Based on the information in K-Log #37433, an additional investigation was conducted to determine whether any AC alarms had occurred.
The IM Heater and CRY Stepper were checked, and AC input undervoltage protection (AC-FAIL) errors were found. After the inspection, all alarms were cleared.

For details about AC-FAIL, please refer to the following website: link

Based on these results, it is presumed that a momentary voltage drop occurred throughout, including the center area, rather than only at both ends, although the extent of the impact varied depending on each device’s threshold.

IM HEATER
EX: AC-FAIL error
EY: AC-FAIL error
IX: No error
IY: AC-FAIL error

CRY Stepper
EX: AC-FAIL error
EY: AC-FAIL error
IX: No error
IY: AC-FAIL error

For reference:

The equipment that can be controlled via the web interface uses the following two regulated DC power supplies.
The PWR401MH is equipped with AC input undervoltage protection (AC-FAIL).
The PMX70-1A does not have this function; therefore, it is assumed to have recovered automatically after the momentary power interruption.

A list of the equipment used is available on the following page:
https://gwwiki.icrr.u-tokyo.ac.jp/JGWwiki/KAGRA/Subgroups/DGS/Projects/IMHeater
 

[Ushiba, Fujimoto]

Abstract

Toward the PRC/SRC length measurement using P-pol., we implemented an SRY lock check and SRM ADS using POP DC instead of AS DC.
We confirmed that the modified configuration works properly in the VERTEX Guardian, ASC_LOCK Guardian, and INITIAL_ALIGNMENT Guardian.

Details

In the PRC/SRC length measurement using P-pol., the beam is extracted from the OMM chamber by a steering mirror after reflection from OMMT2. Therefore, the AS signal becomes unavailable.

As a result, the SRY lock check and the SRM ADS, which use AS DC, no longer work.
To address this, we implemented the SRY lock check and SRM ADS using POP DC instead of AS DC.

Modification of the SRY Lock Check Function

As shown in Fig. 1, we modified the is_src1f_locked() function in lsclib.py so that the lock check is performed using POP DC instead of AS DC.
This implementation was copied from the function used in is_prc1f_locked().

Modification of the SRM ADS

The SRM alignment had been performed so as to maximize AS DC in SRY.
Therefore, it does not work in the current configuration where the AS beam is blocked.

We therefore decided to engage the ADS using POP DC.

Fig. 2 shows ENGAGE_ADS_FOR_SRY in the modified ASC_LOCK.py.

The specific changes are as follows. The previous settings are kept as commented-out lines.

• TRG_MTRX: changed from AS DC to POP DC
• INMTRX: changed from AS DC to POP DC
• INF_GAIN: changed from -0.3 to -0.45

The reason for changing the gain from -0.3 to -0.45 is as follows.
From the time-series data during ADS using AS DC, shown in Fig. 3, the AS DC power was approximately 1.5 times larger than the POP DC power. Therefore, the gain was adjusted to compensate for this difference.

Fig. 4 shows the behavior of the relevant signals when the ADS was engaged.
After the ADS was engaged, the SRM moved, and both POP DC and REFL DC increased.

We confirmed that the modified configuration works properly in the VERTEX Guardian, ASC_LOCK Guardian, and INITIAL_ALIGNMENT Guardian.

Summary

We modified the Guardians so that the SRY lock check and SRM ADS can be performed using POP DC instead of AS DC.

This modification is only a temporary treatment. After the PRC/SRC length measurement with P-pol. is completed and the AS signal becomes available again, the configuration should be reverted to the previous one.

[Fujimoto, Yokozawa, Saito]

A mirror was installed inside the OMMT chamber so that the light reflected from OMMT2 would reach the optical table on the X+ side. The reflected light was then directed into the RFPD. However, the DC output was approximately 1.3 V, with fluctuations in the signal also observed. Since the DC output was several volts when the RFPD was placed at OMC REFL, the position of the RFPD will be fine-tuned next time to maximize the DC output, and the condition of the RFPD will also be investigated.
 

• First, a mirror was installed inside the OMMT chamber so that the light reflected from OMMT2 would reach the optical table on the X+ side (Photo 1). The beam was then lowered using a periscope, and the optical power was measured to be approximately 12 mW. An ND filter with OD = 1.0, a mirror, a lens with a focal length of 50 mm, and the RFPD were then installed. The optical power measured just before the RFPD was approximately 1.5 mW. Next, a rotatable polarizer was placed in front of the lens (Photo 2), and the polarizer angle was adjusted to maximize the optical power. The measured power was approximately 1.0 mW. The BNC cable that had been used for the OMC REFL PD was then connected to the DC output of the RFPD so that the signal could be monitored through CDS. The mirror alignment was adjusted to maximize the DC count, resulting in approximately 2200 counts, corresponding to approximately 1.3 V. When the RFPD was placed at OMC REFL and approximately 1 mW of light was injected, the DC output reached several volts after optimizing the mirror alignment. Therefore, it is likely that not all of the incident light is being coupled into the RFPD. Fluctuations in the signal were also observed. Next time, the position of the RFPD will be fine-tuned to maximize the DC output, and the condition of the RFPD will be investigated.

I checked the guardian and found that this filter was not implemented into the guardian, so this filter should be turned on manually at some point.
I also checked when the filter was turned on (fig1) and checked the klog around that  day but I could not find the related log.

If this filter is really necessary, it should be turned on again, and also be managed by the guardian.

I checked ETMX and ETMY suspensions by engaging local controls step by step.
There seems no serious issue and both suspensions can reach LOCK_ACQUISITION state without any problems.

[Nakagaki, Yasui]

I tried to turn off the heater on the 8K shield, but this power output was already 0W.

We checked the situation and found that the AC alarm was active on this power supply.

It appeart that the heater shut off due to a drop in the input valtage.

According to the temperature of the 8K shield top, the heater appeared to have stopped before 8:31(JST) on Aug 27.

Around this time we experienced a momentary power outage at this site.(klog 37420)

We checked the resistance of this heater and found no issues.

Anyway, we completed the pre-heating work.

[Kimura, Yasui]

We vented OMT-OMMT chamber to atmospheric pressure.

9:17 Turned off the interlock system for GVommt / 3.3*10^-5 Pa

9:24 Closed the GVommt / 9.1*10^-4 Pa

9:32 Started pressurization / opened the ball valve about 15 degrees and the angle valve by two turns)

9:35 3.5*10^2 Pa

9:55 3.1*10^3 Pa

9:59 Increased air flow (opened the ball valve fully and the angle valve by two turns)

10:19 1.0*10^4 Pa

10:35 1.6*10^4 Pa

10:42 Increased air flow ( increased the regulator's secondary pressure from 1.5 to 2.0 kgf/cm^2 and opened the angle valve by two turns)

10:55 2.8*10^4 Pa

11:30 5.7*10^4 Pa

12:00 8.1*10^4 Pa

12:14 9.4*10^4 Pa ->9.6*10^4 Pa

Included for informational purposes

A follow-up investigation found that, according to information from Hokuriku Electric Power Company, a power outage occurred in Kamioka at around the same time (08:22 on August 27, 2026), and power was automatically restored.
Although it isn't included in the town's name, so I can't say for sure, it may have been related.
 

Nakagaki-san, ikeda

The ETMX Picomotor controller had remained powered off.

The Picomotor controller had not been turned back on following the electrical equipment inspection with a blackout in March.

This issue was unrelated to the cause of the real-time PC reboot.

Corrective action:
We powered on the controller and confirmed that it could be accessed via MEDM.
 

Since ITMX NBDAMP_P2 didn't turn off automatically when ITMX went PAY_TRIPPED state, I manually turned off NBDAMP_P2.
If this filter is not managed by the guardian, the guardian should be changed.

I offloaded the BF GAS with the FR.

I offloaded the F0, F1, F3, and BF GAS filters with the FRs.

At approximately 08:22:44 JST on August 27, 2026, a momentary power outage occurred at both end, causing four computers k1ex1, k1ex0, k1ey1, and k1ey0 to reboot. All four computers were back online by 08:25:19 JST.
The ADC and DAC cards were recognized normally, indicating that the I/O chassis likely continued operating without interruption.
Only k1ey0 showed an unusually high IRIGB_TIME value of 27, compared with the normal value of approximately 11. To restore it to normal, all real-time models on k1ey0 were restarted.

Regarding systems other than the real-time models, the Picomotor controller cannot be reached by ping from k1ex1. This issue may be unrelated to the power outage, but it appears to require an on-site inspection and will be checked during the next visit.
 

I updated the Guardian for the DRMI ADS preparation so that the following parameters are automatically set:

• IMMT2 excitation frequency, amplitude, and phase
• Common filters for the BP and I signals
• Notch filters

I modified LOCKING_DRMI_1F state as follows (see also fig1).

1. Change gains of MICH1, PRCL1, and SRCL1 filter bank to 1 to avoid kick of the suspension during the lock acquisition (self.counter = 0).
2. Increasing gains of MICH1, PRCL1, and SRCL1 fiter bank to 9, 1.7, and 1.4, respectively after DRMI is locked (self.counter == 2).
3. Turning off MICH2, PRCL2, and SRCL2 gains when suspension waskicked during the lock acquisition to avoid additional kick before calm down (self.counter == -1).

Ushiba, Hirose, Tanaka

We modified IMMT2 ADS so that we can implement resG filters in PRCL, SRCL, and MICH controls to suppress the peaks at IMMT2 dither frequencies in their error signals.

We changed the dither frequencies from 30.125 Hz and 32.125 Hz to 9.125 Hz and 19.125 Hz and performed the phasing for these frequencies (fig.1).

According to the phasing results, we set the demod. phases. Also, There seem to be some peaks at -0.875 Hz and +1.125 Hz in bandpassed signals. So we implemented the notch filters for 0.875 Hz and 1.125 Hz in DEMOD_SIG_I.

After that, we engaged IMMT2 ADSs. They seem to work well. 

[Hirose, Tanaka, Ushiba]

Abstract:

Roll off filters were implemented into TM_OLDAMP_L filter bank of PRs and SRs.
These roll-off filters reduced the noise in PRCL, SRCL, and AS34 signals.

Detail:

In this morning, DRMI was not stable enough to continue the ASC work, so we checked the suspension motions.
Then we found that SRCL signals had no structure above 1 Hz, meaning that SRCL didn't seem to see suspension motions.

To address this issue, we implemented the roloff filters in TM_OLDAMP_L filter for PRs and SRs.
For PRs, we just implemented roll-off filters at 3 Hz while maintaining overall gains at low frequencies.
For SRs, we reduced overall gains to make the loop stable with 3 Hz roll-off filters.

Figure 1 and 2 show the filters we implemented for PRM and SRM, respectively.
Similar filters are also implemented into the other PRs and SRs.
All filters are implemented at FM3 (LOCK) of TM_OLDAMP_L filter bank.

Figure 3 - 6 show the spectra of MICH error, PRCL error, SRCL error, and sideband biuldup signals, respectively.
Black lines show the spectra before this work.
Red lines show the spectra after this work.
Blue lines show the spectra when all TM_OLDAMP_L filters are turned off.
The spectra of PRCL, SRCL, and AS34 signals improved a lot with this work.

Though blue spectra are slightly better than red spectra, we keep TM_OLDAMP_L ON because it takes a long time to calm down if there is no damping controls.
In addition, since the difference in spectra are less than 3 Hz and they do not limit the RMS, these differences should not affect both LSC and ASC.

Then, we implemented the filter in the VIS guardian.
FM3 of TM_OLDAMP_L will be turned on automatically between ALIGNED state and LOCK_ACQUISITION state.