Around two or three months ago, we found one of the high voltage drivers was broken. Also, we found that the control of phase shifter always saturates. So we decided to buy a new high voltage driver with a larger dynamic range. However, we just realized yesterday that this large dynamic range high voltage driver has a worse phase behavior.

This work is done with IRMC locking, I just exchanged the high voltage driver. We could see from the attached figure. The old high voltage has a better phase margin.

So we are sacrificing phase margin to have a larger dynamic range. But we have a more stable phase behavior, so we should use back the old style high voltage driver to have a better locking performance.

[Yuhang and Aritomi]

We tried to use different control bandwidth for CC2 locking. But the measurement result shows not a very reasonable result.

Usually, the higher the bandwidth the better control. However, we are not having this situation now. But it can also be related to the not well-designed control loop.

[Aritomi, Yuhang, Eleonora]

Recently we had a problem about IR back reflection from filter cavity. After installation of faraday, green phase error signal with p pol transmission is very stable (Pic 1) and we could lock green phase with p pol transmission stably. After locking of green phase, IR transmission of filter cavity is stable (Pic 2). One problem is that IR FC reflection is fluctuating even when green phase is locked and it seems to come from motion of suspended mirrors (entry 1547).

I checked the website of KTP company(RAICOL crystal), the substrate absorption loss should be 100ppm in our case.

Simon

Below you can find the first results of Yesterday's measurement with an S-polarized input beam. The structure of the map is quite similar to the wavefront-error measurements done by Hirose-san last year(?).
The homogeneity in terms of the polarization angle is ~2 degrees which is almost twice as much as for ETMX. However, also this is apparently consistent with Hirose-san's measurements.

The view onto the map is as the incoming beam would "see" the substrate.

BAB before OPO is 114.5 mW and BAB after OPO is 0.231 mW, so current transmissivity of OPO is 0.2 %. When transmissivity of incoupling mirror is T1 = 8 % and transmissivity of HR coating of PPKTP is T2 = 0.025 % and round trip loss inside OPO is L, we can get L = 12 % by solving following equation.

T1*T2/(1-sqrt(1-T1)*sqrt(1-L))^2 = 0.002

Escape efficiency = T1/(T1+L) = 40 % which is very low.

If L = 0.425 % like Marco's thesis P.87 (transmissivity of HR coating of PPKTP is 0.025%, not 0.25%), escape efficiency is 95 %.

[Eleonora, Matteo, Miyakawa (remotely)]

We have finally solved the problem with DGS.

As suspecteted by Miyakawa-san, it was due to a broken ADC timing adaptor. The main reason why we took so long to realize it is that in order to simplfy the configuration I was connecting the timing box (the one that was carefully tested) to only one of the two ADC cards istalled in the PC.  Miyakawa-san pointed out that all the ADC/DAC PCie installed in the standalone have to receive a correct timing signal otherwise NONE OF THEM will work. On the other and it doesn't matter if they are not connect to AA/AI.

So I test also the second ADC timing adapter and found out that it was broken. The pin of the SMB connector on the board (pic 1) was broken and got stuck in the SMB2BNC adapter (pic2). According to Miyakawa-san this is a frequent issue for these components.

Since we don't have any spare ADC timing adapter, I removed the second ADC card (which currenty is not used) from the standalone and connected the DAC timing adapter to the DAC PCie. In this configuration the system could work again.

Miyakawa-san will send to Mitaka two more ADC timing adapter from KAGRA.

Other usefull information:

1) The correct way to connect the DAC timing box to DAC PCie and AI is shown in pic 3 (note that the connectors for the two ports are the same). The first pdf attached shows the pin assignment of the DAC timing box which is usefull to test the connections and detect a possible break of the SMB. 

2) The pin assignement for ADC timing box is show in the second pdf attached. Since in this case the connector from ADC and to AA are different it is easy to identify the correct connection configuation.

3) The correct set up for the timing signal generator is: square wave at 65536 Hz, level 2.5, offset 1.25. It provide a squared wave of 0-5 Volt.

[Yuhang, Aritomi]

We changed plate BS for homodyne to cubic BS and aligned homodyne with s pol. Visibility is 0.988 which means loss from visibility is 2.4%. However, squeeze level is same. It seems that squeeze level is limited by unknown loss. Now we suspect that loss of OPO (escape efficiency) is large.

Simon

As there are some doubts about the linearity of the input-polarization, I checked the HWP and can confirm that it is not an accidentally taken QWP (I changed the angle to confirm its periodicity of 45deg).

In order to further increase the linearity, I have put a QWP in front of the HWP and looked for the S-pol minimum (while maximizing P-pol) on the sensors when the mass was out of the beam-path. I could reach a minimum of ~1.2 mV with having ~384 mV on P-pol-sensor. Turning the HWP by 45 degrees, I reached a S-pol maximum of ~312 mV and a P-pol minimum of ~0.9 mV.
Those are much better numbers than we had before so that we can say to have a good linearization now!

With those changes on the setup and setting the initial polarization to S, I started mapping ETMY.

Matteo, Eleonora, Simon

(Report from July 30th 2019)

As we finished absorption and polarization characterizations of the ETMX spare mass, we exchanched it with the ETMY spare mass (see pictures below).
We tried a kind of new technique for making the exchanging procedure a bit more easier and safer. Therefore, we made use of the holding-structure of the container where the mirror-substrates are transported. We found that we can use those structures to flip the substrate in a horizontal position (which is required for the measurements) and to put it directly on the sample-holder which coincidentally fits very well to these holding-structures.
That way works also vice-versa.

In addition, we recognized that both substrates indeed have a mark (probably) pointing to the thicker side of their wedges.
However, we found that this mark is on different sides for ETMY and ETMX (see last pictures).

Yuhang and Aritomi

Yesterday, we tried to put ND filter so that we could characterize the squeezing level.

We put ND-0.3, which is corresponding to 0.5 loss.

When we were measuring, we were sending 40mW of green. From the measurement of yesterday, we could know the squeezing level we are sending is 13dB. By using this squeezing level, I made the plot of squeezing level change curve. The x-axis variables in the plots is losses. And there are three curves for different phase noise level.

The measurement result of squeezing corresponding to different losses is

From this plot, it seems elog 1523 result is reasonable.

[Aritomi, Yuhang]

We measured loss of each optics between OPO and homodyne BS. We put BAB on resonance of OPO by hand and measured power. The result is as follows.

Loss between OPO and homodyne is 7%. 4% is from dichroic mirror and 3% is from two lenses. According to spec of dichroic mirror (HBSY11), reflectivity for s pol should be 99.3%. We can try to optimize the angle or try another HBSY11. Anyway we need low loss dichroic mirror and superpolished lenses.

I measured the error signal of PDH lock as shown in attached pictures (green line).
Frequency scan was done with 20 Hz triangular wave (3Vpp).
Some details will be reported.

So far I have worked with closed 80K shield, and without most outer shield of cryostat chamber.
Today I closed the most outer shield using crane.

Actually, it did not improve the PDH lock stability.
The viewport window on the most outer shield is detached in order to avoid scattered light.
So this may cause acoustic fluctuation.

We often have large bump in shot noise spectrum like an attached picture. Note that squeezing path is blocked.

Yuhang and Aritomi

It seems that the turbo pump just gives us a narrow peak at 600Hz. The on and off of it doesn't change squeezing level.

To see the contribution of noise at different frequencies, I performed the coherence measurement between squeezing and GRMC locking loop error signal/IRMC locking loop error signal/GRMC transmission signal/Green phase-locking error signal/IR phase-locking error signal.

The result is attached. From these results, we could have the following deduction:

• GRMC locking loop is contributing the noise at 9.5kHz and 6.9kHz
• The coherence between 3.7kHz and 6.9kHz shows up mainly in IR phase and IRMC locking.
• It seems IR phase noise and IRMC locking loop has strong coherence. So I think we can improve IR phase behavior by improving IRMC locking.
• 600Hz noise doesn't show up in the IRMC and GRMC locking loop.
• Sometimes, the squeezing measurement is very good. For example, the one together with GRMC loop error signal. I attach the squeezing level at that time.

All the measurement is using averaging of the meachine of 200, I think we should use more average in the future.

The large increase of phase noise above 300 Hz when the CC2 is closed (pic 3) is very strange and we need to investigate it. The loop should have a UGF of few kHz, but it doesn't seem to work correctly.

[Aritomi, Yuhang]

We measured squeezing and anti-squeezing with green power from 15 mW to 55 mW. Attached picture shows the result. Our case seems that loss is 25% and phase noise is 30 mrad. Note that this is not a fitting.

After this measurement, we checked visibility and found that visibility is 0.935 which means loss is 12.6%. (Note that when we calculated visibility, we didn't consider DC offset of PD. Visibility should be higher than this.) Then we aligned LO and BAB and visibility became 0.973 which means loss is 5.3%. However, squeeze level didn't increase.

[Aritomi, Yuhang]

Recently we turned off lasers when we leave. We accidentally kept lasers ON from last Friday and today we found that squeezing spectrum is very flat and phase noise is much less (attached pictures). Green power is 40 mW and error signal of CC1 and CC2 is 76 mVpp and 120 mVpp. Phase noise is smaller almost by a factor of 10. We guess large phase we had so far is due to unstable lasers. When we achieved good squeezing in May, lasers were always ON. We'll keep lasers on and see what happens.