NAOJ GW Elog Logbook 3.2

[Matteo, Eleonora]
We reinstalled the second ADC PCie into the standalone after it was taken out because of the timing-box issue.
The standalone PCie slots are piled up in vertical are (from the bottom):
1) ADC1
2) ADC2
3) DAC
4) empty
We connected the new timing box to ADC2 and the real time model was restarted without problems.
We installed one of the two new BNC2dsub box in the clean room (to be used for AA signals) and we connected the cables from to the AA module which is in the standalone rack.
Now the situation is:
ADC0
-16 channel in the standalone rack (used for local control)
-16 channel in the clean room. (4 used for FC lock signals, 12 available for AA)
ADC1
-16 channel in the clean room. (available for AA)
-16 channel possibly available by installing an additional BNC2dsub box
Now we need to test if ADC2 can correctely acquire signals. I need to modify the real time model to do this.
Pic1: modules on the standalone rack in the corner of TAMA central area
Pic2: modules on the rack in the clean room

Since we had problem of saturation, I decreased again the power of LO. Then the measurement of visibility if consistent with the two homodyne PDs. But they are not exactly the same.

We are still having the resonance issue of CC2. As we reported, we could lock CC2 wih bandwidth of kHz. However, sometimes if the resonance is excited by some random vibration. We have resonance at 3.2kHz. This limits our CC2 bandwidth. For example, sometimes we could only lock it with bandwidth of 400Hz(first picture shows this). The resonance is shown in the attached picture if we increase the gain.

Aritomi and Yuhang
I attach here the plot of SQZ, Anti-SQZ and shot noise level. These are used for entry 1571 for the estimation of loss and phase noise.

Aritomi and Yuhang
In the beginning of yesterday's recovery of squeezing measurement, we measured LO spectrum with homodyne and found lots of peaks at the low frequency region. We checked the clipping and the centering of beam on homodyne PD. It was fine.
However, since we recognized that the frequency of 24.5Hz comes from scroll pump. And 20Hz comes from clean room fan. The 13.5Hz peak comes from bench horizontal mode. We decide to put back the board wihch is taken away because the demodstration requirement we had on Monday. After that, the measurement of LO spectrum becomes very clean.
Maybe the balance of weight on top of bench helps to remove the bench horizontal mode peak. And the board isolate the sound wave propogates to the mirrors and PDs. These help to remove peaks. But the source of peaks at 16Hz, 18Hz and 33.8Hz are still unknown.


green power (mW) | MZ offset | OPO temperature (kOhm) | p pol PLL (BAB) (MHz) | p pol PLL (CC1) (MHz) | BAB maximum (V) | Demodulation phase of CC2 (SQZ) (deg) | Demodulation phase of CC2 (ASQZ) (deg) |
0 | 315 | 0.114 | |||||
15 | 4.01 | 7.16 | 190 | 190 | 30 | 100 | |
20 | 4.1 | 7.16 | 175 | 175 | 0.54 | 30 | 100 |
25 | 4.19 | 7.16 | 165 | 160 | 0.704 | 30 | 100 |
30 | 4.29 | 7.17 | 157.5 | 165 | 35 | 90 | |
35 | 4.38 | 7.17 | 162.5 | 160 | 1.34 | 48 | 90 |
40 | 4.5 | 7.18 | 165 | 2 | 50 | 80 | |
45 | 4.58 | 7.19 | 175 | 150 | 2.92 | 50 | 80 |
50 | 4.68 | 7.19 | 162.5 | 155 | 4.08 | 53 | 75 |
55 | 4.78 | 7.2 | 167.5 | 155 | 5.52 | 58 | 75 |
60 | 4.88 | 7.2 | 160 | 7.12 | |||
65 | 4.98 | 7.2 | 155 | 9.04 |
Squeezing and anti squeezing plot is attached. The result seems more reasonable. We have 6.1dB squeezing with 40mW green! We estimated loss is 20.8% and phase noise is 26.3mrad. Given that loss between OPO and homodyne is ~8% and loss from homodyne (visibility, quantum efficiency) is ~ 3%, loss from OPO seems ~10% (design is 5%).

Simon
(This is report from Yesterday)
I inspected the OSTM mirror substrate by my eyes in the TAMA clean-room (absorption bench) to make sure that there are no major damages (pictures attached).
So far, it looks good. I packed the substrate again and left it in the shelf in the anteroom.

[Yuhang, Eleonora]
Today we worked on the recovery of the FC after summer break. We recovered easily the alignment towards the end mirror but once we aligned the input we couldn't find any flash.
Since from the signal we suspected a saturation of the stanford research used to amplify oplev signals in the end, we went to check and we found a major water leakage in the end room.
The floor below and around the vacuum chamber (which is a bit lower with respect to the rest of the room) was completely covered with water (~1cm deep).
While switching on and off the air conditioner we observed that a discrete amount of water started to drop from it along and wall, reaching the floor. We suspect that the drain system of the air condition is not working properly.
We tried to dry the water with some paper but a part of the water is still there.
Anyway we decided to go on with the recovery: we zeroed the oplev signal, so that the stanford where not saturating anymore and we could close the loop properly. Than we realigned the end mirror, found the flashes and lock the cavity.
To realign the end mirror, we did the usual trick that is to let the beam pass through the second target hole and look for the reflection from the end mirror on the back of the second target.
Tomorrow we will remove the remaining water and ask Takahashi-san if he can check with us the air conditioner.

As suggested by Matteo, our squeezing measurement issue can be related to homodyne. The idea is to check the visibility of homodyne by using two PD of homodyne and compare them. To check the visibility, I used BAB and IRMCtra to make the beat note. For the BAB maximum transmission, the new PLL locking frequency I got today is 315MHz (without green).
For homodyne PD close to IRMC:
BAB is 500mV
LO is 1.835V
In this case, I found PD is saturated. So I put an OD 0.5 filter in front of IRMC. After this, LO becomes 644mV.
The beat note is shown in the attached figure 1. In this case, visibility is 90.37%
For homodyne PD far from IRMC(everything is the same apart from this PD):
BAB is 498mV
LO is 647mV
The beat note is shown in the attached figure 2. We could see that it is saturated. This is strange because this PD should have the same response with the other. Or we should not use this homodyne in this way because it is designed to use both PD at the same time. Anyway, we should investigate if this is a problem or not.

Today I checked the alignment of SHG, GRMC, OPO, IRMC and homodyne.
Among them, only OPO and homodyne is misaligned. So every misalignment is related to OPO.
Homodyne visibility is measured as 90.37%. (The situation of homodyne will be reported in the following entry)
Fig.1 SHG spectrum
Fig.2 GRMC spectrum
Fig.3 OPO p-pol spectrum
Fig.4 OPO p-pol spectrum(after alignment improvement)
Fig.5 IRMC spectrum
Fig.6 OPO CC spectrum
Fig.7 OPO BAB spectrum
Fig.8 OPO BAB spectrum(after pitch alignment improvement)

Kubo-san, Yoshiyuki, Simon
I inspected today the big furnace at the ATC (located besides the exhaust hood) to see how we can use it and whether it is available or not. Please see also the pictures attached.
The model is: ISUZU DSTR-314K
max. temp.: 1220 degrees Celsius
The inner area seems not very clean. So, in order to use it for KAGRA substrates, we need to create some kind of enclosure to save the substrates from pollutions (especially penetrating atoms/ions).
I will try to find a manual but it seems that the actual usage is relatively simple. Also, since almost nobody uses it now, we can take our time.
Before using it, however, we have to write an Email to Kubo-san and Yoshiyuki.

Here I attach the auto-alignment telescope design circulated by Yuefan. Just copy Yuefan design.
distance from FI to lens1 | focal length of lens1 | distance from lens1 to QPD1 | |
NF | 0.975m | 500mm | 0.65m |
distance from lens1 to lens 2 | focal length of lens2 | distance from lens2 to QPD2 | |
FF | 0.454m | -50mm | 0.6m |

The injection telescope for squeezing injection is changed(entry 1546) because of the existing CVI lens we have is limited. It is good to know the robust of this new telescope. Thanks to the code EleonoraPolini sent me, I easily performed this test.
Since only the injection telescope is changed, I only did the test for this injection telescope. Compared with the previous design(entry 1366), the robust is similar. So I think we can be happy.

Aim: We made some measurements with the quadrants to check if both RF and DC are working fine
For testing the quadrant, we took a beam from the main laser with a beam splitter (BS1). Scheme in figure 4.
At the first tried, we just put another beam splitter (BS2) after the first one, then EOM in one of the path with a modulation frequency of 23MHz. One lens was put between BS1 and BS2 in order to have the beam waist inside the EOM. After combining two pathes with BS3, we put a filter with OD=1 to avoid injecting too much power on the quadrant. Then we place the quadrant with the beam more or less centered.
But we could not see any beats from the RF output of the quadrant. We thought one possible reason could be the phase modulator makes 2 sidebands (one above and one below the light frequency), both of them interfere with the non-modulated beam. Probably two beat signals cancel each other almost perfectly because the beam path is the same (there is no such thing as a cavity that can treat the two sidebands differently).
So we decided to add the AOM on the non-modulated path with modulation frequency of 80MHz. With two beams of similar size and well aligned on the quadrant, we could see the 80MHz beat in the spectrum.
Then we demodulated the RF signal with the box. By checking the in-phase(I) and quadrature(Q) of one of quadrant output , we could see that not only the power between I and Q is changing, but also the total power (both signals are moving up and down together).
Since our oscilloscope was not able to calculate the total power of two signals, we decided to put the second quadrant at the other output of BS3 and accquired one group of data while we knocked on the bench. The raw data we got show in figure 1. From this figure, we confirmed the quadrants are working fine. The I and Q are perfectly out of phase for two quadrants. The fact that the I and Q signals of the same QPD do not have the same peak-to-peak amplitude could be explained by a path length change of less than one wavelength.
Then Martin did some analysis of the data. In figure 2, he calculated the I^2+Q^2 for both quadrants, and also the sum of them. The amplitude of two quadrants have more or less each others inverted signal. The fluctuation in the sum is smaller than that of the individual amplitudes. Hence the total power is almost conserved. The remaining fluctuation in the sum could be due to unequal beat signal amplitude (for instance due to a difference in ND filter, or a difference in overlap of the beams on the different QPDs).
In figure 3, the phase of both quadrants are moving in the same way, it could be caused by the path length difference before BS3, so it has same effect for both the quadrants. But there is this factor of 2 difference in the phase changing, which we didn't really understand. Part of it could be caused by the small DC offset of the raw signal.
After testing the RF signal of the quadrants, we also tested the DC with the galvo. We simply put the galvo in front of one of the quadrant (see the bench picture in figure 5), and did some rough alignment to make sure that the beam is inside the diode. Then by sending four quadrant DC outputs to the galvo controlling board, two outputs of this board will be used to control the x and y direction of the galvo. Then we checked the x and y error signal through the monitor port of this board. As long as the galvo loop is closed, the galvo is able to bring the error signal back to 0, which means the beam is centered on the quadrant.
Conclusion: With two quadrants, we got figure 1 that confirm both of them are working as we expected. For the DC, the galvo is able to center the beam on the quadrant. Now everything has been removed from the bench and packed. We will send them to NAOJ today or latest next Monday.

I checked the CVI lens we should buy if we want to replace the thorlabs lens now we are using between OPO and homodyne.
The result is attached in the figure. Since we could find a lens with the focal length similar to thorlabs focal length, the simulation result is similar to the previous simulation. In this case, the lens we should buy from CVI is PLCX-25.4-46.4-UV and PLCX-25.4-70-UV.
Also if we have again almost 100% match from OPO to AMC, the telescope design for filter cavity injection should be not be influenced.

If we just consider this is a factor of 11 between green and IR filter cavity locking accuracy as pointed in entry 760 figure 6, we could estimate the IR locking accuracy of the filter cavity. (in green locking case)
As reported in entry 1486 and entry 1513, we have green filter cavity locking accuracy of ~4Hz. This is corresponding to 4e-12m.
Then by considering the factor 11(because of cavity pole of green and IR is different), the locking accuracy of IR should be 0.36e-12m. In this case, it meets the requirement of filter cavity length fluctuation.
But we still need to confirm the measurement of entry 1486and 1513.

If we consider an error of reflectivity of HR coating of PPKTP(0.02%). We could have a very different estimation of OPO round trip loss and OPO escape efficiency.
Note that here we use transmissivity of OPO of 0.2% as a prior (as entry 1538).
Note that if reflectivity of HR coating of PPKTP is 99.995%, we can explain 10% loss from OPO in loss and phase noise measurement.

Escape efficiency is T/(T+L) where T is transmission of output coupler and L is intra cavity loss. So escape efficiency should be 0.08/(0.08+0.00425) = 95%. Calculation in Marco' s thesis seems wrong.

From Marco thesis, the escape efficiency is 0.92/(0.92+0.00425). It is 99.5%, it seems fine in that case.