NAOJ GW Elog Logbook 3.2

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

Ushiba-san(remote), Yuhang
With the help of Ushiba-san, I set up an automatic suspension health check script for TAMA filter cavity.
This work is motivated by the recent discovery of mirror magnets falling down issue. We suspected the coils for BS/input/end mirrors reported in elog2712, 2800, 2812. The earthquake happened on 2021/10/07. Afterwards, we checked oplev spectrum, which helps us to check if there are touching problem. However, we haven't checked whether we have coil falling down. The strange behavior of mirror driving was found much later. Therefore, we conclude that we should check both oplev spectrum and transfer function.
Checking oplev spectrum takes about two minutes. But checking the response of each coil would take one hour, which requires opening diaggui many times and clicking buttons. So we would need to have an automatic measurement of these TFs.
The place of the suspension check shell code is at /home/controls/Desktop/TAMA_VIS/check_after_earthquake. The transfer function is in this directory as well. To confirm if the code I copied works well, I took a measurement of transfer function beforehand as Fig.1. This test is done for PR coil H1. A reference is saved with H1 in function. And a measurement with H1 input signal blocked is in Fig.1 as well, which shows noisy and smaller TF and almost no coherence.
Before taking an automatic measurement, I unblocked signal to H1. Then I cd to the directory where we have the shell code (cd /home/controls/Desktop/TAMA_VIS/check_after_earthquake). Next step is to run ./test.sh . Finally, we can get output at command window as Fig.2. Opening diaggui and checking PR coil H1 TF show a measurement result as Fig.3. We can see it is different from Fig.1, which means that the automatic measurement was really done with H1 signal unblocked. This confirmed that the code I copied is working well.
Next step: create .xml file for other coils and modify shell code accordingly.

Katsuki, Marc
For the measurement with 30 deg input polarization angle we swapped the scan and shift directions (ie in the attached figures 2 nearby points in Y are close in time).
We can now see the stripes in the horizontal direction.
This confirms that the power fluctuations are not correlated to the mirror position.

Yuhang and Michael
We attempted to obtain a proper ringdown signal of the cavity by switching off the 1st order diffraction passing through the AOM. We used the following methods:
1) Turn off the RF signal from Anritsu function generator to AOM
2) Amplitude modulate the RF signal from Anritsu by using Mokulab - a square wave of frequency 1 Hz was input to the Anritsu external modulation port such that the Anritsu signal would oscillate between normal and zero, with an "instant" change (quotation marks).
3) Frequency modulate Anritsu signal, which would then make the AOM efficiency effectively zero. But the Anritsu only allows 1 MHz/V modulation, so it wasn't enough.
These methods produced ring down signals that were nowhere as fast as what was expected - we measured microseconds but are expecting nanoseconds ringdown. So even though the AOM can be turned off more quickly than what we could get by mechanically blocking the beam, we are still limited by the deactivation speed of electronic switches.

One magnet (upper) was glued with the jig and released 3 hours later. The other magnet (left) is under gluing. It will be released on the 14th.

As reported in elog2831, we adjusted AOM incident laser relative angle to AOM and achieved power scattered from 0-order to 1-order of only 6%. We suspected the alignment maybe the main issue since usually 20dBm RF power should be enough. To confirm this, I took the RF amplifier from TAMA ZHL-2, which has bandwidth of 1GHz and can accept incident RF power smaller than 15dBm. I used this RF amplifier to provide even larger RF signal to AOM and check whether we can have larger efficiency.
The ZHL-2 has a gain of 18dB for signal around 100MHz. So I used Anritz to provide 2dBm at the beginning, which passed through ZHL-2 and arrived at AOM. Using power meter (with 55mW incident on AOM), I measured AOM efficiency as a function of AOM applied RF power. The result is shown in Fig.1.
We can see that higher AOM efficiency was achieved by applying higher RF power. In addition, the increase of efficiency with RF power is still quite linear and not reaching flat region. Therefore, it should be confirmed that alignment is not the reason of low AOM efficiency. But a higher AOM efficiency is prevented by not being able to provide a higher RF signal due to device limitation.

Michael and Yuhang
In the past two days, we have been setting up AOM for switching off OPO injection beam fastly. A 'AA.MT110 IR 27' was used to provide the first Brag diffraction. According to specification, about 1 degree (lambda*f/2/v) Brag angle is required. Now, a newport pure rotation stage is used to adjust AOM. Using the current rotation stage, we could achieve efficiency of only 6%. Of course, the RF signal power could provide some limitation considering we can provide up to 20dBm now.
But anyway, we would need to replace it with newport alignment stage 9071 to better align AOM. We would need a high-power RF amplifier for providing higher RF signal level. Since bandwidth of 110MHz is not small, the bandwidth requirement needs to pay attention as well.

As reported in entry 2817 we can see vertical stripes in our measurements.
One question is are they correlated to the Y (horizontal) position of the mirror.
For reference, the measurement are done in zig-zag ie we scan in X (vertical) and shift in Y.
I show in the attached figures the sum of the normalized s and p polarizations for various input polarization angles as a function of time.
The color changes correspond to a change in Y position.
To my eye it seems that these figures show that there is no clear correlations between the Y positions of the mirror and the power fluctuations.
In any case I will try to do further investigations next week.

Katsuki, Marc
Recently we found out that there are large stripes visibles in the birefringence measurements.
Furthermore, if we check either the raw or normalized s and p polarizations power, their sum is not at all constant (see figure 1).
We installed a power meter at the beginning of the imaging unit and measured the pump power without sample (fig2) with sample (fig 3) and with sample moving as a birefringence measurement (fig 4).
The power is not stable at all...
While we could see something like +/- 1 % fluctuations during the polarization calibration, we can see that the long term peak to peak fluctuations seems to be around +/- 10 % during 30min and +/- 20 % during a birefringence measurement.
This explains why we see the stripes and why the sum of s and p polarizations power is not constant during a birefringence measurement.
Note that this large power fluctuations does not affect drastically the birefringence measurements as we don't see the stripes in them.
Furthermore, it seems that these power fluctuations are far more important at low power compared to high power (few watts for absorption measurements) because we can not see the stripes in these measurements.
Another by-product of this measurement is that we could estimate the background level of each PSD to be about 1uV.

After the evacuation of input chamber, I aligned the input oplev by tweaking the steering mirror just after the oplev laser source.
Then I measured the input oplev spectra as shown in the attached figure. The input pitch and yaw oplev spectra look very similar now.
I checked the oplev beam height at the injection, readout viewports. The oplev beam height at the injection, readout viewports were 110mm, 106mm, respectively. I tweaked the injection steering mirror and made the both beam height 110mm, but the pitch and yaw oplev spectra still look very similar.

Before I opened the gate valves between input/BS and input/arm, the pressures in input, BS, arm was 1.3e-6 mbar, 9.9e-10 mbar, 3.1e-8 mbar, respectively.
First I opened the gate valve between input/BS. After 1 hour, the pressures in input and BS chambers became 5.5e-7 mbar and 1.9e-8 mbar, respectively.
Then I opened the small/large gate valves between input/arm. The pressures in input, arm became 5.4e-7 mbar,4.5e-7 mbar, respectively.

Michael and Yuhang
18.2% power is not coupled.

Michael and Yuhang
We checked the mode matching condition between BAB and OPO. 13.2% power is not coupled.

I started the evacuation of input chamber. First I used rotary pump until the pressure in the input chamber reaches below 0.1mbar. After I removed the rotary pump, I closed the small gate valve close to arm and opened the small gate valve close to the input chamber. I will wait until the pressure in the input chamber reaches the similar one in the BS chamber.

Yuhang and Michael
We found an issue with the Mokulab unit used in the ATC cleanroom
We used 100 Hz 10 dBm 1Vpp signal from another function generator. Putting the signal into Mokulab IN2 and using the oscilloscope function shows no frequency signal and 4 mV offset. IN1 reads 9Vpp (+5.4, -3.6) at 118 MHz even with no connection.
Using 1 MHz signal into the spectrum analyzer function likewise shows no signal, just -131 dBm floor. Using a T connector from the function generator to oscilloscope and Mokulab, we see that the signal on the oscilloscope is reduced when switching from Moku IN2 to IN1, but no change when switching IN1 input impedance between 50 Ohm and 1 MOhm.
Mokulab function generator output works fine though, we have been using it to scan the cavity.

I have now taken 3 measurements with s polarization at the input that saturated the lockin amplifier..
It seems that we can really easily saturates the p polarization.
Indeed, it is connected to the old lockin amplifier where the range (0 to 1 V) is changed depending on the sensitivity setting we are using.
Now I'm injecting about 160 uW of power but when doing the measurement with s polarization it seems that the p polarization power changes by more than a factor 14...
I'm starting hopefully the last measurement with s polarization at the input where the sensitivity of the old lockin is set to 100 mV despite the value at the center being 0.2 mV.
It would be convenient to have a new lockin amplifier to avoid this issue...

For easier comparison with direct measurement with PCI (see elog 2755), I show here the absolute value of delta n flipped both horizontally and vertically .
It seems that the larger delta n area have somehow a close triangular shape.

I had a look at the normalized Is and Ip data.
As seen in figures 1 and 2 which show respectively the s and p polarizations normalized intensities the stripes are mainly present in s polarization.
This was not the case for the previous measurements (see figures 3 and 4) were stripes were actually visibles in both s and p polarizations...
Also it seems that p polarization is saturating..

Recently I've been trying to compute birefringence from TWE measurements of spare ETMY based on Aso-san's computation.
I'm using the S1thruS2 measurements with roll angle of the mirror of 45, 90, 135, 225, 270 and 315 deg with 20 measurements averaged.
The code I wrote does the following :
1) Find the real center and roll angle of each TWE maps
Hirose-san who did the TWE measurements at Caltech placed 3 markers on the mirrors.
First I overlapped 3 circles on the markers of the map with roll angle of 90 deg as it is the same orientation as the PCI measurement.
Then, I changed the centering of all other maps + rotated them to match the circles positions of the 90 deg one.
See figures 1 to 6 for the TWE maps (piston, tilt and curvature removed) without rotation and 7 to 11 for the rotated maps.
2) Sanity check of the RoC
OSCAR is used to removed the piston, tilt and focus of the TWE measurements but states that the removed RoC is about -716 m..
This is actually due to the Fizeau interferometer setup. Computing the RoC taking into account the clear aperature and position of the reference sphere gives a mean RoC = 1.9 km (as expected).
Especially, we recovered the same RoC as measured by Hirose-san with 270 deg roll angle.
3) Combining TWE maps into birefringence
Using 4 rotated maps, it is then possible to compute the spare ETMY birefringence (delta n and theta the fast axis orientation) as in figure 12.
Actually, because there is an arctan(2*theta) used, theta is only defined between -pi/4 and pi/4 that causes some wrapping of theta (and therefore delta n as well).
So in figure 13 I used an unwrapping algorithm to try to remove this wrapping. Results are reported in figure 13.
We can recognize so similar patterns to the measurements with the PCI setup. Final conclusion should be done after a new direct measurement with the beam at normal incidence).
4) Next steps
- Tune the unwrapping algorithm -> still some issue at the bottom of the map
- Check the orientation of the mirror during PCI and TWE measurements to understand why there seems to be both a vertical and horizontal flip with respect to each other
- Finalize the direct measurement at normal incidence
For easier comparison with direct measurement with PCI (see elog 2755), I show here the absolute value of delta n flipped both horizontally and vertically .
It seems that the larger delta n area have somehow a close triangular shape.

Picture of the mirror without magnets.