NAOJ GW Elog Logbook 3.2
Aritomi and Yuhang
Today we measured the visibility in a more precise way. Because we use PD and we could fix it and measure it for a longer time. Then we take the mean value of two beams power.
After the improvement of BAB to AMC matching, we did this and the result is shown in the attached figure.
Then we tried to measure squeezing. However, we could measure only roughly 3dB of squeezing. And then we checked the visibility. It is only about 88% at the time of measurement. So in principle, we should see more squeezing. But unfortunately, we have again the misalignment of the homodyne. Then I checked these two beams with AMC. I found out both of BAB and LO were misaligned.
We will measure again after the alignment of homodyne. We still don't know why this alignment can be so easy to be misaligned! We had visibility of 99% before all the measurement and then we found 88% after a not decent measurement. This happens within 5 hours!
Yesterday, I found the leaking pipe around the tank in TAMA central area storage room. Today Takahashi-sensei came here and repaired it. We found there is a hole on the pipe. We cut the broken part and reconnect the shorted pipe back to the original position.
As pointed out by Takahashi-sensei, the old air conditioning system is just used for backup. In the future please use only the new airconditioning system.
As shown in the attached picture, the old air conditioning system has some problems.
I put two irises.
One is before the half wave plate and along the squeezing path.(Shown in the attached figure 1 and 2) As you can see this iris equips with a rotateable blocker. We chould also use this block or unblock the beam.
The other is just before homodyne's BS along the LO path.(Fig 3)
I also put a black wall to block scattering lignt.(Fig 4)
Good work! By the way, loss is visibility**2 so loss will improve by ~9%.
By moving the lens position, I could improve the matching from OPO transmission(BAB) into AMC to a decently high level. This will improve our losses situation by roughly 4.5%. Before, the matching of this beam was roughly 95%. Now it is (1.3680-0.0068)/1.3680 = 99.50%
In the attached figure 2, there is some residual s-pol(actually it is p-pol, but we put a half-wave plate. So this residual beam becomes p-pol when it arrives AMC). But it comes from the defect of PBS.
Good work! By the way, loss is visibility**2 so loss will improve by ~9%.
This entry is a log on Apr. 23.
I swapped the mirror to PBS for splitting beam into TEM00 path and HOMs path.
TEM00 path beam has ~2 mW power with slightly tilted PBS (HOMs has 16.6 mW).
The reason PBS has slightly tilted incident angle from 45 deg is that the PBS is 56 deg Brewster PBS, and very low transmitted power with 45 deg incident angle.
Then I confirmed polarization of both transmitted and reflected beam are not circular polarization.
I made the codes on Python to compute the 2D graphs in order to study the robustness of the injection (fig 1) and reflection (fig 2) telescopes, taking into account the correlations moving the two lens.
The mismatch is consistent with the results of entry #1311 and there is a region of positions in which the mismatch is under 10% also for the reflection telescope.
I measured again the squeezing and anti-squeezing after the engagement of cc noise eater. Also, the measurement is done after the implementation of the s-pol GRMC lock. The measurement result is attached.
We could see that we have 3.30dB of squeezing and 16.47dB of anti-squeezing. This precise number is done by averaging the noise spectrum from 30kHz to 500kHz and then subtracting. In the attached figure, we can see there is still a lot of peaks.
Since we have enough green power, we decided to use s-pol again. By changing the gain of GRMC and MZ servo, we could lock both of them again. Also, I changed the integrator of MZ.
We could have 50mW of green light going inside OPO as before.
However, I observed a more stable coherent control 1 loop. This is quite beneficial for the future.
The pitch local control loop of PR has been closed.
The mechanical TF and the closed loop TF are shown in pic 1 and 2. The comparison between the open and closed loop spectrum is shown in pic 3.
UGF is crossed two times at 3Hz and 10 Hz. The phase seems above 50 deg.
Participant: Yuhang, Matteo, Eleonora, Aritomi
We checked many things and want to figure out why we have a 300kHz peak in the spectrum of homodyne.
We tried to remove green by putting line filter(1064nm), tried to investigate the locking of OPO, tried to see the effect of leaked p-pol to homodyne. Finally, we confirmed the problem comes from the coherent control beam directly.
Then we found the noise eater doesn't give any difference when we switch on/off noise eater. So we suspected that this is because we are using not enough power of cc laser. This guess is mainly from the remind of Chienming. Then we tried to increase the cc power. We found the peak disappeared after going beyond a current value of ~1.2A. So we confirmed that increasing the current value above ~1.2A can engage the noise eater.
Then we set the current value of 1.305 and temperature of 34.37 degrees for cc laser. This is done by compromising available ND filters, desirable value 15mW of IR after filters and the avoid of mode hop. As we know, we lose alignment each time after putting the ND filter. We also recovered alignment. The alignment situation is attached in figure 1. We will keep this setting for the future until we find additional problem.
Participant: Eleonora, Aritomi, Matteo, and Yuhang
Today we found the problem why I can have so large noise of cc-PLL. The reason is fiber PD is broken again. We just swap the PD and we could lock cc PLL very well. After the swap, we measured the beatnote level which is 7dBm now(measured by hp-E4411B, so the real amplitude should be -10dBm). This should be a reference for the future.
Then I measured the phase noise of both loops again. The result is shown in the attached figure 1. As you can see, in this figure, the RMS phase noise of cc PLL is 5mrad. This is 30 times smaller than the previous measurement. (Actually, I made a mistake of estimating the phase noise level of the previous measurement) While the measurement of p-pol PLL shows RMS phase noise of 15mrad, which is 3 times higher than the measurement of Marco.
While I was checking the demodulated beat note of p-pol PLL, I found a very low-frequency oscillation. This is shown in the attached figure 2. We should investigate how to remove this oscillation because it brings us almost 1rad of phase noise, which is a lot.
Next step:
buy new power cable for fiber PD or many batteries.
we should also check the level of p-pol beat note.
The DGS input/input output voltage ranges are:
ADC: ± 20 V
DAC: ± 5 V
The volts to counts calibration is 2^15/(Vpk):
ADC: 1 V = 1638 count
DAC: 1 V = 6544 count
I followed the Marco method and measured the phase noise of CC PLL. It shows an RMS phase noise of 149mrad. It is almost 50 times higher than p-pol PLL phase noise level.
I measured the p-pol PLL fast and slow loop correction signal. We can see from the attached figure. Although at that time fast loop is not stable, it shows very low-frequency drift. But slow loop reads this signal can try to bring the loop back to the original state. Since I calculated the correlation coefficient of these two signal, the slope of these two signal is the same. So the correlation coefficient is -1.
I think this is better than the coherent control loop. It is measured and shown in the entry here.
Yuhang and Pierre
We tune the servo for locking the coherent control loops.
For green coherent control, we use 20dB attenuator and 50Om for error in. The measured open loop transfer function is attached as figure 1. We have unity gain frequency of 85Hz.
For local oscillator coherent control, we use 30dB attenuator and 50Om for error in. The measured open loop transfer function is attached as figure 2. We have unity gain frequency of 51Hz.
I did another simulation for injection (-400 mm focal is not a common lens) and for the reflection (avoiding to change the already installed injection telescope into the homodyne).
The robustness for the injection telescope is really good, less than 3% moving the first lens in a range of +/- 5mm and less than 10% for the other one.
The robustness for the reflection telescope is not as good, we reach also 20% mismatch for +/- 5mm movement of one lens.
0- Current configuration:
Notch filter 1, Notch filter 2 ans LP filter are disable.
The Servo-filter must be set only on 1/f integrator.
An attenuator of 30dB with a 50 Ohm load is set on the ERROR IN input.
The gain is set to minimum (position 0).
The unity gain frequency was measured to 50Hz.
1-Setting of switches on the front panel:
* The differentiator shall be disabled on the front panel in setting the switch on "OFF".
* The switch INV/NON INV on the front panel, shall be set on INV.
2-Setting of the 8 straps on the board:
Notch filter 1, notch filter 2 and Low-pass filter are disabled in setting strap on connectors P7, P8 and P9 between pins 2 and 3.
* The transmission signal is ont used.
The strap on connector P4 (3 pins) is set between pin 2 and 3.
* Strap is set on connector P11 (3 pins), between pins 2 and 3, in order to activate the sample-and-hold on the triangular signal, on the locking.
* Strap is set on connector P3 (2 pins) to connect the triangular signal to the output stage.
* Strap is set on connector P2 (3 pins), between pins 1 and 2, for test purpose.
To check notch 1 and notch 2 filters (in scan mode) between TEST IN and TEST OUT. For this test the differentiator, shall be set on "ON" (not intuitive but important). After this test, the differentiator shall be disabled the front panel in setting the switch on "OFF".
* Strap is set on connector P1 (2 pins), in order to be able to tune the offset.
3-Modification of components:
* Integrator 1/f: corner frequency changed to 22 kHz
Capacitor CMS 1206: C38 = 3.3nF
* Integrator 1/f2: unchanged
* Low-pass filter: unchanged
* Notch filter 1: unchanged
* Notch filter 2: unchanged
* Gain adjustment (G): Gmin = 0.0125 / Gtyp = 5
* Input impedance
Resistor CMS 1206 : R145 and R146 removed