np.sin(4*np.deg2rad(80-72.5)))*10**(10./20)  # Calibration from klog #30827, but calibration was done in 30 dB, but now in 20 dB
    PDs2Rad=1/(4*(352.0-24.2)*np.sin(4*np.deg2rad(80-117.4)))*10**(30./20) # Calibration from klog #30827, but calibration was done in 30 dB, but now in 0 dB
    PDp2Rad/=130*10.8/1.1/2  # Same with X (klog #32181)
    PDs2Rad/=130*10.8/1.1/2 

See attached for the original data in counts and the calibrated data for X and Y.

TMS power budget:
  - Using the following, power transmitted from ETMX and ETMY would be 420 mW and 395 mW, respectively. Since we have 4 BSs to IR_PDA1 now (JGW-T1808962), power at IR_PDA1 would be 26.2 mW and 24.7 mW, respectively.

PIMC=10.8
PRG=13.0
BS=0.5
T_ITMX=0.444/100 # PhysRevApplied.14.014021
T_ETMX=6.8e-6  # PhysRevApplied.14.014021
T_ITMY=0.479/100  # PhysRevApplied.14.014021
T_ETMY=6.92e-6  # PhysRevApplied.14.014021
RTLX=50e-6  # Round-trip loss klog #30823
RTLY=60e-6

def cavitypowertransmission(T1,T2,L):
    L=L-T2   # Remove ETM transmission from round-trip loss
    t1=sqrt(T1)
    t2=sqrt(T2)
    r1=sqrt(1-T1)
    r2=sqrt(1-T2)
    rloss=sqrt(1-L)
    return (t1*t2)**2/(1-r1*r2*rloss)**2

T_XARM=cavitypowertransmission(T_ITMX,T_ETMX,RTLX)
T_YARM=cavitypowertransmission(T_ITMY,T_ETMY,RTLY)
P_TMSX=PIMC*PRG*BS*T_XARM
P_TMSY=PIMC*PRG*BS*T_YARM

 - During 10.8 W PRFPMI lock
K1:TMS-X_IR_PDA1_INMON = 10200 cnts
K1:TMS-Y_IR_PDA1_INMON = 9530 cnts
  These corresponds to 10.6 mW for X and 9.83 mW for Y, using 40/2**16 V/cnts, 1.5e3 Ohm transimpedance gain, and 0.39 A/W for PDA100A2. These are roughly a factor of 2.5 smaler than the estimated power above.

Past TMS power measurements and discussions:
  - According to measurements in October 2022 (klog #22456, klog #22339), power around RLNS2 was 529 uW for X and 480 for Y with 2.9 W input, single arm. For 2.9 W single arm, using T_PRM=0.135 (JGW-L1605744), they should be 1170 uW for X and 1100 uW for Y. Here, measured values are a factor of 2.2 smaller. Note that these measurements were done with a power meter. So, the discrepancy is consistent between power measurement methods.
  - This was also the case in December 2018 (klog #7415, klog #7419).
  - Note that all the calculations above assume 100% modematching of IMC transmission to the main IFO. This should be a good approximation due to DARM shot noise estimates and past modematching measurements.
  - We are consistently loosing half the power since 2018. Something inside BRT???

Next:
  - Redo polarization calibration under 10 W PRFPMI configuration
  - Set HWP angle to reasonable values and make PD gains same for X and Y

[Kimura and Yasui]
　To improve the convenience of cryo-cooler maintenance,
the CRY Group has installed special tools for the cryo-coolers in the X-end and Y-end machine rooms.
 The special tools are stored in the KTC tool box shown in the photos.

I offloaded the F0 and F1 GAS filters with the FRs.

Yokozawa-san turned off the FFU on the pre-booth of the PR3 at 15:02 today.

I compared the accelerometers' ASDs before and after this work. There are no significant changes.

I analyzed the magnetic injection data on April 3rd, based on the coupling function model.

• Fig.1: ASDs of the strain and the magnetometer (at BS), for the reference and each injection time
• Fig.2: Coupling Function upper limit,  (Bx^2 + By^2 + Bz^2)^0.5 -> strain
• Fig.3: Projection upper limit

CAL group

We performed DARM calibration.
The measurement dir: CAL/current/measurement/2025/0407/
The measured files: 1454_*xml
Measured TF:

• ETMX_TM -> DARM
• PCAL -> DARM
• DARM OLTF
• ETMX_IM -> DARM
• (ETMX_MN -> DARM)

During the last measurement (ETMX_MN -> DARM), the lock was broken. Probably our excitation kicked a resonance  around 40Hz.

Abstract:

We observe a significant reduction in the height of peaks around 20 Hz and 40 Hz.
Inserting resistors may be effective in reducing the remaining prominent peaks caused by pitch coupling in IX and IY.

Detail:

Thanks to recent improvements in the DARM sensitivity, we can now observe a significant reduction in the height of several suspension resonant peaks around 20 Hz (Fig.1) and 40 Hz (Fig.2).

All of the peaks around 20 Hz, along with some around 42 Hz, have decreased by more than an order of magnitude.
This suggests that these peaks likely originate from the tower suspension resonances.

Currently, the most prominent peaks around 40 Hz are the pitch resonances of IY at 44 Hz and IX at 45 Hz (klog:33236), as indicated by the blue vertical cursors (Fig.2).
These may be excited by the IM coil drivers.
Therefore, inserting additional resistors after the coil drivers for IX and IY might help reduce the height of these peaks, in addition to recent efforts to damp or subtract them using ASC signals (klog:33266, klog:33275).

I will estimate the current vertical-horizontal coupling and pitch coupling based on these peaks and calculate the associated nonlinearity as klog:33131 and klog:33162 soon.

Abstract:

Reducing the TM coil driver noise would be highly effective to improve the BNS range.

Detail:

I created a simple noise budget plot (Fig.1) using the measured DARM spectrum from 04:10 UTC on April 6th, during one of the best sensitivity periods observed so far.
The budgeted noise sources—suspension thermal noise, shot noise, and actuator (coil driver) noise—are identical to those shown in the figure in klog:33015.

The measured noise in the 70–90 Hz band, which currently limits the BNS range the most critically, can be almost explained by a combination of suspension thermal noise and actuator noise.
We still likely see some contribution from input jitter noise in the 100–400 Hz region.
The next steps should include replacing the TM coil driver to reduce the actuator noise and optimizing the interferometer alignment to suppress the input jitter noise.

By the way, there is a hypothesis that the actual actuator noise might be 1.8 times larger than expected (klog:32132).
However, if we assume a 1.8× increase, the estimated total noise would exceed the measured noise (Fig. 2), suggesting that this hypothesis might be ruled out.

It seems like DHARD_Y has better SNR than DHARD_P to see peaks around 45 Hz. Interestingly, POS_SPOL also sees this motion (could be due to ITM birefringence). See attached. These channels would be useful for noise subtraction as well (sadly, they are 256 Hz DQ channels). See attached for noise subtration result using DHARD_Y with simple bandpass filters around these peaks.

The sudden decrease of the temperature of 50K_REFBRT_HEAD happened. However, it is still around 80K, while 55K for the arm side

Abstract:

I designed the new hierarchical actuators from scratch.
MN and IM crossover frequency is 0.4Hz and IM and TM crossover frequency is 4Hz in the new design.
I will test it when the IFO can be used next week.

Detail:

As reported in the previous post, it seems very hard to optimize the hierarchical actuator by modifying current filters, so I designed the filter from scratch.
First I measured the ETMX TF by exciting each stage from CAL excitation port to keep the actuator balance as LSC actuators.
During the measurement, I turned off MNH, IMH, BFH dewhitening filters to avoid saturation.
After the measurement, I designed the filter at foton file for SUSMOD as shown in fig1.

figure 2, 3, and 4 shows the overplots of measured TF and suspension models I made for TM, IM, and MN, respectively.
TM TF seems well modeled while phase of IM seems different around 2-2.5Hz.
Also, MN TF seems to have lower gain at high frequency but it should be fine because MN will be only used for low frequency.

According the the suspension model, it is hard to avoid crossing TM and IM actuators around 2.2Hz if we set the crossover frequency at 3 Hz, which is current nominal value.
So, I slightly increased the crossover frequency and set it at 4Hz in the new design.
Also, it was found that setting crossover frequency between IM and MN around 1Hz due to the complex resonant structure around 1Hz, the crossover frequency between MN and IM is at 0.4Hz in the new design.

Final design of the hierarchical actuator can be seen in fig 5.
I will try to test the new hierarchical actuators next week.