Akutsu-san, Marc, Matteo

In agreement with Akutsu-san we plan to use the ATC clean room to assemble the new OPO.

Fortunately, there is an IR laser installed on one corner that we can use. After the laser source, there are some optics (Faraday Isolator, lenses) so that the beam should be collimated.

The procedure to turn on the laser is the following :

- Connect power supplies

- Turn on the +5V supply (largest black button)

- Press 'stand-by' button on the laser source to get out of the standby mode

There are few components around this laser setup that we can displace for the assembly (scale, forks, pillar) so that with 2 steering mirrors we have enough space to use this laser.

PS :

- We have to bring new clean suits and we can also bring the old ones of this clean room for the next time we clean the clean suits

- we'll need a periscope, beam profiler, mounts, forks, few lenses, safety glasses, etc ...

Marc, Matteo

Last Friday I flipped the surface reference sample and tried to further tweak the alignment without much improvement.

Yesterday, we decided to measure the beams profile.

Unfortunately the automated LabView program had some error and we had to do measurement one step at a time...

I installed first the razorblade vertically (2 cm in front of the last hole of the translation stage toward the imaging unit) and scanned both the red and IR beams on the Y/Z axis.

[Note that connecting the DC output of the powermeter went smoothly despite troubles to read the powermeter through the other cable]

Then I turned the razorblade horizontally (54mm above the translation stage) and scanned both the red and IR beams on the X/Z axis.

Actually the razorblade cut the beam not on the razor edge... meaning that we have to take into account the 28mm width of the blade.

I finished to write some MATLAB code to automatically extract all measurement data from yesterday, fit them using a program wrote by Manuel.

But It is still needed to check the screenshots 1 by 1..

Marc, Matteo

Yesterday we removed the first contact applied on KAGRA spare, flipped the mirror and applied first contact on the other surface (Fig 1)

Michael and Yuhang

We replaced the big mirror mount in front of AMC with a compact one. (attached figure 1: the replaced mirror. attached figure 2: the gap between replaced mirror and board)

We have made homodyne DC cable, homodyne power cable, AMC DC PD power cable go through the appropriate hole. (figure 3: extended cable on homodyne power side. figure 4: current cables situation around homodyne)

To close bench, we still need to drill holes on board to let GR inj/ref and IR inj/ref go through (three holes are expected). 

Marc, Michael and Yuhang

We have characterized some signal and noise of CCFC PD in elog2384. However, the pump power was 30mW in that case. According to the simulation and experiment, for the current system, 18mW of pump power would be better. Therefore, we did more test with 18mW pump.

In the attached figure, some signal and noise levels are shown. From this plot, 18mW pump power seems to be not large enough. Indeed, we also tried to close anyway the loop with such small error signal. But we couldn't find a good filter configuration to make the loop closed proporly.

I confirmed the coating of the surface reference sample is on the side with the written part. Therefore the correct orientation is with that side first (i.e. facing the periscope).

The discrepancy between the current and old measurement as well as the slight asymmetry can be caused by not optimal alignment of pump beam.

Aritomi, Marc, Matteo

Today we cleaned a bit the computer room and brought the KAGRA spare to the clean room. We putted first contact on the surface facing outside the protection box.

Then, we followed the realignment procedure described on the wiki and installed the surface reference sample to realign the setup with the goal of reproducing the calibration done in elog 1619 : R_surf = AC_surfref/(DC_surfref*P_in*abs_surfref) = 16.9 [1/W]

First, we putted the surface with the sample name facing the laser (PS : mirror center in z is 38 mm). Results are presented in figure 1 and gives (with abs_surfref = 0.22) : AC_surfref = 0.17V, DC_surfref = 3.91V, P_in= 16mW -> R_surf = 12.4 [1/W]

Then, we flipped the reference sample (PS : new center at 33.25 mm) to check if it could explain the discrepancy wrt to elog 1619.

During a scan, there were some troubles with some connections and we got several error messages from LabView... We had to restart the computer to use LabView again but there is still some troubles with the power-meter ( detected but not able to connect to the computer)

The result of this configuration is presented in figure 2 and gives : AC_surfref = 0.31V,  DC_surfref = 4.01V, P_in= 15mW -> R_surf = 23.4 [1/W]

This time the AC measurement is not as symmetrical as expected so we'll further tune the alignment tomorrow.

Marc, Matteo

We also notice that the readout photodiode is misaligned even without sample.

Michael and Yuhang

Today, we investigated various signal and noise level from CCFC PD. In addition, the servo noise is also added to be compared with signal level.

Attached figure shows their comparison. There are two signals measured in the figure. One comes from picking off signal with beam sampler(~1%), the other comes from BSN11(~3%). The splitting ratio was measured with power meter (offset has been subtracted).

From this measurement, BSN11 would be suitable for the signal pick-off. By using this signal, probably a bandwidth of ~50Hz could be achieved without reinjecting noise. We will try to lock FC with BSN11.

We made further anti-squeezing and squeezing measurement. This measurement indicates optical losses to be 19.1%, phase noise to be 20.5mrad.

Therefore, we still have 3% optical losses which are not known.

The ideal sqz-asqz for current set-up and improved set-up are shown as well.

Marc, Michael, Yuhang

Recently, homodyne alignment was optimized. Frequency independent squeezing was remeasured. From this measurement, 6.5dB squeezing was measured. Compared with the measurement in elog1837, 1dB more squeezing has been achieved. The main difference between these two measurements is the homodyne alignment.

In this measurement, 37mW pump power was used. Therefore, there should be 15dB original squeezing.

If we assume 20mrad of phase noise, 6.5dB squeezing corresponds to 19% optical losses.

If we assume 30mrad of phase noise, 6.5dB squeezing corresponds to 17% optical losses.

The old measurement was indicating 26% optical losses. Therefore, at least 7% optical losses has been reduced.

The known losses are 1-(1-7e-2)*(1-1e-2)*(1-3e-2)*(1-2e-2)*(1-3e-2)*(1-1e-2) = 16%. 7% OPO intra-cavity losses, 1% dichroic mirror, 3% Faraday isolator, 2% mirror and lens losses, 3% homodyne efficiency/quantum efficiency, 1% classicial noise. Therefore, about 1% to 3% optical losses are not figured out. We will make more characterization tomorrow.

Marc, Yuhang

Today we got the FC flashes back.

We acted on all picomotors to realign the beam and remove the OpLev offset.

We then took spectra of the OpLev signals and saw that the End Mirror one showed a broadband increase of the noise floor.

This was due to one of the steering mirror screw being almost loose. Therefore we put it to a better situation and compensated it by acting on the steering mirror on the OpLev laser bench.

We also changed the optical density before the End Mirror PSD as it was previously not so well fixed (fig 1 : current situation)

The current Oplev spectra are represented in Fig2 with green/brown the references and blue/red the current ones :

  -  PR and BS show the high frequency noise increase that we hope to solve by using the KAGRA QPDs

  - End pitch exhibits a new peak on pitch around 6 Hz

  - BS pitch exhibits a new peak on pitch around 9 Hz

Before locking the FC we would like to further tune the OpLev. Indeed, we fixed all their optics but then did not retune the PSD positions nor check the diagonalizations.

These may explain the new peaks on BS and End Mirror as their optics were moved quite a lot

PS : the second target is on remote mode and End mirror picomotors have been disconnected

[Takahashi, Matteo, Yuhang, Marc, Michael, Aso]

We opened the BS chamber and recovered the suspension.

1. Opend the BS chamber from the top.
2. Checked the status of suspension. The gap between the south-west side of IM and the IRM (magnet box) was too small (<0.5mm).
3. Shifted the suspension point of IRM to south-west. The gap became larger (>1mm).
4. Measured the TF of suspension. It was consist with the previous one.
5. Aligned the BS using the IR beam.
6. Closed the chamber and started evacuation. Opend the GV for the TMP on BS after the evacuation with the big RP.

The KAGRA QPDs will be used for the TAMA OpLev.

In order to prepare their installation, I checked the available space for these QPDs.

The most critical one is the Input Mirror ones : there is only a lateral space of 70 mm between the various beams. Also, the longitudinal space between the first QPD and the end of the window optical table is 60 mm.

As the KAGRA QPDs have  horizontal size at the order of 62 mm (and vertical 73 mm) , it will be required to design our own cover box.

For all the others OpLev there sould not be any critical space constraints

We also took some measurement of PD spectrum.

Attached figure 1 shows the measurement of CC1 PD noise after demodulation for different incident power. In this figure, op-amp LMH6624, R1 1.1kOhm, R2 13kOhm are used.

We could see that noise becomes to be shot noise limited when laser power reaches around 300uW. Therefore, I confirmed that electronic noise was still limiting in this case. Then I checked again the simulation, which shows the noise is limited by the resistor. However, according to the simulation, shot noise should start to limit after laser power reaching 3mW.

360mV pk-pk corresponds to -4dBm, after a 21dB amplifier, it becomes 17dBm (50mW). According to the specification of frequency mixer (ZX05-1L-S+), it may have permanent damage if the RF power is more than 50mW. Therefore, we have been already reaching this threshold due to this 160~180MHz oscillation. This seems to be the reason of demodulation problem when (R1=11Ohm, R2=130 Ohm) are used. A filter to remove this oscillation may help to solve this problem.

We checked again the simulation of this PD up to 200MHz (see attached figure 1). It has voltage noise increase around 120MHz, but this peak is not very sharp. Therefore, we still don't quite understand why we have such large oscillation when (R1=11Ohm, R2=130 Ohm) are used.

Marc, Yuhang

Today we pursued the investigation of this photodiode. Especially, we investigated if there is some offset present that could saturate the mixer used for the demodulation.

First (with high resistors of previous entry) we measured an offset of -1.68V.

Adding a DC block reduces it below 2mV.

Then, we replaced the resistors (low in previous entry) :

We measured an offset with mean value -90mV and also a clear frequency modulation around 180MHz and peak to peak around 148 mV (Fig 1)

Adding a DC block reduced the offset mean value to -2mV but the signal around 160MHz had an increased peak to peak amplitude around 360mV (Fig 2)

As there is a 20dBm amplification after the mixer, this high frequency signal is larger than the 14MHz one (at -10dBm) and is close to saturation of the mixer.

We'll try to compare this result with simulation (performed up to 100 MHz for now)

Today, we checked again this moving peak. At the beginning, this moving peak appears to be very close to 14MHz again. But after 10min, on spectrum analyzer, it goes away from 14MHz peak and stay there until almost 30min.

The CC1 PD is used to detect RF signal at 14MHz. While checking this 14MHz signal peak on the spectrum analyzer, I found another peak that is moving in the frequency domain, and its peak height also changes.

This moving peak can reach a level of almost about -63dBm (before amplification). The frequency of this moving peak seems to be +/- 10MHz around 14MHz. Besides, it seems to become even larger when it approaches the 14MHz peak. More importantly, the RF signal from CC1 PD is only -51.3dBm (before amplification), which means that this moving peak may introduce non-negligible noise into the CC1 loop. There is a movie in this link shows the situation.

I also found this peak still exists even when all optical cavities are unlocked. This situation is in the movie in this link (after 20dB amplification). At around 8 second, I switched off CC PLL, the moving peak disappeares. So it seems this moving peak is related with PLL loop.

I also confirmed that this peak is not RF signal cross talk. Because when I block the light incident on CC1 PD, this moving peak disappears.

In the CC1 PD (TAMA 14MHz resonant PD), LMH6624 is being used (this modification was done in 2019). However, a relative large resistor (1.1kOhm) is being used to connect + of LMH6624 and ground. This connection should introduce lots of thermal noise. Therefore, I would like to replace it with a 10Ohm resistor. Accordingly, the other resistor which is used to amplify signal is changed from 13kOhm to 130Ohm.

Before this resistor replacement, I made several measurements as a benchmark. The green pump power used in this test is always 30mW.

Signal from PD: -51.3dBm. Noise from PD: -83.2dBm (SNR: 31.9dB)

SIgnal after amp: -30.0dBm. Noise after amp: -70.4dBm (SNR: 40.4dB)

After demodulation, the signal time-series is measured by oscilloscope. Its pk-pk is 420mV. Thickness of singal line is 76mV. (SNR: 5.5)

After resistor replacement, the same measurements were also performed. And I got:

Signal from PD: -49.1dBm. Noise from PD: -83.4dBm (SNR: 34.3dB)

SIgnal after amp: -28.0dBm. Noise after amp: -77.4dBm (SNR: 49.4dB)

After demodulation, the signal time-series is measured by oscilloscope. Its pk-pk is 88mV. Thickness of singal line is 20mV. (SNR: 4.4) 

(The demodulated signal is a bit strange. Although the 14MHz peak becomes larger but the demodulated signal is smaller.)

It seems the SNR improvement is obvious by replacing resistor. However this improvement is only visible before demodulation. The demodulated signal (checked from oscilloscope) even becomes worse. For demodulation, we upgraded DDS to provide saturated LO for each demodulator/mixer. But maybe we still have some issues about signal demodulation.

I have already changed resistor back to the original situation. And put CC1 PD back.