NAOJ GW Elog Logbook 3.2
Today I moved the clean booth that was in the storage room on the test setup for the Soleil-Babinet.
Indeed we want to install it later on the PCI setup and want to limit possible dust contamination.
All the optics that were below (seems to be TAMA mirrors) and screws, magnets, etc... are now inside the clean booth of the microscope.
The setup is now ready for calibration of the Soleil-Babinet and we plan to use pure s-polarization for this.
Marc Michael
Today we brought 2 SR560 to input oplev (one from the OPO test setup, one from the elec shop).
We reinstalled the shift PSD and a steering mirror on the rail and centered the PSD.
Note that the tilt PSD and its steering mirror are really close to the rail so the newly installed one have barely enough room to move.
Marc, Matteo, Michael
One possible reason for our very low coherence could be that we injected uniform noise with 4000 counts.
To improve both the driving and sensing it is actually better to inject noise through swept sine.
We did so with noise amplitude of 10000 counts and found that the length mechanical resonance at 0.9 Hz was strongly excited in every dof.
Either the noise is too strong, the PSD is at a wrong position or the driving matrix needs to be tweaked.
We are currently taking again this measurement with lower noise amplitude at 4000 counts.
Marc, Michael
We checked the END oplev SR560 and found out that the length oplev one was saturating.
We decreased the gain of all down to 1 and realigned all signals. After that we increased the gain back to 100.
For reference, the tilt PSD sum is about 4500 counts and shift PSD sum is about 3000 counts.
Marc, Michael
Because we are missing some SR560 now to check the input oplev we decided to check the end ones.
We recentered the beam on the PSD and confirmed the good condition by measuring the usual spectrum of the END oplev for pitch and yaw.
We tried to measure the TF between END mirror yaw excitation end the 3 dofs of the oplev.
However, we did not manage to have coherence about 0.1 between yaw excitation and yaw signal from the oplev.
We checked the coils and they are working fine.
We injected white noise from about 500 to 4000 counts amplitude but it did not change the coherence value.
We will try to check with excitation over longer duration.
We also tried to realign the length PSD but could not find proper values from MEDM.
We will try again this alignment removing the SR560 to avoid too large gain of the signal.
This entry reports absorption measurement of AZTEC #3 with R = 0.5793 cm/W and about 7.7 W of incident pump power.
Its absorption is in agreement with KAGRA requirements.
Recently, the C disk is almost full due to large size of windows folder.
This paused several times on-going measurements.
We will try to update windows (maybe to version 11), to see if it solves the issue.
In the meantime, I changed the default folder where the map data are saved to D:\PCI_measurements.
Marc, Michael, Yuhang (remote)
It was reported in elog 2828 that input oplev spectra are similar for pitch and yaw.
First we wanted to realign input oplev but found out that diaggui was timed out.
This can be fixed by restarting the standalone pc.
Before that, we need to take a snapshot of the medm configuration but the disk was almost full (99%). In the past, this prevented proper snapshot (eg in elog 2881) so we deleted the data from the disk beforehand.
We followed procedure from elog 1979 where the delete command is rm -r 13*** to delete all files starting with 13. We removed all trend/second and full data of May that freed 7% of storage.
After that we could investigate the input oplev behavior.
We found out that there is now only one SR560 (the other one is in ATC cleanroom for the new OPO test setup).
In that case, only pitch is increased by a factor of 100 and we guess that the yaw signal is somehow normalized by the sum (effectively pitch) making the 2 signals identical.
When removing the SR560 connection, we could see the expected mechanical resonances of pitch and yaw.
Note that now x is yaw and y is pitch.
Because we will not replace the OPO so soon, we plan to take back the SR560 from ATC to the input oplev.
Marc and Michael
The frontend PC was reinstalled with Debian 11 - we are going to go with the latest software and hardware that will be used in O5 rather than current versions. The frontend was altered with the following manual partition setup (figure 1):
/boot - default size
/ (root file system) - 50 GB (maybe some large things will be installed here...)
/swap - 32 GB - same as RAM size
/tmp - 5 GB
/var - 10 GB
/home - remaining space
Also:
Language - English
Region - Japan
Locales - en_US.UTF-8
Keyboard - American English
hostname - FEcentral
domain name - mtk.nao.ac.jp (default)
Network interface - eno1 Intel Corporation ethernet controller 10G X550T
root password - starts with a 1 instead of 0
ops user 1000 (same password as usual)
controls user 1001 (same password as usual)
package manager - deb.debian.org
HTTP proxy - blank
From the ALIGO respository, we are now using the debian bullseye 1.0.9 cdssoft instead of buster 1.0.6 as in the last entry on this topic.
For the data concentrator Debian 11 installation, I am still not so sure about the partitioning. I still haven't seen any guidelines on it. The preinstalled storage hardware is shown in figure 2, from the BIOS settings. Also, we are using the minimal Debian installer and connect to the internet during the installation. But, the Myrinet controller needs some other firmware to run (figure 3). I remember that LIGO has some link to it but can't put it on their repository. In this case we have two ethernet controllers to choose from (figure 4). But, the first one on the list also didn't work. I skipped that and went to the partitioning. It already threw out the preset partitioning with some filesystem that I don't know of (figure 5), and I'm not entirely sure how to do this one either. So I just stopped there and didn't write anything to the disk, i.e. it still boots CentOS 7
I changed the root password back to normal. For me it was just Linux paranoia to make the root password different but I guess it doesn't matter here.
Marc Matteo
We removed FC from the AZTEC #3.
Note that the imaging unit black cover was modified so that we don't need anymore to disconnect the cables.
We checked the bulk reference sample and measured R = 0.5794 cm/W with Pin = 27.67 mW and Pt = 15.06 mW.
It is still consistent with previous estimation so we installed the AZTEC #3 on the translation stage.
We did a long z scan and estimated the 2 surfaces at z = 26.05 and 114.5 mm so that z_center = 70.275 mm
We started XY measurement at Z_center.
From the long z scan, absorption seems promising.
Marc, Matteo
Today we removed AZTEC #2 and installed the bulk reference sample.
We measured R = 0.5780 cm/W which is compatible with previous measurement within 5%.
This is reasonable so AZTEC #2 measurements are fine.
We used again first contact on the AZTEC #3 as there was some remained of the previously applied first contact and peeled.
Note for future to never use the sticky tape given by first contact.
We also applied first contact on the second surface of the 1.5 inch sapphire used for calibration.
We cutted holes in the black box covering the imaging unit in order to avoid to have to remove the cables every time we want to intall/remove KAGRA size substrates.
We started measurement of AZTEC #1 at exactly the same position as before (XY map at Z center).
Measurement is on-going but preliminary results seems compatible with previous measurements meaning that they should be fine.
Report of AZTEC #2 absorption measurements.
For reference the arrow is on top pointing towards the imaging unit.
Absorption is larger than AZTEC #1 and we can also quite clearly see the growing seed on the xy2 map (fig 2 that is measured at the z center of the substrate)
Two photos (under gluing and after gluing) are attached.
[Takahashi, Aritomi, Marc]
We opened PR chamber and fixed one falling magnet of PR mirror and released the suspension. During this work, we noticed the oplev laser was hitting off center of PR mirror. We aligned the oplev laser to make it center of PR mirror. Currently the oplev beam is not reaching the PD and we need to align it.
We will open BS chamber on Friday next week (5/27).
Two photos (under gluing and after gluing) are attached.
Date: 2022/5/13
With Homare Abe and Takayuki Tomaru
At KAGRA
-
We took pictures of 2 sapphire samples with a green injection laser beam:
- One of them showed a green scattering light
- The other one showed a red scattering light
-
We also took pictures of ITMY mirror with the same green injection laser beam:
- The results showed a green scattering light
Analysis will be performed.
Pictures: klog
Aritomi, Marc, Michael
We installed the new picomotors drivers. They worked fine for PR picomotor but did not work for BS picomotor.
We need to open BS chamber to check the picomotors.
With PR picomotors, we recovered the old reference in BS chamber (eg in elog 2794). In that configuration (and maybe BS in a 'random' position, the green beam is hitting quite below the gate valve window between BS/input but at least we don't have anymore clipping).
However, the PR oplev laser is now hitting the edge of the steering mirror before the PSD. It means that the PR reference at BS chamber that we are using might not be valid anymore for the current alignment inside the PR chamber.
We realigned PR and END oplevs and ran the coil health check codes.
All magnets of PR, INPUT and END mirrors are fine except PR H2 and we can not assess BS situation as the oplev PSD is broken as reported in elog2775.
We found a spare PSD inside the cleanroom and tested its dark noise (see attached figure where red and blue curves are the dark noise).
It is fine so we plan to replace it after the recovery of BS picomotors.
Since the beam spot at GV between BS/input depends on both PR and BS, we should align PR and BS at the same time by using the GV between BS/input and the first target as references.
[Aritomi, Michael]
We found that one of the green references at PR chamber was not aligned as shown in Fig. 1, which means we need to align the green injection beam. We removed the cover of the optical bench and aligned the green injection beam as shown in Fig. 2.
Then we aligned PR to make the green beam at the center of GV between BS/input. However, in this situation, the green beam at PR reference in BS chamber is clipped by a mirror behind the BS as shown in Fig. 3,4.
Since the beam spot at GV between BS/input depends on both PR and BS, we should align PR and BS at the same time by using the GV between BS/input and the first target as references.
Marc, Matteo
Following the recovery of the PCI, we installed the AZTEC #2 sample.
First we measured X_center = 400.735 mm, Y_center = 122.105 mm.
We did a long z scan and got the 2 surfaces at z = 26.28 mm and 119 mm meaning that Z_center = 72.64 mm.
We have incident power = 7.69W and transmitted power = 6.636W.
We started a XY measurement at Z_center and hope to finish all 5 maps by monday.
Marc Matteo
This entry summarizes these past days activities.
We found a mistake in the fitting code (the wavelength was hard coded to 633 nm instead of 1064 nm making the previous estimation wrong by a factor sqrt(1064/633))
To avoid this issue, we coded a more flexible function 'fit_blade.m' that is saved in the PCI scripts folder on the desktop
With this function, we could finally tuned the pump beam telescope and recover Manuel's parameters (fig 1)
We also check the vertical and horizontal angle of incidence to be -0.2 deg and 2.5 deg respectively.
We also checked the probe beam parameters by placing a power meter in front of the absorption PD.
We also got same parameters as Manuel (fig 2)
The angle of incidence is 3.47 deg meaning that the relative angle between the probe and pump beam is also correct.
We installed the surface reference sample and tried to maximize the R coefficient by changing both the translation stage and imaging unit z positions.
While doing these motions, we could find the expected maximum at x = 35 mm and z_iu = 68 mm.
However, R = 14 /W meaning that we still had not optimal conditions.
After investigating other possible issues (chopper frequency, laser power, pd, lock-in are all fine) and repeating several times this calibration, we could not improved this value by much...
We installed the bulk reference sample and got R = 0.5536 cm/W.
Both surface and bulk calibration are 10 % lower than expected but the reason is still not clear to us...
We decided to use the 1.5 inches sapphire that Manuel used to check the setup calibration (eg check entry 1132).
We did the exact same measurement and got similar values (fig 3) : mean absorption ~ 40 ppm/cm.
However, the sample was really dirty so it we cleaned this sample by wiping alcohol on it and repeated the measurement that gave identical result (fig 4)
It means that despite some issue with the setup, our calibration gives us reasonable results and can be considered to be working again.
Marc, Michael
We checked the END oplev SR560 and found out that the length oplev one was saturating.
We decreased the gain of all down to 1 and realigned all signals. After that we increased the gain back to 100.
For reference, the tilt PSD sum is about 4500 counts and shift PSD sum is about 3000 counts.
Marc, Matteo, Michael
One possible reason for our very low coherence could be that we injected uniform noise with 4000 counts.
To improve both the driving and sensing it is actually better to inject noise through swept sine.
We did so with noise amplitude of 10000 counts and found that the length mechanical resonance at 0.9 Hz was strongly excited in every dof.
Either the noise is too strong, the PSD is at a wrong position or the driving matrix needs to be tweaked.
We are currently taking again this measurement with lower noise amplitude at 4000 counts.