Marc, Yuhang

First, we checked BAB and LO alignment inside the AMC.

LO mode-matching is about 99.9% and BAB is 99.12 +/- 0.3 %.

We measured visibility by installing a PD before the AMC and got 95 +/-1.7 %, but we expect to find same value as mode-mismatch into the AMC..

We then measured the visibility directly from the homodyne and got 98.1 +/- 0.4%.

While this value is closer, there is still some discrepancy and we are not so sure about the reason for that.

We measured squeezing and anti-squeezing with higher pump power (45mW, 50 mW, 55 mW).

The result is attached in figure 1.

From the fit, we have optical losses = 20.74 +/- 1.16 % and phase noise = 34.64 +/-3.15 mrad.

We tuned PLL frequency and OPO temperature before each measurement but forgot to record their values..

DDS RF channels are summarized in elog1488.

This information is summarized in my Ph.D. thesis Appendix B. I take them and put them here in a table.

Yuhang and Marc

As reported in logbook3064, the noise coupling from 532nm pump relative intensity noise (RIN) to anti-squeezing was characterized. In this logbook, we report the same noise coupling way, but it's from RIN to squeezing.

For two situations of no noise sent to MZ and 100mV sent to MZ, we measured four spectra, including RIN spectrum, squeezing spectrum, transfer function spectrum (from RIN to squeezing), coherence spectrum. The comparison of each spectrum in the two situations is shown in attached figure 1 to 4.

We can see that the noise injection is coherent below about 10kHz. So the transfer function measurement should be reliable below this frequency. Therefore, we take the measured transfer function when 100mV is sent to MZ for calculation in the next step.

The next step calculation is firstly a subtraction of measured squeezing spectrum (orange in Fig. 5) and expected squeezing level (black in Fig. 5). This expected squeezing level is an estimation of a constant value of 6.4 dB since squeezing is frequency independent, considering the flat region of measured squeezing spectrum above 20kHz. The result of the subtraction is the green curve in Fig. 5, which should represent all the calssical noise, such as homodyne electronic noise, pump amplitude/phase noise, and LO amplitude/phase noise. On the other hand, we use the transfer function to transfer the RIN, measured when no noise is sent to MZ, to squeezing measurement. Thus we get the blue curve in Fig. 5, which should represent the classical noise coming from the pump amplitude noise (RIN). However, we see in Fig. 5 that the green curve is even lower than the blue curve, which seems not very reasonable for me since total noise (green) should not be smaller than a signle noise component (blue).

To solve the mismatch of green and blue curves issue in Fig. 5, we adjusted the expected squeezing level by hand to 9.4 dB. Then we get Fig. 6 which makes the derived total noise (green curve) overlap with a single RIN noise (blue curve) at peaks around 10kHz. We should note that this is just a guess of squeezing level of 9.4dB, but it could explain better the classical noise coupling to squeezing. The green curve has some frequency regions higher than the blue curve, which could be also reasonable if the pump phase noise, LO amplitude/phase noise can contribute to that. However, the noise contribution from these other noise sources needs to be verified.

Yuhang and Marc

It was found that the relative intensity noise (RIN) contribute substantially to the squeezing noise spectrum, such as the peaks around 10kHz in the squeezing noise spectrum reported in logbook 3062. Thus we suspect that the amount of RIN could limit the squeezing measurement at TAMA.

In my understanding, the RIN could introduce noise through the noise amplification of the bright field of CC sideband. On the other hand, the squeezing field is generated from the DC power of the pump. Therefore, the noise from CC field and RIN should be uncorrelated with the squeezing field.

To investigate the effect of RIN, we took anti-squeezing noise spectrum measurement in two conditions. One is the measurement as usual. In the other condition, we inject 100mV white noise (from ~50 to 51200 Hz) to the 'pertubation in' point of the MZ locking servo. In both condition, we took four measurement. They are the spectrum of anti-squeezing, the spectrum of RIN (without normalization), the transfer function from RIN to anti-squeezing, the coherence between RIN and anti-squeezing.

As a fast analysis, we attach in this logbook the comparison of those spectra. A detailed analysis will follow after considering more carefully the noise coupling mechanisms.

Sorry for my late reply.

It's better to leave the actual value of beam radius [mm] with an error.

Marc, Yuhang

First we checked the LO alignment into the AMC and found misalignment in both pitch and yaw.

We tuned the Homodyne balance and checked the visibility.

After some tweaking we measured visibility about 0.92 % which is lower than expected (99%).

We found that the BAB was heavily misaligned in the AMC and after tweaking we could reach about 95.7 % visibility.

However, because our goal is to make sure that the system is in an understood state we decided to go towards FIS measurement.

(Actually we also found out that there were large scattering noise below 100 Hz, maybe due to grass cutting.. Even if the situation improved with the better homodyne visibility, we should also try to check this measurement at lower frequency)

We measured FIS with following conditions :

Results are computed in figure 1 where we could reach 5.9 dB squeezing with 40 mW of green power.

With the fit of ASQZ versus SQZ in figure 2, we have optical losses = ( 20.9+/-0.93 ) % and phase noise = (25.46 +/-3.14 ) mrad.

From our wiki, we expect 19.1% optical losses but taking into account the lower visibility, the expected losses become 23.6 %.

For the phase noise, we have quite large uncertainty because from 45 mW of green power the CC1 became really noisy. We will investigate this issue tomorrow.

During the alignment tuning, I made a mistake and moved the pitch of the steering mirror close to the mirror labelled 'FIS'. Before measuring FDS we should make sure to recover its alignment using the irises on the table.

Marc, Yuhang

We found that one SR560 that was used for the input oplev yaw was broken (increasing a lot the noise above 10 Hz) so we replaced by the one used for the length sensing oplev.

We want to find the good PSD position and first tried to measure the natural length mechanical resonance in the tilt PSD without excitation but we could not see any signal at the expected frequency (0.94 Hz).

We decided to inject white noise in the length of INPUT at this frequency.

First we used 5000 counts amplitude but we could see harmonics so we decreased the amplitude to 1000 counts.

The attached figure report the 0.94 Hz peak amplitude in the pitch and yaw spectrum while moving the tilt PSD (larger position correspond to moving the PSD closer to the INPUT mirror).

For the yaw, we can see the expected 'V shape' in addition to some constant noise which might be due to non perfectly decoupled driving.

The minimum of this 'V shape' is at about 7.3 cm which is really close to the waist position.

For the pitch, we can clearly see the 'V shape' but the minimum position is about 3 cm which seems to be really off.

We have to investigate the reason for this discrepancy.

Marc, Yuhang

In order to prepare for OPO measurement, we measured some preliminary FIS spectrum.

New measurements with better tuning are planned.

We checked the OPO non-linear gain after realigning the BAB and GR beam into it.

we measured without green 550 mV in transmission and with 25.6 mW of green 2540 mV which corresponds to non-linear gain of 4.47.

This is a bit lower than expected (should be about 5).

We tried to tune the OPO temperature (starting at 7.130 kOhm with 10 Ohm increment) but it seems we were already around the optimal temperature. The p pol PLL frequency in this condition was 225 MHz.

We also had to realign the GRMC and had some issue with the error signal sign before locking it.

We tuned the Homodyne balance by injecting noise into the IRMC (sine with 1.2 V Pk2Pk at 800 Hz) and trying to minimize it from the homodyne sub DC channel.

We checked the LO and BAB matching into the AMC and after some tuning could reach about 99.9 % mode-matching for each field into the AMC.

Finally we measured shot-noise and FIS.

We tried to change the phase of the FIS but could not reach more than 4 dB.

Possible reasons are that we forgot to optimize the OPO non-linear gain or the Homodyne visibility.

Furthermore, we had to tweak alignment at several places so we might have to retune it today to have a more stable alignment.

Despite the fact that they have a power knob, the laser model that we use for oplev don't allow to tune the output power. 

Aritomi, Marc, Michael, Yuhang

We reset the DDS rack, restarted the control PC and reloaded all configurations.

We checked input oplev and found that the laser output was really low (sum read by PSD was about 150 counts while usual value is about 5000).

We replaced the laser head by a new one and could reach up to 7000 counts.

To avoid the saturation, we tried to reduce the laser power, but we couldn't tune the laser power at all...

In any case, we installed 2 OD1 and could reach about 5000 counts on the PSD sum.

What I did : I measured gaussian beam radius and fited beam propagation using beam profiler and Python.

Why I did: trying for Pound-Drever-Hall method.

Experiments: I set up the following in a attached picture.

Result:The result of fitting is also in a attached picture.

Conclusion: I could have measured the beam radius.

Problems:There are some possibility that I mixed optical elements in different wavelength bands (1550nm and 1064nm) in ATC.

What I did: I checked the suspected  optical elements  used for 1550nm or not.

Subjects: Two HWP and one QWP that were suspected in 1064nm,but one HWP(named "1") is cleaned by using first contact. So I checked one HWP(named "2") and one QWP(named "3").

Experiments: I set up the following in attached pictures. In experiment for the suspected HWP(named "2"),maximum mW and mimimum mW showed in Power Meter was as follows.

In experiment for the suspected QWP(named "3"), maximum mW and mimimum mW  showed in Power Meter was as follows.

Conclusion: Based on the above results, it can be concluded that HWP(named "2") and QWP(named "3") are for 1550nm wavelength.

Next: I will check the suspected HWP(named "1").




Addition: I checked WHP(named "1"),WHP(named "2"),and QWP(named "3")again.I set up the following in additional attached pictures.Additional experiments were set up the same as before.

 In experiment for the suspected HWP(named "1"),maximum mW and mimimum mW showed in Power Meter was as follows.

 In additional experiment for the suspected HWP(named "2"),maximum mW and mimimum mW showed in Power Meter was as follows.

In additional experiment for the suspected HWP(named "3"),mW showed in Power Meter at point A,B,and C was as follows.

Based on the above results, it was concluded that HWP(named "1"),HWP(named "2") and QWP(named "3") are for 1550nm wavelength.There was no mixing.

I installed Debian 11 on the Data Concentrator/Network Data Server/Frame Writer computer in TAMA. In 2951 we ran across a problem where we couldn't complete the software installation for the Debian 11 install, and for some reason couldn't install a boot loader properly. The problem was that the installation medium was running in BIOS mode as opposed to the more appropriate UEFI mode. Switching to UEFI installer upon startup made the installation go smoothly as for the frontend computer.

The following options were chosen during installation:

Language - English
Region - Japan
Locales - en_US.UTF-8
Keyboard - American English
hostname - DCCentral
domain name - mtk.nao.ac.jp (default)
Network interface - eno1 Intel Corporation I210 Gigabit Network Connection
root password - normal KAGRA controls password
ops user 1000 (same password as usual)
controls user 1001 (same password as usual)
package manager - deb.debian.org
HTTP proxy - blank
 

The following drive partitioning scheme was chosen:

Disk 1 - 1 TB Western Digital ATA - for operating system and user files
512 MB - EFI System Partition (ESP) - in the failing step last time, using the BIOS mode installer would not let us install this partition. Now it's fine.
128 GB swap space
Remaining (~870 GB) - ext4 / - perhaps there's no real need to separate into /home, /var, /tmp here, so the system can just figure that out by itself

Disk 2 - 24 TB Avago MegaRAID virtual drive (4x 7.6 TB disks with RAID5 redundancy) - for mass storage of channel data
24.0 TB - ext4 /data0 - used for frame writer. In KAGRA, there are two frame writer PCs k1fw0 /data0 and k1fw1 /data1 with about 25 TB storage each, for extra data redundancy. But we only have one server PC, so I just assume we are using one frame writer.

Afterwards, I added ops and controls to the sudoers file and installed cdssoft version 1.0.11 (was 1.0.9 when installing frontend) from the aligo caltech Debian Bullseye repository.

*I will comment later in a bit more detail about what some of these technical terms mean

Marc, Yuhang

We have to reload DDS2 config everytime to be able to lock.

Then we are looking at the GR_corr spectrum to investigate the driving matrix of the INPUT but we found out that the noise level is not stable at all (see figure 1).

Our reference level is in blue with injected line on INPUT_PIT_ex2 at 5 Hz and 5000 amplitude is blue curve of figure 1.

At other times, we can see noise increase up to 10 Hz or even at higher frequency.

This is true with or without line injected, with or without AA or pointing loop.

Same after restarting laser.

Not affected by closing AA or pointing loops.

After some time, the noise excess disappeared and we could start some measurements but it quickly reappeared again..

PIT_ex2 4500, GR_corr 7.52

4000, 6.70

3500, 6.47

3000, 5.97

2500, 5.38

Marc, Yuhang

We found that GR reflection was far from overlapping with injection.

This was due to huge offset on input and end because AA loop was still closed.

After opening the loop, we could see TEM00 flashes but we had to have about 100 offset in input and end pitch.

We could not lock the FC so we checked the GR power while changing the SHG temperature.

GR power after FI (mW): 41.7, 45, 35, 44, 45

SHG temperature: 3.085, 3.071, 3.06, 3.0665, 3.075

Finally, we tuned the SHG alignment and could recover more than 50 mW after the FI.

We could still not lock so we tried to tweak the servo parameters :

gr lock servo, attenuation = 0.2, gain = 8, both changed no diff

gr reflection power > 200 uW

Finally, after changing the lock threshold we locked on a sideband. Turning off the servo seems to change the sign of the error signal and we could lock the FC.

In the end, we changed sign on DDS2 by reloading the latest configuration.

I set IR beam power to 1.6mW, I checked vertical and horizontal AOI with the razor blades and got respectively -0.008 deg and -0.022 deg.

I tuned the QWP and HWP angle to minimize p polarization while injection s polarization.

I did the calibration and started measurement of #4 (was cleaned with first contact) with s polarization at the input.

Yuhang and Marc

We would like to have 'tilt driving' coupling to 'shift driving' as small as possible. To achieve this, we worked on finding optimal driving matrix for H1 and H3 for pitch today. 

At the beginning, the driving matrix for pitch is: H1 0.794, H3 1. This was optimized in the past. However, we would like to optimize it again since we glued new magnets at the beginning of this year.

To check how much coupling we have, we sent 5Hz 5000 amplitude for INPUT_PIT_ex2. We checked coherence between GR_CORR and excitation, no coherence was found.

Then we checked when driving matrix is: H1 1, H3 1. We found no coherence even in this case. We checked the connection, we found later on that actually GR_CORR is not connected.

After putting back good connection of GR_Corr, we take data when H1 and H3 are both 1. We saved as reference as blue (REF0 and REF1).

H1 0.794, GR_corr 17.964 (REF2, REF3)

H1 1,         GR_corr 30.475

H1 0.6,      GR_corr 13.38

H1 0.5,      GR_corr 15.82

H1 0.55,    GR_corr 13.82

H1 0.65,    GR_corr 13.09

H1 0.65, H2 0.02,  GR_corr 12.63

H1 0.65, H2 0.05,  GR_corr 13.61

H1 0.65, H2 0.01,  GR_corr 12.30

H1 0.65, H2 -0.01,  GR_corr 12.82

When we drive H1, we get GR_corr 87.08

When we drive H3, we get GR_corr 63.24

We found a ratio of 0.73. Then we did H1 0.73, we get GR_corr 10.71

Attached figure shows a comparison of GR_corr when H1 is 1 or 0.73.

We still have coupling as we see the number in GR_corr 10.71. We still would like to understand it better what is causing this number. One thing we will check is the coupling factor when different amplitude signal is sent to H1 and H3. We will also test this coupling number by sending excitation at a different frequency.

Aso, Marc, Matteo

We put marking on AZTEC #1 and #3 to indicate the orientation of the sapphire during shaping.

One large arrow on the top pointing towards the front surface during our measurement + 2 long lines on the side at the level of the holder.

We packed the samples inside their wooden crates with foam to prevent their movements during shipping.

Marc, Matteo

We rotated the sample by 90 degrees and fine tuned the rotation angle to minimize p polarization power at the readout.

Results are attached to this entry.

It seems that previous position was better so today we rotated the sample back to this position.

Actually, trying to minimize the p polarization after the sample, we rotated clockwise from injection part by an additional 2.54 deg.

From HWP, it seems that we are now at 0.6 deg from theta = 0 deg.