>To address this, I reduced the IMC CMS in1gain by 3 dB

Did you reduce IN2 gain as well?
To keep the relative gain between IMC loop and CARM loop, we also need to change IN2 gain if IN1 gain is changed.

Abstract

We measured the sensing noise (dark noise and shot noise) of the CARM RFPD at REFL as part of the CARM noise budget.
By analyzing the power dependence of the demodulated noise spectra, we extracted the dark noise level and the shot noise coefficient, as well as the effective transimpedance of the RFPD.

Details

As part of the CARM noise budget, we measured the sensing noise (dark noise + shot noise) of the CARM RFPD at REFL.
The procedure was as follows: the reflected beam from the PRM was injected into the CARM RFPD, and the demodulated noise spectrum was measured for several PD input powers.
By fitting the power dependence of these spectra, we estimated the dark noise and the coefficients associated with shot noise.
The demodulated spectra were first amplified with a SR560 (high-pass @ 3 Hz, gain = 100) and then measured with a Moku:Lab.

Figure 1 shows the demodulated noise spectra taken at different PD input powers (the amplification by the SR560 has been corrected).
To extract the noise-floor values for each input power, we computed the mean and standard error over the frequency band from 50 kHz to 100 kHz for each spectrum.
This frequency region was selected to avoid the peak near 40 kHz.

Figure 2 shows the power dependence of the obtained noise-floor values.
The data were fitted using the following model to extract the power dependence of the PD dark noise and shot noise:

Fitting model:
$\delta V_\mathrm{RFPDnoise} = \sqrt{\delta V_\mathrm{dark}^2+\delta V_\mathrm{shot}^2}$
Definitions:
$\delta V_\mathrm{dark}$ : Dark noise of the RFPD
$\delta V_\mathrm{shot} \equiv R\sqrt{2e\alpha}\times\sqrt{P_\mathrm{in}}$: Shot noise after RFPD demodulation
$R$: effective transimpedance including demodulation
$e$: elementary charge
$\alpha$ : responsivity [A/W]
$P_\mathrm{in}$: input power to the RFPD

Result of fitting:
$V_\mathrm{dark}=3.96(1)\times10^{-8}\,\mathrm{V/\sqrt{Hz}}$
$R\sqrt{2e\alpha}=2.271(8)\times10^{-8}\,\mathrm{V/\sqrt{Hz\,mW}} =7.18(2)\times10^{-7}\,\mathrm{V/\sqrt{Hz\,W}}$

Assuming $\alpha=0.77\,\mathrm{A/W}$ (QE = 0.9), the effective transimpedance $R$ and the power-to-voltage conversion efficiency including demodulation $R\alpha$ are
$R=1.447(5)\times10^3\,\mathrm{\Omega}$
$R\alpha=1.114(4)\times10^3\,\mathrm{V/W}$

These results will be used for the CARM noise budget soon.

[Fujimoto, Tanaka, Komori]

Abstract:

We measured the transfer function and dark noise of the CARM CMS.
The results appear reasonable, and we have now completed the measurements necessary to perform the CARM noise budget.

Details:

We carried out the same set of measurements as those performed for the IMC CMS, as reported in klog:35669.
For the CARM CMS, we focus only on the fast path and ignore the slow path for now, because the slow path is used for MCE feedback and is negligible for the CARM noise budget.

The dark noise of the CARM CMS fast output was measured with a ×100 amplification using the SR560, and the result is shown by the blue line in Fig. 1.
We measured the spectrum in two frequency ranges, similar to yesterday’s IMC CMS measurements, and this time there was no discrepancy around the 1 kHz boundary.

The measured transfer function (blue dots) is compared with the modeled one (cyan line) in Fig. 2, and they agree well.
Using the modeled transfer function, I calculated the input-equivalent noise of the CARM CMS fast path (Fig. 3).

We also investigated the mismatch between spectra measured in two frequency bands reported in klog:35669.
Today, we found that one of the terminal resistors had a poor connection and was introducing excess noise. After replacing it, we successfully measured the IMC CMS slow-path dark noise with good agreement between the two spectral measurements (red in Fig. 1).

On the other hand, we observed a mismatch in the dark-noise measurement of the IMC CMS fast path (magenta in Fig. 1), even though it matched well in previous measurements.
We found that the dark noise is sensitive to mechanical stress applied to the CMS board connector.
Since the high-frequency measurement is reproducible and the IMC CMS fast-path dark noise is negligible at low frequencies, we will use the previous result.

The mismatch in the Mokulab dark noise is likely due to a degraded noise floor in the low-frequency band caused by additional downsampling.
Furthermore, we found that the overall gain discrepancy between measured and modeled transfer functions originated from the impedance setting of Mokulab: we should have divided the output by two. With this correction, the measured and modeled gains now match well.

We have completed the necessary measurements for the CARM noise budget and will proceed to the calculation phase.

As discussed in klog35456 and klog35472, high CARM gain might introduce the instability during increasing laser power.
So, 40kHz UGF would be good at least for achieving the stable switching of CARM input gains.

During the work reported in klog35472, I set CARM UGF as 40kHz, so the initially measured UGF seems unexpectedly too high.
I'm not so sure why the CARM UGF (and also IMC UGF) became so high, so it would be nice to investigate the reason.

This afternoon, we found that the lock loss was occured whenever the IN1GAIN value was changed from -23 dB to -22 dB after increasing the laser power itself. This process is performed after increasing the IMC output power. 

At first, we measured the CARM OLTF just before changing the gain values related the CARM loop (K1:IMC-SERVO_{IN1,FAST}GAIN, K1:LSC-REFL_SERVO_SLOW_GAIN). In this state, the OLTF itself should be the same as the state after passing through the INCRASING_LAS_POWER state. Fig. 1 shows the CARM OLTF, dark curve is the one before the adjustment. The UGF at that time is ~60 kHz, and the phase margin seems to be roughly 30 degrees. The gain changing process is performed to keep this OLTF so we considered that the phase margin is not enough to perform the changing process. Therefore, we attempted the process in the state which the overall gain reduced -3 dB by decreasing K1:IMC-SERVO_FASTGAIN, K1:LSC-REFL_SERVO_SLOW_GAIN to -3dB in advance. That is, the process started when IN1GAIN was -24dB, FASTGAIN was 1 dB, and SLOW_GAIN was 0.32

After changing the process, we succeeded in transiting the INCREASING_LAS_POWER state manually. The bright curve is the one after the adjustment. Current CARM UGF is ~40 kHz, phase margin is 43 deg. I wonder that the gain above ~ 150 kHz seems to be not changed even though we adjusted the overall gain. We need more investigation. 

Then, we modified the script in the INCREASING_LAS_POWER of the LSC_LOCK guardian. We tested the process by the guardian at twice, the first trial is failed on the way which FAST_GAIN decreased from 1 dB to  -2 dB. The second trial was succeeded. So we would like to know the succeess rate of this process. We left the IFO with this state.

[Fujimoto, Yokozawa, Ushiba (remote), Tanaka, Komori]

Abstract:

We resolved the lock-loss issue that occurred when increasing the laser power by tuning the IMC loop gain.

Details:

As reported in klog:35668, lock loss consistently occurred when the input power was increased.
Yokozawa-san found that the CARM open-loop gain exhibited an unexpected peak around 200 kHz (Fig. 1), which was a strong candidate for the cause of the lock loss.
Ushiba-san suggested that this could be due to an excessively high IMC open-loop gain.
I measured the IMC loop and confirmed that this was indeed the case: the UGF was around 200 kHz, and the phase margin was small—approximately 20 degrees or less (Fig. 2).

To address this, I reduced the IMC CMS in1gain by 3 dB, lowering the UGF to around 100 kHz and improving the phase margin to ~30 degrees (Fig. 3).
After this adjustment, we were able to increase the laser power without losing lock.

Tanaka-san updated the guardian setting for in1gain in the LOCK PREP state of the IMC guardian.
With this change, the system can now transition through the INCREASING LAS POWER state without lock loss.

I checked the rain collector of the Atotsu weather station. It was fullfield by today's rain.

I and Sawada-san cleand it (removing algae).

We accepted the following SDF change caused by klog35676.

KAGRA Pcal-X updates (2025/11/26)

Workers: Dan Chen, Shingo Hido

We performed monthly Pcal-X calibration on 2025/11/26.

After the calibration, we updated EPICS parameters related to the Pcal-X system. No issues were found.

I analyzed the PEM injection data:

2. Shake injection to TMSX table EXC1 (Fig.3.)
11/25 13:07:00 - 13:41:00 (JST)
EXC : K1:PEM-EXCITATION_EX0_RACK_1_EXC
REF : K1:PEM-POTABLE_EXC_RACK_EX0_ADC0_DSUB25_OUT_DQ
info : TMSX EXC1 50 - 900 Jz 171 point 5 Hz resolution 100cnt excitation

Note that the REF channel name was wrong.

Since a significant excess was not found for almost all frequency injection, the upper limit based on the Coupling Function model (w/o frequency-conversion) is also plotted. 

The channel name in the klog35662 was wrong

I analyzed the PEM injection data:

1. Acoustic injection to PSL room
11/25 12:31:00 - 13:05:00 (JST)
EXC : K1:PEM-EXCITATION_MCF0_RACK_13_EXC
REF : K1:PEM-MIC_PSL_TABLE_PSL1_Z_OUT_DQ
Info : PSL 50 - 900 Hz 171 point 5 Hz resolution 100 cnt excitation

The projected noise was dominant around 400Hz widely.

[Fujimoto, Tanaka, Komori]

Abstract:

We measured the transfer functions and dark noises of the IMC common-mode servo (CMS).
The input-referred noise of the IMC CMS seems reasonable, and we found several issues to be resolved in the measurement.

Details:

As a first step toward performing the full CARM noise budget, we measured the transfer functions and dark noises of the IMC CMS using settings as close as possible to those during observation mode.
The power spectra were taken with Mokulab in two frequency ranges: 10 Hz–1 kHz and 1 kHz–100 kHz.
We first measured the dark noise of Mokulab itself (Fig. 1) and compared it with the dark noise of the IMC CMS.

For the measurement of the dark noise of the IMC CMS slow path in the 1 kHz–100 kHz range, we inserted an SR560 because the noise was otherwise hidden beneath the Mokulab noise floor due to the generic filter with a DC gain of 0.1.
The SR560 was configured with a first-order high-pass filter at 300 Hz and a gain of 100 above that frequency.
The input-referred noise of the SR560 is negligible and is also plotted in Fig. 1.
The measured dark noises of the CMS channels are compared as well.

Ideally, the dark noise should be measured with the three boost filters in the common path enabled to fully reproduce the real operating condition.
However, enabling the boost filters caused circuit saturation, so we instead measured the noise without the boost filters and will compensate for their effect later.

The measured transfer functions of the slow and fast paths are shown in Fig. 2.
The corresponding input equivalent noise spectra for both paths are shown in Fig. 3.

Our next steps will be:

• Resolving the mismatch between the spectra measured in the two different frequency bands.

• Resolving the discrepancy between the measured and modeled transfer functions.

• Performing similar measurements for the CARM CMS.

• Measuring the power dependence of the RFPD noise floor to experimentally determine the shot noise equivalent power.

After today's works, 'INCREASING LAS POWER' state stopped working properly.
The laser power does not increase in this state now, and we do not understand the reason at this moment.
I requested PRFPMI RF LOCKED state.

I found the rain collector data at Atotsu looks strange (too small) since June or July this year.
Other channels of the weather station (e.g., temperature) are not strange.

So I suspect some obstacle material are stacked.

[Kimura and M. Takahashi]

 We inspected all vacuum pump units witout the V-chamber areas.
No abnormalities were confirmed in the vacuum pump units of the central mirror room, Y-arm, or Y-end mirror room.
However, the following abnormalities were confirmed:
1. Mortar spray peeling from the ceiling near 530m on the X-arm
2. Looseness in two jacks on the X-10 ion pump support frame
3. Looseness in GVex anchor bolts
Tightening work was performed for items 2 and 3.
Photos are attached for reference.

The previous photo session was done on 11/14 (klog). So, this shift happened between 11/14 and 11/25.
I heard from Kenta-san, when he performed the initial alighment on 11/21(Fri.), he thought the mirror reference position of ITMY has already shifted. So, I'm not sure this shift is related to the earthquake close to the site (klog).

Date: 2025/11/25

Member: Dan Chen, Shingo Hido

We performed our usual WSK calibration at UToyama.

The results look no problem.

Results

Comparison with Previous Results

Comparing with previous results, no significant issues were found.
Attached graph is the result summary including the latest measured data.