As we were worried, the temp of EY 50K REFARM HEAD started increasing again because of the violation of the naive condition of cooling.

The water level for the FB laser is near the lower limit. We need a refill.

We need regular checks as CRY and VAC on Friday.

The 10 hours later data from the recent report. EX 50K REFBRT HEAD temp is still fluctuating around 93K as Fig.2

In the case of EY 50K REFARM HEAD temp in Fig.1. it also started fluctuating around 86K and one day later, it started cooling.

In addition, the vacuum level (EXT Vacuum level ~ 2*10^-5 Pa) is not as bad as in 2023 (EXC: ~10^-3 Pa at its peak), and it is also fluctuating at this level. Be careful not to have data at EXC because of some trouble.

50K REFBRT HEAD temp stopped increasing and is now fluctuating at ~94 K. According to the EY Arm side case, this is a good sign. 

EXY Im temps. EX Im temp is increasing. EY Im temp is near constant.

I performed the lockloss investigation for the recent lockloss from the OBSERVATION state of the LSC_LOCK guardian, between 2025/03/18 04:43:0 UTC and 3/19 15:40:20 UTC. The previous lockloss investigation was posted in klog33036.
During this period, there were 47 lockloss. The plots are summarized in wiki.

1Hz seismic motion made the IMC lockloss

For 8 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.

Oscillation of MICH and PRCL with 1.65 Hz

For 1 lockloss (2025-03-20 01:27:34.437500 UTC), there was an oscillation in the MICH and PRCL error signals with 1.65 Hz (Fig). I didn't find the coincident oscillation in the oplev signals. 

Oscillation of ITMY and ETMY with 0.9 Hz (or 0.88Hz?)

For 3 lockloss, there was an oscillation in the ASC signal (especially CSOFT) with 0.9 Hz (Fig). For one lockloss, there are coincident oscillations in the PRM pitch, PR2 pitch, and ITMX pitch with 0.88 Hz (Fig).

• 2025-03-20 05:56:20.812500 UTC
• 2025-03-20 06:14:23.062500 UTC
• 2025-03-20 09:07:25.562500 UTC

Oscillation of OMC PD over minutes

The amplitie of OMC PD looks growing over several minutes (Fig).

Others

Sometimes, we saw the saturation of OMC DCPD signals just before the lockloss. Sometimes, there was no saturation. In addition, the time it takes to saturate depends on the case.  The cause of the OMC saturation is still unclear.
Based on Ushiba-san's suggestion, I will try to characterize whether the OMC alignment is good or bad before the lock loss by using the other PD channel at OMMT.

Memo

Even though the lockloss occurred in the end, there was a recovery from the OMC oscillation twice.

• 2025-03-20 00:06:20.812500 UTC (Fig)
• 2025-03-20 14:45:35.562500 UTC (Fig)

2025/03/28

Kimura and Yasui

 We turned on the TMP at 15:00.

[Yokozawa, YamaT]

Abstract

We installed Coil Driver Switcher for the TM stage of ETMX.
Now we can reach LOCK_ACQUISITION state and excite the TM stage only via HPCD.
For switching LPCD, model update is required (maybe on the next maintenance day?).
 

Details

Installation of Coil Driver Switcher board
As an operation test of the Coil Driver Switcher was completed in klog#32943, we installed it for ETMX_TM. The Coil Driver Switcher hasn't been assembled to a formal chassis yet, S-number is not assigned for the chassis. A S-number of the used circuit board is S2516468. After preparing front/rear panels and a formal chassis, it should be replaced for a proper management of S-number for installed circuits. From the view point of gain adjustment, coil balance, and calibration, it may be better to keep using the same board until finishing O4c. S2516468 was installed rear side at U42 of EXV2 rack as shown in Fig.1. Output port of this board is connected to the feedthrough at the EXV chamber (a cable which had been connected to the output port of TM HPCD was moved to the output port of this board).

Connection with Coil Driver chassis
Two DB9 inputs of this board are connected with the output ports of TM HPCD and TM LPCD for CRYp. HPCD (S2315283) has been used in the past and wasn't changed in this work. LPCD for CRYp (S1604776) which had been installed for the IM blade damp and hasn't been used until today was reused for this purpose (see Fig.2). So there is no new installation of the coil driver chassis.

BIO cables
BIO cables were stolen from ones for BFV coil driver chassis for this purpose (see also Fig.3). Though we cannot enable the de-whitening filters for BFV now, it should be no impact for the daily operation because we haven't used de-whitening filters for the tower stages. k1iopex1 has now 5 BIO cards and one more BIO card may be able to be installed. If we will try it and will return stolen BIO cables to BFV, we probably need a whole day as a maintenance time.

Operation check
After the installation, we requested LOCK_ACQUISITION for VIS_ETMX guardian and it works well. Because there is no DAC output for TM stages in this state, I ensured that HPCD path is now enabled by using awg excitation. Figure 4 shows the TM Oplev signals with excitations from a HPCD path of the TM stage. For switching to a LPCD path and checking the injection via LPCD, we need to modify the real-time model. It will be done in next(?) maintenance day. Anyway, IFO lock with HPCD is probably now available.

Note
[1] DC power for the switcher board is derived from LPCD chassis. So we need to turn on LPCD before HPCD. On the contrary we need to turn off HPCD before LPCD. (see also Fig.5).
[2] A cable between the feedthrough and the coil driver output of IM blade damp is still left in EXV2 rack as shown in Fig. 6. If we don't need IM blade damp for future, they should be uninstalled.
[3] 'open' and 'close' of BIO are assigned for using HPCD and using LPCD, respectively in the current connection. In the case of Z-switch of OMC, 400Ohm (BIO 'close') has many peaks compared with 100Ohm (BIO 'open'). Though we haven't been sure the actual reason yet, BIO 'close' may induces undesirable noise due to additional connection to GND. If we will face many peaks when we swtich to LPCD, it may be better to try swapping the BIO assignement as 'open' for LPCD and 'close' for HPCD.

Abstract:

The nonlinearity of the DC readout generating the DARM error signal probably starts to limit sensitivity in the 60–90 Hz region.
Additionally, force noise applied to the type-A intermediate masses (IMs) may contribute to the sensitivity.
I propose unplugging the cables or inserting a resistor immediately after the coil driver for the IMs to reduce actuator efficiency, and we can quickly assess whether this improves sensitivity.

Details:

During recent observations, I noticed that when a resonant peak at 40 Hz or 50 Hz was excited, additional peaks at higher frequencies also appeared.
This behavior suggests the presence of nonlinear effects in the DC readout.
To investigate further, I analyzed the DARM spectrum in a quiet observation state.

The DC readout is fundamentally quadratic, or more precisely speaking, follows a 1 - cos(x) function.
The relationship between DARM power P and displacement x is given by

P/P_0 = 1 + 2 x/x_0 + (x/x_0)^2,

where P_0 is the offset power of 15 mW, and x_0 is the offset displacement, approximately 2 pm (dividing 7 pm, estimated from klog:31350, by the square root of the input power ratio) in recent KAGRA operations.
Typically, we assume x ≪ x_0, neglecting the quadratic term.

To model the nonlinear response, I considered prominent peaks at 8 Hz, 20 Hz, and 40 Hz (IM vertical and pitch modes), as shown by the blue curve in Fig.1.
These peaks were modeled based on a white force noise input, appropriate vibration isolation from the IM and TM, and the DARM open-loop transfer function.
The Q-values for the 20-Hz and 40-Hz peaks were assumed to be 10³ and 10⁴, respectively, with the latter inferred from recent cryogenic Q-value measurements (klog:32476).
These modeled curves were then visually fitted to the measured DARM error signal (OMC-TRANS_SUM_DC_OUT), normalized by P_0.
This spectrum corresponds to x/x_0, controlled by the DARM loop.

Using ifft in matlab, I generated a time-series representation of the spectrum, added the quadratic nonlinearity, and reconstructed the spectrum.
This revealed that some small peaks in the 60–90 Hz range were caused by the nonlinearity, as shown in Fig.2.
The peaks around 60 Hz originate from the nonlinear coupling between the 20-Hz and 40-Hz peaks, while those around 80 Hz correspond to the second harmonic of the 40-Hz peaks.

A simple order-of-magnitude estimate suggests that the spectral amplitude of the 20-Hz and 40-Hz peaks is 10⁻³ and 10⁻⁴, respectively, leading to nonlinear contributions in the range of 10⁻⁸ to 10⁻⁷.
To suppress these below shot noise (which is 5×10⁻⁹ /√Hz for an offset power of 15 mW), all peak amplitudes must remain below 10⁻⁴ /√Hz.

Additional Concern: Excess Force Noise on IMs

Another issue is that the noise tail from the 40-Hz peaks might be limiting the spectral floor at 60–80 Hz.
While I modeled the observed peaks using a white force noise input, the inferred force noise level is unexpectedly high.

For instance, assuming a reasonable vertical-horizontal coupling of 1% or a pitch miscentering of a few mm, the expected force noise should be on the order of 10⁻⁹ N/√Hz. However, the inferred force noise level is two orders of magnitude higher than the expected thermal noise and the coil driver noise, which is calculated by approximately 10⁻⁸ V/√Hz in the output voltage spectrum.

Proposal:

To test whether IM actuator noise contributes to the observed sensitivity limitation, I propose:

1. Unplugging the cables connected to the IM coils from the coil driver, or

2. Inserting resistors (e.g. 50Ω) immediately after the coil driver to reduce actuator efficiency.

Since the IM actuators at EY, IX, and IY are actively used for only a few narrow-band damping controls, this test can be conducted more easily, and we can confirm whether the sensitivity will be improved or not.

Yea and Yer compressors stopped due to cooling water problems. Restarted after cooling water was restored.
P-53, P-54, and P-55 compressors continued operation, but alarms were displayed, so compressors were temporarily stopped and restarted in order to reset the displays.

I took the TCam photos for four mirrors at 8:24 ~ 8:31 this morning. I checked the images.

• ETMX
  • original
  • fitting result
  • overlay with previous reference center
• For ETMX, I updated the reference position. I also re-draw the yellow line shown at the k1mon4. For the other mirrors, we don't need to update the reference positions.

Finally, the BRT4 temp reached the critical temp of 157 K. Then water seemed to start evaporate.

I increased the heater power as shown in fig1.

Just for your information.

Fig. 1 show the spectra as

• Red and blue: 2025 Mar 27 13:37 UTC; IFI-IMM-PRM chambers are in air but all the flanges closed; IO Guardian state is PROVIDING_STABLE_LIGHT
• Green and brown: 2025 Mar 24 13:21 UTC; the other conditions should be the same as above.

For the conditon evidence, Fig. 2 show the IO Guardian state (1000=PROVIDING_STABLE_LIGHT) an the pressure for the IFI-IMM-PRM chambers.

Seems very getting quite.

 A trend graph of Q-mass in EXC is attached.
The partial pressure of each residual gas in the EXC has decreased due to the switch to turbo molecular pumping.

 Attached are trend graphs of the temperatures in the compressors of the cryocooler and Q-mass in the EYC.
The temperatures in the compressors temporarily rose due to the cooling water outage, but have now returned to normal values.
It was also confirmed that there was no change in the partial pressure of residual gas in the EYC after the cooling water was stopped.

I check if HPBD at IFI chamber can catch the PRM reflection beam whe PRM is in the MISALIGNED_BF staate.
Figure 1 shows temperature of HPBD and PRM guardian state.
Just after reaching the PRM at MISALIGNED_BF state, temperature of HPBD was increased significantly.
So, HPBD seems to catch the PRM reflection.