Following the cleaning of the SHINKOSHA evaluation plate with first contact, I performed absorption measurements in XY,YZ and XZ planes at the same location with the previous measurements.

The results are presented in the first 3 figures.

Using the calibration computed just before the measurement and reported in entry 2510 together with the power measurements : Pt = 2.764 W and Pin = 3.193 W.

Without any fit we got :

I think that a more precise estimate of the XZ and YZ planes measurements could be done by only considering the data inside the sample.

For instance, using the equation 3.19 of Manuel's PhD where the effective thickness of the sample is computed.

Note that I checked that the lockin was not saturing before starting the XY map. However, after the last measurement in the YZ plane I found out that the lockin was saturated...

I'm wondering if it arised because of point defect/ new dust... Anyway I started a new XY plane measurement after changing the lockin gain. Sadly, the PCI computer got a windows update and stopped this measurement...

Marc, Yuhang, Michael

We measured the OPO nonlinear gain vs input green power.

The measurement was performed by modulating green phase at about 1 Hz and looking at the BAB transmission from the OPO. The signal was triggered to maintain its position on the oscilloscope window and the "persist" function was used to keep the resonance peak on the screen. The peak oscillates up and down as a consequence of modulating the green phase. The maximum value of the peak represents the amplification and the mimnum value represents the deamplification. By plotting max and min gain vs green power we can find the OPO threshold power shown in figure 1.

The fit looks a bit odd, and is quite imprecise on the deamplification fit. I found it especially quite difficult to discern what the minimum deamplified power was on the oscilloscope - the persist function of the oscilloscope was used to keep the mimnimum and maximum values of the resonant peak visible, but it also oscilalted slightly on the frequency axis, which obscured the minimum values visible.

At face value, the nonlinear gain seems slightly reduced from the previous value of 80.56 +/- 0.14 mW (Yuhang/Aritomi thesis)

Marc, Michael, and Yuhang

When we want to lock filter cavity today, we found it was harder to lock even with filter cavity z correction loop on.

To understand why this happens, we took measurement of oplev signal and checked weather information.

1. Oplev signal. In the attached figure one, four suspended mirrors oplev spectrum is shown. We can see the mirco-seismic noise between 0.2 and 0.8Hz is increased by a factor of 5 for BS/Input/End.

2. Tide/seawave/wind information: We checked Yahoo tenki, as shown in the attached figure 2,3,4,5, the tide is small tide, the sea wave height is around 3m, the sea wind is around 12.5m/s, the ground wind is around 5m/s.

Attached figure is theoretical CCFC FDS curve with optimal detuning (54Hz). With the following parameters, frequency at which the anti squeezing crosses shot noise is 44Hz.

sqz_dB = 10.5;                    % produced SQZ (dB)

L_rt = 120e-6;                    % FC losses

L_inj = 0.35;                     % Injection losses

L_ro = 0.24;                      % Readout losses

A0 = 0.05;                         % Squeezed field/filter cavity mode mismatch

C0 = 0.05;                         % Squeezed field/local oscillator mode mismatch

ERR_L =   1e-12;                  % Lock accuracy (m)

ERR_csi = 30e-3;                  % Phase noise (rad)

phi_Hom = [0/180*pi, 30/180*pi, 60/180*pi ,90/180*pi]; % Homodyne angle (rad)

det = -54; % detuning [Hz]

According to elog2514, the current CC detuning should be ~72Hz and we have to change it by 18Hz to have the optimal detuning. Using the formula in elog1727, the CC PLL frequency should be changed by 2*18/1.91 = 18.85Hz. Since the current CC PLL frequency is 6.99701252 MHz, the optimal CC PLL frequency should be either 6.99703137 MHz or 6.99698367 MHz. By checking the CCFC error signal, I confirmed that 6.99703137 MHz is the correct one (In DDS, 6.99703139 MHz was set).

Here is the new CC PLL setting. I saved the DDS setting as "20210520_dds3_CCFC_check" for characterization of CCFC error signal and "20210520_dds3_CCFC_FDS" for CCFC FDS measurement.

Fig 1,2 show the measured CCFC error signal and locking accuracy, respectively. The CCFC calibration amplitude is 182mVpp. Now the CC detuning is 60Hz.

[Aritomi, Yuhang]

We changed the filter setting for CCFC to remove the peak at 170Hz. The new filter setting is 30Hz LPF and gain of 1000.

Then we measured FDS with CCFC (attached figure). The peak at 170Hz disappeared with the new setting.

Unfortunately, I couldn't find the anti squeezing quadrature. We will measure it again.

Degradation parameters:

sqz_dB = 10.5;                    % produced SQZ (dB)

L_rt = 120e-6;                    % FC losses

L_inj = 0.35;                     % Injection losses

L_ro = 0.24;                      % Readout losses

A0 = 0.05;                        % Squeezed field/filter cavity mode mismatch

C0 = 0.05;                        % Squeezed field/local oscillator mode mismatch

ERR_L =   1e-12;                  % Locking accuracy (m)

ERR_csi = 30e-3;                  % Phase noise (rad)

Marc, Michael, Yuhang

In the past, we usually check the mode matching between filter cavity (FC) reflection and homodyne LO without considering the beam jittering. However, FC reflected beam jittering is an issue which degrades homodyne detection efficiency.

To check this issue, we first lock filter cavity with green with AA/pointing/length control loops on. Then we half-detune BAB and check its spectrum on oscilloscope when it arrives AMC. Due to jittering, there are peaks going up and down in the AMC spectrum. We used oscilloscope persisit function to record the spectrum for about 200 seconds (as attached figure). In this situation, we measure the highest value of these peaks.

The peaks height are

All the peaks in the above table are taken in the same manner. But we firstly took TEM00, then we zoomed in and checked higher order modes. Since they are taken in the same manner, we do division as (all HOMs)/(all HOMs+TEM00) = 10.57%.

If we use this value, the mode mismatch in homodyne detection will be 0.8943*0.8943 = 80%. Considering the optical loss in elog2511, the total optical losses will be 1-0.8*0.807*0.904*0.99 = 42.2%. This value is larger than the optical losses we used in PRL paper, but closer to the derived optical losses from SQZ/ASQZ measurement in this link. However, it is noted that the evaluation of mode mismatch in this entry should be a pessimistic one. Because we take the highest HOMs, which only tells us the worst mode matching in the 200s measurement. We also conceived to take many instant AMC spectrum of FC reflected BAB, which should give a more reasonable evaluation.

In addition, we can also use visibility measurement to double check the mode mismatch induced homodyne in-efficiency.

The mode-mismatch was slightly over estimated as I divided by tem00 power and not total one...

The corrected values are : misalignment = 3.9%, mode-mismatch = 1.2% and total 5.3%.

I measured FDS with CCFC (attached figure). The DDS setting for FDS with CCFC is saved as 20210519_dds3_CCFC. During this measurement, CC2 input test mass feedback was engaged with gain of 3.

There is a large peak around 170 Hz introduced by CCFC. I will try to change the filter for CCFC if we can remove the peak.

Unfortunately, the homodyne angles in this measurement are in the wrong side. So I will measure again with correct homodyne angles.

Marc and Michael

We measured the beam profile after the mode matching telescope simulated in 2486 and placed as shown in 2501. The measurements are plotted in figures 1 and 2. Figure 2 includes an outlier, perhaps the distance was recorded incorrectly? Figure 1 shows the result without the outlier. Either way, it seems the beam size and position is close enough to the prediction that fine tuning can be done using the lens rails of the mode matching telescope. Note that in the OPO replacement measurement, the beam will be shifted by a periscope after the f=75mm and then the OPO will be mounted on a rotation stage.

Before this measurement, we checked the alignment of the beam path. It is well centered on the steering mirror before the f=75mm. Without the lens, the beam goes along the screw holes on the table. However, when the lens is placed, the beam diverts about 1cm towards the edge of the table over a distance of about 25-30 holes. It diverts in the same direction when the lens is flipped.

Also I made a slight bump on the EOM rotation stage adjustment knob when handling the beam profiler cable, but it doesn't seem to have affected the beam path.

[Aritomi, Michael]

This work is done on 20210518.

We decreased the pump green power from 40mW to 20mW for CCFC. First, we checked the nonlinear gain with 20mW green.

The measured nonlinear gain is 5 with 20mW green. When we assume the OPO threshold is 80.6mW, the theoretical nonlinear gain is 4. It seems that the OPO threshold is lower than 80.6mW now.

Then we measured CCFC error signal and locking accuracy (Fig 1,2). The CCFC calibration amplitude is 182mVpp.

I found that a huge offset is sometimes injected in Z correction and the Z correction unlocks due to it. I also found that the huge offset appears when I touch the LEMO cable between rampeauto and SR560 in Z correction loop. I replaced the LEMO cable and the problem was solved.

Aritomi, Michael, Yuhang

We tried to optimize filter cavity reflection mode matching to homodyne LO. Now it is about 5%, which is still larger than last year.

After that, we tried to measure FDS. Considering a reasonable degradation budget, we did a fit for all the measurements.

We found ~3.3dB squeezing above rotation frequency, but almost no squeezing was observed below rotation frequency.

The degradation parameters:

squeezing level: 11.2dB

round trip loss: 120ppm

total optical loss: 38%

Matching to filter cavity: 5%

Matching to local oscillator: 5%

Locking accuracy: 5e-12m

phase noise: 30mrad

Marc, Michael, Yuhang

Yesterday we started characterizaton of the fds degradation budget,

PROPAGATION LOSSES :

First we checked the propagation losses by measuring the BAB power at various places on the bench :

after the waveplate just after the OPO : 462 uW

before homodyne : 369 uW

The ratio of these powers gives an overall propagation loss of 19.3%. This value is compatible with previous measurements.

We also compared at the edge of the bench before injection ( P=437 uW ) and just after reflection (P=374 uW ). This ratio gives the in vacuum propagation loss as 14.4%. Note that the BAB was not resonating inside the FC for this measurement.

We checked at the edge of the bench in reflection ( P = 368 uW) and just before homodyne ( P = 363 uW ). This gives losses on this part of 1.4%. There is only 5 optical components there and better quality ones are already bought.

We also checked the power after the waveplate after the OPO ( P = 462 uW ) and just before injection at the edge of the bench ( P = 446uW). This gives 3.5% of losses.

BAB/FC MODE-MATCHING :

We locked the FC with green and tuned the AOM frequency around the various resonances of BAB.

We placed a photodiode just before the one used for CCFC in order to get a larger gain.

TEM00 was scan with AOM speed 60mHz and deviation 6kHz while TEM01/10 and LG01 were scanned with same speed and deviation 600Hz. Taking into account these factors we can calibrate these signals with calibration = 2*deviation / (1 / (2/speed)) = 2 * 2 *speed  * deviation where the first '2' comes from green to IR conversion and second one to get the half-period of AOM scan.

In figure 1 you can see these 3 scans.

The mode-matching was estimated by computing the area under each curve after removing the offset (90 counts).

It gives 4.1% misalignment and 1.3% mode-mismatch and overall value of 5.4%

ROUND-TRIP LOSSES :

We did as reported in the RTL estimation paper (namely switch on/off resonance of BAB).

We got the results attached in figure 2.

In addition to the mode-matching, we also assumed 8% of RF sidebands power and 1% lost due to laser fluctuations.

It gives round-trip losse of 116 ppm in good agreement with previous estimation.

It seems that the bulk calibration was overestimated. This is especially apparent when computing its transmission that was 45% instead of the expected 55%.

I performed again the bulk calibration and got :

AC_bulkref = 0.062;
DC_bulkref = 4.14;
P_in = 26.4e-3;
P_t = 13.1e-3;
T_bulkref = P_t/P_in
abs_bulkref = 1.04;
R_bulk = AC_bulkref/(DC_bulkref*sqrt(T_bulkref)*P_in*abs_bulkref) = 0.7743 W/cm

I used this new calibration to compute again the absorption map of the sample (see the 3 attached figures.

In the figure, the absorption is extracted from a fit using 2 normal distributions.

Here I also add the overall mean and standard deviation of each map (ie without any fittting) :

I created a page in our wikipage to share information of the old measurement of optical losses and phase noise for FIS and FDS.

https://gwpo.mtk.nao.ac.jp/wiki/FilterCavity/losses%20and%20Phase%20noise

Mode mismatch between filter cavity and LO was found to be relatively high. And this results in homodyne detection effeciency to drop by about 13.3%. Together with the bad mode matching inside filter cavity reported in elog2503, we could explain worse FDS measurement.

As shown in the attached two figures, the TEM00 peak is 1.19V while the LG01 peak is 0.104V. This corresponds to 8.0% mode mismatch. Note that this spectrum is after the optimization of alignment and filter cavity half-detuned.

We have tried to reduce this LG01 mode by moving the mode matching lenses. However, the mode matching can be barely improved.

To search for the reason of this mode mismatch, we checked the beam position on every in-air optics, we found no clipping issue.

We have also tried to measure the power loss, the total power loss from after the in-air Faraday to before homodyne is about 19%. This is in-agreement with the old measurments.

For the in-vacuum part, we could try to scan the injection steering mirror yaw or pitch slightly and see if there will be a clear power drop. We will try with this method to check if there is in-vacuum clipping.

If there is not clipping found, we need to first understand why this could happen. In principle, for our optical system, there should not be such large mode matching change. In the worst case, if we couldn't figure out what is causing this problem, we will need to measure the beam parameter again and redesign the telescope.

I measured CCFC error signal with 20% pick off. The CCFC calibration amplitude is 452mVpp. Fig 1 shows the CCFC error signal with different demodulation phases. The CCFC error signal agrees well with theory, but the CC detuning changed by 20 Hz from the previous measurement. This means that the filter cavity length changed. The CC PLL frequency can be written as follows:

CC PLL frequency = 14*FSR + CC detuning

From this formula, the CC detuning change of 20 Hz corresponds to the FSR change of 1.4 Hz and the filter cavity length change of delta L = delta FSR /FSR * L = 0.8 mm. We need to tune the CC PLL frequency.

Note that I fixed the mode mismatch between OPO/FC to 6% in the calculation.

Then I locked CCFC. The filter setting is gain of 10000 and LPF of 0.03Hz. The CCFC can lock only for a few minutes due to the CC1 saturation.

Fig 2 shows the IR locking accuracy with/without CCFC. Now the IR locking accuracy with CCFC is 1.2 Hz and the high frequency noise shape looks similar to the best locking accuracy we obtained on 20201211. I compared the IR locking accuracy on 20210517 with the one on 20210514. The difference of these is whether the laser is kept on for more than one day or not. It seems that keeping laser on makes the high frequency noise better.

I used strip tool to check how long time AA/pointing loop needs to use to go from unlocked point to locked point. This tells us rouhgly the bandwidth information of these loops.

The AA loop filter and gain are as following:

Note that these filters and gain are not optimized yet. The time to go from unlock to lock is shown in the attached figure 1. We can see it took about 5 second, which means the bandwidth is about 200mHz.

The beam pointing loop filter and gain are as following:

Note the pointing loop gain is not optimized yet. The time to go to the good point is shown in the attached figure 2. We can see it took about 2 min, which means the bandwidth is about 8mHz.

The absorption distribution is fitted with 2 normal distributions.

I thought it could be useful for the case of XZ and YZ maps (where there are measurement points outside the sample) because it allows to remove the effects of absorption outside the sample and point defects/dust on the surface.

But I agree that it might not be the most suitable distribution, especially for the shinkosha samples...