NAOJ GW Elog Logbook 3.2

Pengbo, Simon
We have started to analyze the coated OSTM from Shinkosha regading the absorption of the HR side.
First, we took out the OSTM and inspected the mirror visually. We found pencil marks on the barrel and among them an arrow that indicates the thicker side of the wedged substrate and shows toward the HR side of the mirror (see attached photos).
The actual orientation of the sample inside the sample holder was a little bit tricky mainly because of the wedge and the size of the sample-holder which is basically too large for such a mirror.
- At first, we tried to put the thicker side upside to have a somewhat parallel orientation between sample-holder and mirror-plane (as can be seen from the pictures, we are using a - with optical tissue - covered ruler as a spacer). However, we recognized that this will distract the pump-beam so that it cannot be measured anymore regarding its power.
- Therefore, secondly, we rotated the mirror by 90 degrees so that the distraction of both pump and probe beam is only parallel to the optical table which can be counter measured by a respective relocalisation of the IU and the photometer. However, with this position, the HR coating is facing the pump but also the spacer, which we originally wanted to avoid
- Especially regarding the probe, we had to change the IU position by ~3mm less than it would have been the case for a wedge-free substrate
After the alignment of the sample, we looked for the exact position of the HR-coating by applying Z-scans. We carefully increased the laser power to take care that the coating is not damaged (initially, we did this on the outer edges of the mirror of course).
Then, we ran a map-scan in the center with 15mm radius in the position were we identified the coating (Z = 48.8). The results of that scan can be seen also in the pictures attached. Our main result is a quite homogeneous mean absorption of 16 ppm (+/- 3ppm) with some point-like excesses indicating the positions of either dust or defects within the coating, most likely.

Now we could lock CC2 loop with unity gain frequency of 2kHz. To see the difference of CC2 phase noise with different gain. We measured phase noise with different gain.
As expected, higher gain make noise lower at low frequency. But also the higher gain excites resonance at higher frequency.
From the measurement, it seems the gain of 0.2 is the best case. (Although they are all quite similar)

[Aritomi, Yuhang]
We measured CC2 free running phase noise when filter cavity is locked and aligned with dithering (attached picture). We cannot lock CC2 stably since piezo actuation range is not enough. We'll try to feedback CC2 error signal at low frequency to input mirror of filter cavity.

R_surf = AC_surfref/(DC_surfref*P_in*abs_surfref) = 18.1 [1/W]
where AC_surfref = 0.425V, DC_surfref = 3.55V, P_in = 0.030W and abs_surfref = 0.22
R_bulk = AC_bulkref/(DC_bulkref*sqrt(T_bulkref)*P_in*abs_bulkref) = 0.741 [cm/W]
where AC_bulkref = 0.072V, DC_bulkref = 4.2V, T_bulkref = 0.55, P_in = 0.030W and abs_bulkref = 1.04/cm

Aritomi and Yuhang
We checked power at several points this Monday and make IRMC transmission set at 1.7mW.
Actually, this means we increased also the error signal(or increase gain). Today we checked the error signal, and actually, it was quite close to oscillation. So we adjusted the gain while looking at the error signal and measured transfer function again. In the end, we put the gain value of the control board from 1.3 to 0.8.
The open-loop transfer function now is as the attached figure 1.
We also measured the IRMC error signal spectrum while IRMC is unlocked and locked. As shown in the attached figure 2. From this locking performance, we could see that the loop suppresses the even harmonics of a fundamental 9Hz oscillation while there are still some odd harmonics left. Also, the 50Hz and its harmonics are introduced after closing the loop. So the PD doesn't introduce any 50Hz noise.

Aritomi and Yuhang
Since the unity gain frequency of CC2 was only ~400Hz because of some resonance. We decide to put some more damping material.
So we put some double layers bent rubber(as shown in the attached photo1) in the position where there may be some spring resonance(as shown in attached photos 2 and 3).
Then we measured OLTF, which is shown in the attached photo 4. And it is shown that the unity gain frequency is around 2kHz now. We also tried to increase the gain, we measured the resonance frequency when there is oscillation. As shown in the attached photo5, the oscillation is at ~23kHz.
We also measured OMTF. By comparing the measurement we did before putting this new damping rubber, we found the peak around 1 and 2kHz disappeared(as shown in the attached figure 6).

[Aritomi, Yuhang]
First we measured squeezing when filter cavity is locked/unlocked (Pic. 1). We expected more squeezing since we improved reflectivity of dichroic mirror by 4%, but squeezing level is still around 6.1dB when filter cavity is unlocked. Squeezing level when filter cavity is locked/unlocked is similar, but phase noise seems suppressed when filter cavity is locked.
Here are some information of this measurement.
green power (mW) | OPO temperature (kOhm) | p pol PLL (MHz) | Demodulation phase of CC2 (SQZ) (deg) | Demodulation phase of CC2 (ASQZ) (deg) |
40 | 7.19 | 165 | 105 | 135 |

[Eleonora, Matteo]
Today we did good progress on the dithering.
We gave up on measuring the sensing matrix and we tried to close the loops by simply feeding back each demodulated signal to "its" mirror.
This seemed to work surprisingly well. (See pic 1-2). The transmitted power increased and became quite stable as well as the lock.
I tried to move one by one the BS, the INPUT mirror and the END mirror when dithering is engaged and the loops seem able to move the cavity mirrors to recover the good alignment. (See pic 3, 4, 5)
From the last two plots we can also check the level of coupling of our error signals, and possibly improve the driving.
The main problem I see is that the error signals are not oscillating around zero but they have an offset and the dithering lines are still well visible in transmission. It shouldn't be like this but I'm not sure about the reason.
Currently the corrector filter is a simple pole at 0.001 Hz. Maybe we should move it to zero. I will also try to put a offset in the loops and see if it can improve the transmission and reduce the dithering line in transmission.
Even if the transmitted power is very stable because we were mostly affected by pitch misalignment, I also tried to close the loops in yaw. They seem not to work well. The error signal is always around zero, so it is not easy to tell the difference when I close the loops, but when tried to misalign the BS in yaw they didn't recover. I didn't spend much time to investigate the problem but I will do it soon.
Since the error signals get crazy if the cavity unlocks I decided it was safer to stop the lock for the night. The cavity kept the lock for more than 5 hours (best record ever) and the lock was stopped on purpose before I go to sleep. It will be good to implement some kind of guardian that will open the dithering loops when the cavity unlocks.
Some details:
- The gain of the input and end loops are respectively -1 and -0.5.
- I reduced the amplitude of both dithering lines from 10000 to 5000 counts.

The wire inside the AOM driver box for supplying DC voltage was disconnected.
I fixed it by soldering, then it was conducting.
Also I soldered the power supply cable with 2 capacitors.
I forgot to take pictures.
I will upload them tomorrow.

We have one flat silicon mirror which will be used for measurement.
It has a wedge, 50arcmin.
On the other hand, fused silica input and output mirrors' wedge are 30min.
We have to take it into account to make a mirror cap.

Aritomi and Yuhang
We measured the green power reflected from the quadrant(used for the AA system). The incident power is measured as 1.2mW while the reflected power is measured as 0.46mW.
Almost 40% of power is reflected from it.

I took a picture of the green beam shape from the filter cavity transmission more than one week ago (11 days ago). Shown in attached figure 1.
I took the same picture last Thursday after the flatting of green beam height. Shown in attached figure 2.
I also took the same picture again today after correcting the beam cut issue on AOM/iris. Shown in attached figure 3.
I think it is clear that the filter cavity transmission has less astigmatism.

[Aritomi, Yuhang]
Current mode matching is as follows. Mode matching is around 90%. TEM00 is fluctuating a lot due to alignment drift of suspended mirrors and it's difficult to improve the alignment without auto alignment.
Mode | AOM frequency (MHz) | IR transmission |
TEM00 | 109.03593 | 2400 |
HG10 | 109.43126 | 180 |
HG01 | 109.43207 | 250 |
IG02 | 109.82923 | 115 |
offset | 94 |
Then we checked IR reflection and it was cutted. We measured BAB reflection from filter cavity when BAB is on/off resonance.
Reflectivity when BAB is on/off resonance is 44% and 77%. "real" cavity reflectivity which is ratio of these reflectivity is 57% and this is too low compared with ~80% in paper we published last year.

[Eleonora, Matteo]
MODULATION
Excitations are sent in pitch
INPUT MIRROR: 15.5 Hz
END MIRROR: 18.5 Hz
Amplitude: 10000 count
Both lines are well visible in the transmitted power. (Pic1, bottom)
DEMODULATION
Before the demodulation the transmitted power is filtered with a resonant filter at the modulation frequency (see pic 2-3)
The demodulation phase is chosen by looking at the transfer function between the injected line (seen on the oplev) and the transmitted power. This phase is ~12 deg for both input and end (pic 4-5)
The demodulated signal is filtered with a first order low pass (simple pole) at 0.3 Hz.
The spectra of the demodulated signal after and before lowpass look as expected. (Pic 1 top. Blue and Red)
ERROR SIGNALS
Error signals are very noisy (see some examples in the attached pdf). To investigate their goodness we have misaligned in turn the input and end mirror of a known amount of counts and measured the change in count of the two error signals. The results are not very clear, nevertheless we have tried to compute a sensing matrix.
demod INPUT | demod ERR | |
input misaligned | -300 | 1500 |
end misaligned | -700 | 13000 |
It seems quite unbalanced and not very reasonable. Another suspicious thing is that error signal behavior is not very reproducible.
It seems they are affected by some variables which we are not considering and controlling.

Yuhang and Aritomi
Power reference form
GR power before green EOM | 200mW |
GR power before AOM | 36mW |
GR power before MZ | 152mW |
GR power after AOM | 12.6mW |
IR power before ND filter | 20mW |
IR power after ND filter (before IRMC) | 4mW |
IRMC transmission | 1.7mW |
We set the power after AOM by changing the AOM modulation depth. The maximum RF signal we should give to AOM is 30dBm(1W). The signal is generated from a signal generator then pass through a RF amplifier(amplify 32dB). So the maximum RF signal power we should generate should be -2dBm. Then we measured the GR power after AOM while changing AOM modulation depth(see attached form).
AOM modulation depth | Green power (injected to FC) |
30dBm | 28.2mW |
29dBm | 28.5mW |
28dBm | 27.8mW |
27dBm | 26mW |
26dBm | 23.5mW |
25dBm | 20.7mW |
24dBm | 17.8mW |
23dBm | 15.11mW |
22dBm | 12.65mW |

[Aritomi, Yuhang]
We recovered IR flash and alignment. However, when IR is aligned, IR reflection seems a bit cutted. We measured BAB reflectivity from filter cavity. Filter cavity was not locked with IR. IR injection is 343uW and reflection is 286uW. Reflectivity is 83% which is lower than before.
Current mode matching is as follows. Mode matching is 1706/(1706+343) = 83%. Pump green power is 60mW and green phase is scanned.
Mode | AOM frequency (MHz) | IR transmission |
TEM00 | 109.03549 | 1800 |
HG10 | 109.43087 | 200 |
HG01 | 109.43131 | 300 |
IG20 | 109.82841 | 125 |
offset | 94 |

I tried to solder the cable to another AOM driver.
But the wire inside the box was disconnected for some reasons.
Tomorrow I will repair this.

I soldered the twisted cables as power supply for AOM driver with two capacitors, 1nF and 100nF, which are implemented for high frequency noise reduction.
Then I connected it to DC power supply and added the DC voltage, 28V.
However, the AOM did not work properly at first.
This was due to the current limit of DC power supply was not enough high.
I increased the current limit, then the AOM driver worked properly.
Then I injected the laser light into the AOM and I could see the diffracted beam as shown in 2nd picture.
I have not yet tuned the shifting frequency, alingment, and so on.
Anyway, I could confirm the AOM can diffract the laser beam.

[Yuhang, Yuefan, Matteo, Eleonora]
Motivated by the transmission issue reported in entry # 1672 we have investigated the goodness of green beam alignment and matching into FC.
First, we have checked that the beam was centered on both input and end mirror. To do this we have steered the beam with the BS on each side of the mirrors in order to make it more visible since it was hitting the suspension frame. We have recorded the BS position for the beam at the two sides and choose the middle point at a reference position. We confirmed that the beam position was already good even if the precision of this measurement is limited to, let's say, ~5mm. This was done to center the yaw, but it also allowed to assess the pitch centering while the beam was on a side. We could also have done the opposite and steer the beam from top to bottom with BS using the mirror top and bottom frame as reference.
Note that for the end mirror we could reach only the right (back toward input mirror) leg of the suspension, while the beam was hitting the pipe before reaching the left leg. So we used the earthquake stoppers as side references. The camera at the end was aligned to make the beam centered in this configuration. Actually, it was already in a good position.
Then, we scanned the cavity after having aligned it at our best. We kicked the end mirror by applying a large offset to the coils and check with the length oplev the excitation of the pendulum motion.
Pic 1 shows the flashes in transmission (top) and the length op lev (bottom). As expected the height of the peaks is smaller and the density is higher when the cavity is moving faster. Note that there seems to be a delay of the oplev signal wrt the transmission one. Actually they arrive from the end room with two different fiber system (old tama fiber system for oplev and new fast optical fiber for transmission)
We zoomed in a region when the velocity is constant so that it is easier to identify FSR. (See pic 2). It seems that HOM are reasonably small (the fact the higher peak is TEM 00 is confirmed by the flashed in the camera).
We have misaligned on purpose the pitch and observed that the HOM become higher as expected. (Pic 3-4). I think this measurement confirms that we don't have a major problem with matching and alignment that can justify a drop of the transmitted power to half of the previous power. Anyway a more refined measurement, with the identification of HOM would be nice.

We reconstructed the history of the cavity transmission.
At the beginning of June we locked again the FC, after more than 6 months (entry #1385), at that time the injected power was 34 mW and the PD in transmission read a bit more than 4 V (which corresponds to ~ 7000 counts). (entry #1404)
After then we decided to reduce the injected power to 12 mW and the transmitted power was reduced accordingly ( ~2300 counts). (Not reported on logbook!)
Recently Yuhang found that by reducing the gain of the locking servo the transmission could reach 5000 counts. (entry #1644) We didn't remeasure the power at that time.
Few days ago, after some realignment work on the bench (entry #1647), we found out the transmission was back to 2200 counts. (entry #1655). In that occasion (even if it is not reported on the logbook) the AOM was also realigned and this brought to a large increasing of the power injected into FC (30 mW). The power was reduced by tweaking the half waveplate after the main laser.
Last Thursday (26/09) we spent half a day to carefully check and asses the alignment level (which is reported in entry #1674). The conclusion of that work is that the cavity seems well aligned.
It seems to me that this mysterious change in the cavity transmission is connected to the change in the injected power, which is largely affected by AOM alignment condition. Also, the only straightforward reason why a reduction in the locking servo gain should bring an increasing of the transmitted power is that the loop was oscillating, and this is likely to be due to an increased power.
Anyway, I think that from now on we should better monitor (and report on the logbook) the level of power injected into FC and the cavity transmission.