NAOJ GW Elog Logbook 3.2

I measured the beam profile of the laser in the ATC clean booth. Yuhang and Michael are doing another experiment, so Aso-san and I will use the reflected light on the beam splitter(see figure 1).
Results:
width | weist position* | |
x | 0.135 mm | -39.4 mm |
y | 0.136 | 6.07 mm |
* start position is two holes (~50 mm) distant from the center of the BS.
I measured the beam profiles again on March 3rd. I took 10 samples for each point and used an average of them.
width | weist position* | |
x | 0.1347+- 0.0006 mm | -35.6 +-0.5mm |
y | 0.139+-0.004 mm | 11.8+-2.3mm |
* start position is two holes (~50 mm) distant from the center of the BS.

In order to get a more precise comparison between birefringence (integrated along z) and absorption (at a given z position) measurements, we performed absorption measurements at several z positions.
Note that in that case the shinkosha 7 orientation is still the same as the measurement done by Manuel ie arrow at the top and pointing towards the imaging unit.
All results are attached to this entry where I used same colorlimit as Manuel (0 to 200 ppm/cm) and similar colormap.
Similar patterns are visibles.
However, it seems that maximum absorption is quite lower than what was measured before...
One difference with the previous measurements is that we were using 0.5s waiting time and 70mm radius..
I'm now starting new measurements with differents lockin amplifier parameters to investigate this issue.
Note that the z values indicated here correspond directly to the translation stage values (therefore different than Manuel measurements where he corrected the z value to match the real position in the mirror)
We tried to investigate possible explanations for this discrepancy.
First we performed along z scan to be sure that we are able to see the 2 surfaces of the samples.
We could find S1 at 34.8 mm and S2 at 122 mm along z.
We can see the ac/dc signal decreasing with an increase of z (same as Manuel's measurement) but the signal is roughly half of what he got.
We have the same chopper frequency, we're injecting pure s polarization, but differences are that he was injecting about 10 W vs our 8.5 W, he set the DC to about 2.5 V vs 4V now and in his computation he is using 1.16 /cm instead of the 1.04 /cm later measured and I'm not sure how the transmission was taken into account.
For reference Manuel's measurements and analysis are in the KAGRA#7 folder.
One strong possibility is that we have a too large pump beam size. Indeed Manuel found out that it could cause some factor discrepancy when he upgraded the setup.
We characterized the beam size with the razor blade as reported in figure 1. The beam waist is 48.5 um instead of the expected 35 um.
Following Jammt simulation that indicates that the beam waist of 35 um by moving the last lens by ~5mm we started to realign but without clear improvement so we'll continue on Monday.

Abe, Katsuki, Marc
We purchased a Soleil-babinet compensator that will be installed in PCI for future birefringence measurements.
We prepared a test setup on the optical table just in front of PCI clean room (the one where there is the reflectance measurement setup).
We will use the FC spare laser.
We installed 2 OD to have ~ 40 mW of power then 2 steering mirrors(Newport 5204) to have proper alignment above a lign of holes.
We also brought the required componenents (1 PBS, 1 QWP and 4 HWPs).

As reported in elog2852, the CC detuning and CCFC demodulation phase should be adjusted. Since the CC detuning in elog2852 was 76Hz, first I changed CC PLL frequency by ~22Hz, but the shape of CCFC error signal was strange. Then I decided to change CC PLL frequency by 10Hz. The setting of CC PLL is as follows.
channel | function | frequency (MHz) | binary number |
CH0 | CC PLL | 20.99112421 | 1010 10111111 01011001 01000000 |
CH2 | CC1/CCFC demod | 13.99408281 | 111 00101010 00111011 10000000 |
CH3 | CC2 demod | 6.99704140 | 11 10010101 00011101 11000000 |
The setting of LEMO cables for demodulation is as follows.
Connection | Color of LEMO cable |
Between CCFC RF amplifier and mixer RF port | green+yellow |
Between DDS and mixer LO port | green |
Fig. 1 shows CCFC error signal. The CCFC calibration amplitude is 138mVpp. The mode matching is fixed to 0.9.
Today FC was quite stable and I could lock CCFC for the first time since last August! The CCFC filter gain is 1000 with 30Hz LPF. The Z correction, AA, BS pointing were engaged.
Fig. 2 shows the locking accuracy with CCFC.

First I aligned BAB to FC. The optimal p pol PLL frequency without green was 300MHz and BAB power before FC was 435uW. The maximum IR transmission of FC was 470.
Then I checked the CCFC error signal. I optimized the p pol PLL frequency to maximize the CCFC error signal. The optimal p pol PLL frequency with 20mW green was 240MHz and the CCFC error signal was 118mVpp. The setting of CC detuning is reported in elog2521. The setting of LEMO cables for demodulation is as follows.
Connection | Color of LEMO cable |
Between CCFC RF amplifier and mixer RF port | brown |
Between DDS to mixer LO port | green |
The attached figure shows CCFC error signal. The mode matching is fixed to 0.9. We need to adjust the demodulation phase and CC detuning.

I found that CC PLL could not be locked. I changed the phase detector polarity of ADF4002 from negative to positive. Then the fast loop of CC PLL could be locked, but slow loop could not be locked.

Today I replaced the Qubig PD back to the one with DC output (simply called DC-Qubig later), whose change was done about one month ago (elog2801). Note that I just temporarily set up DC-Qubig, whose cables are still not deployed properly since it's hard to do by myself
After putting DC-Qubig back, to have a decent loop gain, I adjusted DDS2 channel CH1 amplitude from 1/4 to 1/2. After that, I took a measurement of open-loop transfer function. The unity gain frequency was around 13kHz. At the same time, the Rampeauto attenuation is 0, the Rampeauto gain is 8.
The setting for AA is shown in Fig.1. The setting for z_corr is shown in Fig.2.
In the first minute of Fig.3, the filter cavity is controlled with PDH and z_corr. After using setting of Fig.1 for AA, the filter cavity transmission is stabilized. This figure shows about five minutes. But longer lock such as about 20min was observed tonight as well.

According to Pierre, the GAIN PIEZO in ACTUATOR module controls also the saturation of PIEZO ELEC signal and should not be small value. Pierre suggested to set this gain as 7 and adjust the servo gain with INPUT ATTENUATOR in SIGNAL module. I set the GAIN PIEZO as 7 and the INPUT ATTENUATOR as 0.2 for 1/f^4 filter. UGF is 13 kHz and phase margin is 60 deg (Fig. 1).
Fig. 2 shows the correction signal (CH1) and error signal (CH2). The correction signal does not saturate even at 2.36V (4.72V for PIEZO ELEC) and FC gets more stable. However, FC still often unlocks even without saturation.

As described in elog2846, FC easily unlocks due to saturation of FC GR correction signal sent to laser PZT.
Fig. 1 shows the FC GR correction signal (CH1) and error signal (CH2) with FC servo gain of 0.8. The FC GR correction signal is half of the signal sent to laser PZT. When the FC GR correction signal reaches +/- 0.44V (signal sent to laser PZT reaches +/- 0.88V), the FC unlocks. This saturation can be relaxed by Z correction as shown in Fig. 2 and we could lock FC for a few minutes with Z correction today.
Fig. 3 shows the same measurement with FC servo gain of 2. The error signal becomes noisy because of too high gain, but the saturation does not occur at +/- 0.44V.
Fig. 4 shows the FC GR OLTF with servo gain of 0.8 and 2. UGF is 14kHz with gain of 0.8 and 23kHz with gain of 2.

I confirmed that BS pointing is working. The setting for center in camera is as follows.
offset | gain | |
pitch | 26 | 50 |
yaw | 14 | 10 |

After the recent recovery of FC alignment, FC GR lock is very unstable. I turned off the END picomotor driver, but it didn't change the situation.
I checked the FC GR correction signal and laser PZT signal. The FC GR correction signal is PZT mon signal amplified by 50 with SR560. For laser PZT signal, I divided the signal sent to laser into two. One is sent to laser and one is for monitor. The relation between FC GR correction signal and PZT signal is as follows.
FC GR correction signal = PZT mon*50 = (1/100*PZT)*50 = 0.5*PZT
Fig. 1 and 2 show the FC GR correction signal (CH1) and PZT signal (CH2). As you can see, CH1 with 500mV range and CH2 with 1V range are completely overlapping, which means the FC GR correction signal is exactly half of the PZT signal as expected.
Fig. 1 shows that when the FC GR correction signal reaches +/- 1.1V (PZT signal reaches +/- 2.2V), FC unlocks (the signal becomes 0). According to a manual of laser, the PZT can accept +/- 65V (PZT tuning coefficient is 1MHz/V and PZT tuning range is +/- 65MHz). So the PZT signal is well within the PZT range.
Fig. 2 shows that FC also unlocks even without saturation.
We will investigate the reason of this unlock.

The attached figure shows the dark noise and shot noise of CCFC error signal. The CCFC error signals were measured on 20210622 and the dark noise was measured on 20220218. Note that the DC port of CCFC PD should be terminated with 50Ohm for lower CCFC dark noise. This dark noise was measured by blocking the light before CCFC PD. This dark noise includes the noises of PD, RF amplifier, and mixer.
Estimation of shot noise
Measured CCSB power at CCFC PD when CCSB are on resonance: 0.8 uW
Typical efficiency of InGaAs photodiode at 1064 nm: 0.45 A/W
According to Yuhang's simulation, transimpedance of PD at 14MHz is 5900 V/A
Shot noise at 14MHz before amplification: sqrt(2*A*q)*R = sqrt(2*0.45*0.8e-6*1.6e-19)*5900 = 2e-9 V/rtHz
Shot noise at 14MHz after amplification: 2e-9*10^(34/20) = 1e-7 V/rtHz
Shot noise of CCFC error signal = amplitude of CCFC error signal*(shot nosie at 14MHz after amplification/amplitude of 14 MHz signal) = (154mVpp/2)*(1e-7 V/rtHz / 0.045 V) = 1.7e-7 V/rtHz
where the amplitude of CCFC error signal and 14MHz signal (-14dBm = 0.045 V) are measured values.
I tried to measure the shot noise directly by injecting BAB to CCFC PD. The injected BAB power was 73uW. However, the shot noise could not be measured since the dark noise was still limiting.

As reported in elog2801, we replaced a PD for FC green lock. The new PD has only RF output, but has better SNR. First I checked the PDH signal of FC green lock with the new PD and found that the demodulation phase was not optimized. I changed the demodulation phase of FC green lock (DDS2 CH1) from 354.99 deg to 270 deg. I saved this DDS2 setting as 20220215_dds2.stp.
The FC could be locked with gain of 2.4 for 1/f filter and 0.8 for 1/f^4 filter. The injection green power was 21.1 mW. Since the FC servo gain becomes half with the new PD compared with old PD reported in elog2770, the new PD seems to have larger gain by a factor of 2.
Then I measured OLTF and error signal (EPS1, 1Vpp fixed range) of FC green lock with 1/f and 1/f^4 filters (Fig. 1,2). Both UGF are around 12kHz. The dark noise with new PD is 3e-5 V/rtHz, while the one with old PD is 2e-5 V/rtHz as shown in elog2770. Since the signal of new PD is 2 times larger, the SNR of new PD should be better than that of old PD.
There are many harmonics of 50Hz. These harmonics are not present in dark noise, so they should come from green.
During this measurement, Z correction with gain of 10 was engaged. BS pointing and AA were not engaged since they are unstable. We should investigate these loops.

I tried to move END YAW with motor A, but it didn't move. Maybe the motor A is also broken.

Today I opened the gate valves between END/arm. The pressure in END and arm before I opened the gate valves was 8.7e-7 mbar and 9.4e-8 mbar, respectively. After I opened the gate valves, the pressure in END and arm became 6.6e-7 mbar and 1e-6 mbar, respectively.
After the END mirror alignment with picomotor, we finally recovered the green flash!

Katsuki, Marc
We installed the new ion gun and used it to clean shinkosha7 surfaces.
We removed the first steering mirror after the lens and realigned the pump beam path.
We got R_surf = 18.47 /W and R_bulk = 0.7647 cm/W with Z_IU = 66 mm and Z_crossing ~42 mm.
We resinstalled shinkosha 7 and used the DC signal to find the mirror center as X = 399.08 mm and Y = 122.175 mm.
We started long z scan (between 25 and 110 mm) to find the 2 surfaces z positions with P_t ~ 6.4 W.
The first surface is clearly visible at z = 34.5 mm but not so much the second one.
In any case because we plan to take several XY absorption measurements for comparison with birefringence measurement we started the first map at z = 50 mm and Pin = 8.61W, Pt = 7.37 W.
This map just finished and we can recognize the usual star pattern.
We will take measurement every few centimeters during this entire week at least.

Aritomi, Yuhang
While setting up the magnet health check scripts for all magnets, I did tests for all mirror magnets. And I found PR magnet H2 seems to be fell down. As attached figure, we can see that PR oplev has no response to H2.
Aritomi-san told me that he didn't check PR magnets in the past, which is why he didn't report about PR magnets issue in the past.
But usually we don't use PR coil-magnet actuators. In addition, Aritomi-san has some urgent test to do. So we decide to check PR magnets not now but a bit later.

I checked the total time required to check all magnets we have for filter cavity, which is 863 seconds (14.4 minutes)
This check is a count of total time required to run the script. This count is done by writing some lines of codes inside the script. We can see this 863 seconds as attached Fig.1.
I also checked the time required for checking one magnet, which gives output as Fig.2 (54 seconds). Considering we have 16 magnets to check, the total measurement should be 54*16 = 864 ~ 863 which is consistent with the total time measurement.
Matteo suggests to do such check not only after earthquake, but maybe every month.

[Aritomi, Yuhang]
Before the evacuation of END chamber, we checked the END suspension. We excited each coils of end mirror and measured transfer function from the excitation to pitch/yaw oplev spectrum (Fig. 1-4). All coils seem to be working. We also checked END oplev spectra (Fig. 5, 6). The reference in Fig. 5 is one we use for long time, while the reference in Fig. 6 is one measured on 20220128 before we fixed the END magnets. Fig. 6 shows the END oplev spectra become smaller after we fixed the magnets, but Fig. 5 shows that the END oplev spectra are different from the nominal one and still there is a 6.4Hz peak in pitch spectrum.
Anyway, we brought the movable rotary pump from center area to end and started the evacuation. After the pressure in END reaches below 0.1mbar, we switched to the turbo pump. We will open the gate valves between arm/END tomorrow.

The other magnet (left) was released. The TM was released and the chamber was closed.