Coil driver output voltage noise might be comparable with the shot noise of the effective current applied to the coils.
If so, further actuator efficiency reduction, except for ETMX, might not be effective to reduce actuator noise and reducion of RMS of the feedback signals would reduce it..
I evaluated the effect of shot noise from DC current applied to actuator coils.
According to fig1, RMS pf TM actuator feedback signals during OBSERVATION state are roughly 6000 cnts for ETMX, 2000 cnts for ITMY, 1500 cnts for ETMY, and 2500 cnts for ITMX.
If we considered effective current as 1/sqrt(2) of RMS, effective current for each suspensions can be calculated as follows:
ETMX: 6000 [cnt] * 310e-6 [V/cnt] / sqrt(2) / 150 [Ohm] = 8.8 mA
ITMY: 2000 [cnt] * 310e-6 [V/cnt] / sqrt(2) / 300 [Ohm] * 2 = 2.9 mA (last 2 is a compensation gain, which was engaged after reduction of actuator efficiency to keep the actuator gain from COILOUTF input)
ETMY: 1500 [cnt] * 310e-6 [V/cnt] / sqrt(2) / 300 [Ohm] * 2 = 2.2 mA
ITMX: 2500 [cnt] * 310e-6 [V/cnt] / sqrt(2) / 300 [Ohm] * 2 = 3.7 mA
So, shot noises of the current applied to TM coils are 5.3e-11, 3.0e-11, 2.7e-11, and 3.4e-11 A/rtHz for ETMX, ITMY, ETMY, and ITMX, respectively.
Here, I used sqrt(2*e*I), which is the relation between shot noise and DC current, where e is an elementary charge and I is effective current applied to coils.
If we considered 1e-8 V/rtHz of HPCD output voltage noise, current noises applied to the coil are 6.7e-11 A/rtHz for ETMX and 3.3e-11 A/rtHz for the others, which is close to the estimated shot noise of the effective current.
If the above consideration is true and voltage noise and current noise is independent (summation of the noise can be calculated by square root of square sum), actuator efficiency reduction by inserting resistance reduces actuator noise by a factor of 1.3, which is slightly lower than we expected (1.5) by only considering coil driver output noise.
Also, current noise cannot be reduced by reducing actuator efficiency, so further actuator efficiency reduction, except for ETMX, might not be effective to reduce the actuator noise, and reduction of RMS of feedback signals might reduce the actuator noise.
Since I'm not so sure using effective value of the current is suitable for calculating the shot noise, my discussion would be wrong.
[Oshino-san, YamaT-san, Ikeda]
Summary
Several problems involving IO chassis have been reported. (K-Log#33348,#16759)
The old IO chassis's use HIB cards, but there are no replacement cards in stock.
Therefore, replacement with a new IO chassis is mandatory for the future.
The DGS has confirmed that a configuration equivalent to K1TEST0 will run for a long time on the test bench.
First, we will replace K1TEST0 with the new IO chassis to investigate the impact on KAGRA.
Details
We prepared to replace the IO chassis of K1TEST0 with a new IO chassis(S2416128).
The DC cables we have on hand have different terminal configurations, so we returned to the old IO chassis.
The placement within the rack is tentative.
K1TEST0 and the old I/O chassis have been moved slightly upward within the rack.
The BIOS of K1TEST0(V3) was Version 3.1.
The test bench results confirm that 3.1 recognizes four Adnaco cards.
With 1.0, only one card was recognized.
We plan to replace it with a new IO chassis next week.
[Guo Chin, yuzu]
Ushiba-san pointed out the hypothesis that the recent interferometer is sensitive to seismic motion, that weak seismic motion can lead to oscillation of the IMC length, and that frequent lockloss with IMCL label occurs. We want to test this hypothesis by comparing the lockloss data between 10 W operation and 1.2 W operation.
We focused on the lockloss from the OBSERVATION_WITHOUT_LINES (called silent run) and lockloss with the IMCL labels, because it's easy to find the lockloss related to the excess of the seismic motion (3~10 Hz).
The data in two kinds of periods were checked.
During 10 W operation, there were 21 lockloss with the IMCL label. 19 lockloss coincided with an excess of the seismic motion of 1~10 Hz bands.
During 1.2 W operation, there were 16 lockloss with the IMCL label. 6 lockloss coincided with an excess of the seismic motion of 1~10 Hz bands.
We collected the seismic motion data of the 3~10 Hz band around the lockloss. Actually, we checked both frequency bands of 1~3 Hz and 3~10 Hz. The excess of the 3~10 Hz frequency band looks to be the dominant source of the oscillation of the IMC length.
In some cases, we observed the two-step peaks due to the p-wave and s-wave (for example, the lockloss of 2024/12/28 19:36:42 UTC), and we recorded the peak value closer to the lockloss.
Figures 1~5 show the time series around the lockloss and peak value of the seismic motion.
Figure 6 shows the histogram of the peak amplitudes for two periods.
In the case of 10 W operation, even though the seismic motion is small, such as < 0.5 um/s, the oscillation of IMC length frequently occurred, and the lockloss occurred.
In the case of 1.2 W operation, the IMC length oscillation didn't happen frequently. It makes a smaller number of lockloss with the IMCL label (6 times).
Next week, I will plan to check the seismic motion of 3~10 Hz during lock and collect the excess of 3~10 Hz data that didn't make the lockloss. We can test the hypothesis if we collect many examples where the lock continued even though there was an excess of 3-10 Hz seismic motion, in 10 W operation.
And also, I will check the frequency of occurrence of the excess 3~10 Hz at 2024/12, 2025/02, and 2025/04. It's possible that the seismic motion of 3~10 Hz occurred frequently in April.
Ikeda, Tanaka
As reported in klog32957, the relative phase between each coil of the beacon excitaion seems not to be in phase maybe due to the steep phase rotation in 24 kHz modified AI filter. So we implemented phase rotator for each coil of beacon excitation to be able to compensate relative phase difference due to the phase rotation by 24 kHz modified AI filter.
Fig. 1 shows the new medm screen of beacon excitation. We add the phase rotator function between EUR2COIL and COILOUTF. Now, we use the SIN output as the beacon excitaion signal instead of CLK. So if you want to change the excitation amplitude, you input the value not in CLKGAIN but in SINGAIN. Fig.2 shows the screen of the rotator. Then, if you input the same value in COSGAIN as the one in SINGAIN, the rotator shifts the phase of the beacon signal applied the coils to the value in K1:SEN-{TypeA TM name}_H{1,2,3,4}_PHASE_R.
I tested the function. The left lower panel in Fig. 3 shows the phase transfer function at 23605.5 Hz between H1 and others and the phase configuration. As you can see, the rotator shifted the phase to the value which we set.
I took the TCam photos for four mirrors at 9:10 ~ 9:15 this morning.
Update model files
Request from Ushiba-san
[K1BS, K1SR2, K1SR3, K1SRM]
model: k1visbst, k1vissr2t, k1vissr3t, k1vissrmt
Modifications
BS and SR change all GAS and IP BIO connections to the same connection.
GAS: H32_C2
IP: L32_C2
TOWER_MASTER/TPPEB_TOWER_MASTER
Modify input to COILOUT_CTRL,MASK of F0 from IP_COILOUTF_CTRL,MASK to GAS_COILOUTF_CTRL,MASK.
Request from K.Tanaka-san
[K1EX1, K1EY1, K1IX1, K1IY1]
model: k1sendbeacon, k1sendbeaconey, k1sendbeaconix, k1sendbeaconiy
Modifications
BEACON_LIB/BEACON_SENDER
Input to LOMON was changed from CLK of OSC to SIN.
Added PHASE as an input to COS of OSC before input to COILOUTF.
k1sendbeacon
Replaced with BEACON_SENDER.
[Kimura and Yasui]
A lack of cooling water was found in the chiller unit for TMP in Y-27, so the cooling water in the other chiller units was checked for water quantity.
As a result, a cooling water shortage was also found in the chiller unit for TMP of X-10.
Therefore, approximately 3.5 L of water was added to the water tank inside the chiller unit.
(See attached photos)
up + down
[Kimura and Yasui]
We installed an interlock circuit in the #-39 pumping unit at X-end on Apr. 24..
After installation, we confirmed that the interlock circuit works in the event of a power failure and that the shutoff valve between the dry pump and the TMP is closed.
After this operational test, the gate valve between the pumping unit and the duct was opened and the TMP of the #-39 pumping unit started pumping.
We also installed blue and red double-sided magnetic sheets to indicate operating status of vacuum pumps.
The blue side of the magnetic sheet indicates " operation", while the red side indicates " shutdown".
up + down
> My guess to explain this result
A day later, I read this description again. But, I don't understand it. Please ignore this explanation.
I tried to recover OMC ASC with beacon. I scanned ETMX and found the current drumhead frequency is 23605.5 Hz. Also I performed phasing for 17 MHz demodulation with this frequency and phasing for 24 kHz demodulation with OMMT2 PIT excitation at 0.5 Hz. Then I decoupled OMMT2 and OSTM dofs. But I will try to close the loops yet because today is already late.
Details will be reported later.
[Tanaka, Ushiba]
Current OMC length fluctuation might be limited by sensing noise feedback through OMC PZTs.
So, it would be better to optimize OMC LSC to reduce the residual OMC length motions.
Since OMC LSC residual motion was one of the large coupling path to the OMC transmission (klog30657 and klog30613), we evaluated current OMC length residual motion.
To evaluate the length motion, we used following two ways:
1. Length fluctuation by vibration of optical table
2. Length fluctuation by sensing noise through OMC LSC.
To evaluate the length fluctuation by #1 path, we measured transfer function from OMC geophone to OMC length fluctuation by white noise shaker injection.
We measured the transfer function of each 10Hz bands and combined it to avoid saturation.
OMC length was calibrated into displacement in the unit of meter by comering the condition with old optical efficiency measureent (klog30535).
Calibration factor of 5.1e9 cnt/m was obtained 300-cnt dither amplitude with 29mW OMC trans.
During the measurement, we used 30000-cnt dither amplitude with 30.5mW OMC trans.
Since the actuator efficiency of dither PZT was reduced by a factor of 10 due to the change of PZT driver for OMC DC PD protection (klog32549), calibration factor during the measurement is 5.1e9 *(30000/300)/10 * 30.5/29 = 5.4e10 cnt/m, which is 1.9e-11 m/cnt (note that current nominal OMC lock was done by 500-cnt dither amplitude, so calibration factor during the OBSERVATION state should be 1.1e-9 m/cnt).
Figure 1 shows the measured transfer function from optical table displacement measured by in-vac geophone to OMC error signals.
Bottom left and middle right show the TF from OMC geophone signals (calibrated into the unit of meter) to OMC error signals (also calibrated into the unit of meter) at 130-200 and 50-130 Hz, respectively.
Top right and bottom right shows the coherence between OMC error signals and geophone signals during the measurement.
By using these TF and geophone spectrum during the OBSERVATION state, we can estimate the OMC length fluctuation due to OMC stack vibration as shown in fig2 (blue).
As discribed in the above, calibration factor of OMC error signals during the observation is 1.1e-9 m/cnt, so the current sensing noise of OMC error signals in the unit of meter can be plotted as shown in fig3.
Current UGF of the OMC LSC loop is about 9Hz (klog31780), so I made a OLTF model according to the filters implemented in K1:OMC-LSC_FB_FLT filter bank and calculated the OMC length fluctuation due to feedback of the sensing noise.
Figure 4 shows the result with OMC length fluctuation estimated from OMC geophone spectrum.
According to this calculation, current OMC length fluctuation seems to be limited by OMC LSC feedback noise upto 120Hz except for several peaks.
So, it might be worth trying to modify the OMC LSC loop to mitigate the OMC length fluctuation around 100Hz.
Currently, CARM noise coupling somehow increased and noise around 5kHz is larger than the best sensitivity.
So, reducing CARM coupling and improving the sensitivity around 5kHz also contributes to reduce the OMC length fluctuation through sensing noise reduction.
[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.
[Guo Chin, Yuzu]
We compared the number of lockloss with the IMCL label between 1.2 W and 10 W operation. The frequency of occurrence was 4.5 times per day during 1.2 W and 1.3 times per day during 10 W operation, respectively. This number is smaller than the number in the case of 1.2 W operation (last Feb.).
Note that we collected the lockloss only from OBSERVATION_WITHOUT_LINES (called silent run).
We will summarize the comparison ot the ampitude to trigger the lockloss of the IMC oscillation,
> My guess to explain this result
A day later, I read this description again. But, I don't understand it. Please ignore this explanation.
I also evaluated the projection from the Leg ACC to the OMC error (length)
Scanning plots for each frequency injection are here
https://gwdoc.icrr.u-tokyo.ac.jp/cgi-bin/private/DocDB/ShowDocument?docid=16649
scanning plots (Leg ACC -> Strain) for each frequency injection
https://gwdoc.icrr.u-tokyo.ac.jp/cgi-bin/private/DocDB/ShowDocument?docid=16648
I analyzed the same analysis using SEIS_Z or ACC_Z and compared the results for both 2025-04-20 and 2024-11-09 data.
For both days, the results of using SEIS_Z and ACC_Z are consistent.
By the way, even before the shroud installation (2024-11-09), the linear-like coupling and its siderobe were found.
On 2024-10-16 (before the stack connection), a ground shaker injection was performed (klog31337), but almost no significant excess was found in the DARM. (due to the shaking amplitude not being enough?)
The reason of SR3 oscillation seems F1 GAS control.
Fgure 1 shows the F1 LVDT signals with nominal gain (blue), -3dB (green), and -6dB from the nominal (red).
Since -3dB and -6dB seems no significant difference, I engaged -3dB at FM7 of F1_DAMP_GAS filter bank.
Then, we can engage 2-stage dewhitening filters for SR3 tower as well as the other Type-B suspensions.