NAOJ GW Elog Logbook 3.2
I started some analysis for finesse of the folded cavity.
The finesse was obtained by fitting the transmitted signal.
The finesse at room temp. was about 1.678*1e4(+/- 11) though that at 8 K was 1.579*1e4(+/- 26).
Further analysis is ongoing...
Aritomi and Yuhang
There was END mirror pitch/yaw coupling problem found in AA sensing matrix, as reported in elog2165. Besides, it was found the END mirror driving from coil/magnets has already been not clean, as reported in elog2173.
So we decide to find a good driving matrix for END mirror. Before doing that, the check about if the END mirror PSD is sensing well the pitch/yaw was done.
The method is to excite yaw of END mirror and check the response of signals in pitch/yaw channel relative to this excitation. If the sensing is not good, the resonant peak will go from yaw channel to pitch channel.
The test result is attached. By comparing the resonant peak at 1.58Hz, the sensing coupling was found to be around only 1%. This means that sensing coupling of oplev is not a problem.
What I did
- Ringdown and doppler measurements @room temp.
- Open loop TF measurement of PDH locking loop
Results
The obtained data are still in USB and floppy disk...
But the UGF of TF was about 3.1 kHz and phase margin was much larger than 30 deg.
Therefore, the lock seems to be stable one.
Next
- Cooling down to 10 K
I am planning to cool down the cavity to 10 K which is the designed temperature for the ET-LF.
And then monitor the finess in order to estimate the optical loss at 10 K.
[Aritomi, Yuhang]
We found that FC misalignment changes the shape of CCFC error signal and/or adds offset. This can cause the detuning fluctuation and we think that auto alignment is necessary to obtain the stable detuning fluctuation.
We measured CCFC error signal again (attached picture). CCFC error signal today is larger than last week at 100Hz-10kHz region. This may be related to the fact that we turned on the lasers today.
What I did
- Ringdown and doppler measurements
- Injected He gas to raise the temperature
Results
The attached file shows the very preliminary results of the ringdown measurements.
As shown in the figure, the finesse at 134 K, and 165 K is smaller than that of 170 K (judged by my eyeballs).
This might imply that the molecular layer i.e., amorphous ice can desorb around between 165 K and 170 K.
It should be noted that the fitting has not yet done...
Next
- TF measurement of PDH loop
Tomorrow, I will measure the TF of PDH locking loop.
Then I will re-cooling the cavity.
At this moment, the target temperature is 10 K and monitor the behavior of finesse in order to estimate the loss induced by cryogenic molecular layer.
I think this result would provide some implications to the Einstein Telescope.
I found one of the integrator was not working due to the insufficient soldering.
I remedied it, then the TF of the servo seems what I intented.
However, I found that the gain was too high for stable lock...
I will modify the servo or locking scheme.
On Friday I modified one of the two QPD demod box adding high pass filter (SHP-50+) and amplifier (ZKL-1R5+) for each QPD segment channel (picture1). See PDF pag.2 for current components of the demodbox. The second QPD demodbox is not yet cabled (picture2).
[Aritomi, Yuhang]
We replaced a flipping mirror for CCFC with flipping 50:50 BS. We originally had 36% loss and we have 50% additional loss from BS. The total loss will be 68%.
We measured FDS with CCFC lock (attached picture). We also tried to measured anti squeezing, but we cannot have anti squeezing larger than ~11dB so it is not shown in this plot. It seems the nonlinear gain is not optimized.
For HOM angle -1,-14,-28 deg, detuning is around 60Hz and more or less stable, but for HOM angle -38 deg, detuning is 44Hz. I'm sure that green alignment is fine for HOM angle -1,-14,-28 deg, but not sure for HOM angle -38 deg.
We will measure FDS again and check green alignment at the same time.
In this measurement, we assumed that green power is 56mW and generated squeezing is 21 dB, but nonlinear gain was not optimized in this measurement and actual nonlinear gain (or generated squeezing) should be lower. I assume that generated squeezing is 16dB in this measurement and fitted the FDS measurement again (attached picture). In this case, homodyne angle changes from 0 deg to 90 deg and detuning is between 89-98 Hz. The detuning fluctuation might be better with CCFC.
Aritomi, Matteo and Yuhang
Matteo implemented the RF amplifiers and filters.
The comparison of each segment is attached. Two orders of magnitude amplification is observed at low frequency.
This file is saved in DGS system in Desktop/AA with name WFS_amplify.
Just a reminder, sometimes we still suffer from the unlock of CC1 loop. This is mainly due to the saturation of CC1 correction signal. We need to investigate how to improve this.
I plotted the obtained ringdown data at room temp., 120 K, and 8 K, respectively.
The results shown in this entry are preliminary though, the decay time at 8 K is less than that at room temp.
It should be noted that the measured data at 120 K was 3days after the cavity reached 120 K.
I will analyze the optical loss of the folded cavity.
In addition, I am planning to compute the dynamical response of the cavity.
[Matteo, Eleonora]
In order to transform one of the broken TAMA demodulation board into a simpler phase shifter (to be used for CCFC lock) we removed the mixer and shortcut the output of the phase shifter to RF IN.
We measured the trasfer function between LO in and RF IN which is now an output. See attached picture. The TF is in agreement with what we expect from the phase shifter data sheet (SPH-16+).
[Aritomi, Yuhang]
Carrier and CC AOM frequency are as follows.
We changed a cable length for CCFC LO and checked the CCFC error signal. The result is as follows (Pic.1,2,3 are 0,2,4 m cable length).
| CCFC LO cable length | phase |
| 0 m | 0 deg (I phase) |
| 2 m | 90 deg (Q phase) |
| 4 m | 180 deg |
Then we added the CCFC I phase error signal to perturb of green FC servo. We used SR560 with lowpass filter 0.1 Hz and gain 200 before injecting the perturb. Then CCFC stably locked!
When SR560 gain is 1000, it oscillates with 88Hz (Pic. 4).
We measured CCFC phase noise. For CCFC calibration, we used Pic. 5. Since AOM scan speed is 1600 Hz/s, the calibration factor (Hz/V) is 1600 Hz/s*70 ms/40 mV = 2800 Hz/V. Measured CCFC phase noise is shown in Pic. 6. Note that free run CCFC error signal is out of linear range and the spectrum could be wrong. I also compared with IR locking accuracy we usually use (Pic. 7).
We measured the ratio of CCFC OLTF and green OLTF with SR560 (Pic. 8). The crossover frequency is 40 Hz.
When the CCFC demodulation phase is optimal, CCSB frequency separation can be obtained from the distance of two dips in Q phase signal since the two dips correspond to CCSB resonance (Pic. 9). From Pic. 2, time difference of the two dips is about 120 ms and therefore frequency separation of CCSB is 120 ms*1600 Hz/s = 192 Hz. This is a bit larger than optimal value 108Hz. Note that when the CCFC demodulation phase is not optimal, it will also change the distance of the two dips.
Just a reminder, sometimes we still suffer from the unlock of CC1 loop. This is mainly due to the saturation of CC1 correction signal. We need to investigate how to improve this.
I attached OLTF of CCFC and green lock. Note that I flipped the sign of measured data to match the measurement and theory. The measured phase is not consistent with theory.
What I did
Today, I tried to measure the finesse of the cavity by locking the laser.
However, I could not lock and found the servo had some problems at low-pass parts.
I tried to remedy the circuit but could not solve the problem.
So I decided to make another circuit.
Actually this work has not finished yet...
Besides it, I did some measurements without locking the laser, i.e. doppler method which depends on the dynamical response of the cavity.
Also I measured the frequency split between transmitted p- or s- polarized beam.
As shown in the attached picture, there is a frequency shift between the s- and p- polarization.
The injected beam is pure s-polarized beam, but by playing with a HWP one can inject p-polarized beam.
This frequency spilt measurement corresponds to a cavity enhanced ellipsometry.
Results
To be honest, the analysis has not yet done...
The frequency spit was about 150 usec by 100 Hz 3 Vpp triangle wave.
Notes
The temperature of the cavity was about 120 K.
At first, the reflected (and transmitted) power was about 70 % less than I expected.
I doubted that the misalignment, but it seemed not.
Though I optimized the alignment, the reflected power did not improve.
After one or two minuites illumination, suddenly both of them recovered.
I could not figure out the reason, but the desorption of moleuclar layer is one of the possible reason.
Further investigations are needed.
I found one of the integrator was not working due to the insufficient soldering.
I remedied it, then the TF of the servo seems what I intented.
However, I found that the gain was too high for stable lock...
I will modify the servo or locking scheme.
[Aritomi, Yuhang]
We replaced a TAMA mixer with a new mixer (ZX05-1L-S+) and a low pass filter (SLP-1.9+) for CC1. CC1 error signal with the new mixer and TAMA mixer are shown in Pic. 1,2. SNR with the new mixer seems much better and there is no offset in the new mixer. Note that CC1 servo offset should be 5.
Then we locked CC1. CC1 gain is 2 for 25mW green. Pic. 3 shows CC1 OLTF. UGF is 3.7 kHz and phase margin is 50 deg. Pic. 4 shows CC1 phase noise. CC1 rms phase noise is 84.2 mrad and squeezing phase noise from CC1 is 84.2/2 = 42.1 mrad.
A similar test for input mirror with elog2173 was done.
| pitch to yaw coupling | 1.8% |
| yaw to pitch coupling | 1.7% |
Almost no coupling was found. This is consistent with the result reported in elog2172.
This entry is just a memomrandum.
Last week I made a sum-amp circuit to add the offset to PDH feedback signal.
But the output from the circuit did not contain the input signal, i.e. insensitive to the input signal and could not use though I confirmed all the conduction was fine.
Actually, there was a problem at one op-amp port labeled 5 on the board.
By changing the port, this problem was solved, but what caused this was uncertain...
One possible reason is due to the poor soldering around the op-amp but not sure.
[Aritomi, Yuhang, Eleonora]
First we amplified CCFC RF signal by 33.6dB with RF amplifier port which was used for AOM before. We put a DC block (BLK-89-S+) before the RF amplification. The result is as follows.
We demodulated it with a new mixer and measured CCFC signal when CC1 is scanned and CCSB are off resonance. The signal is 120 mVpp and no offset (Pic. 1).
Then we locked CC1 (CC1 gain: 0.5 for 56 mW green) and changed PLL setting as before. Carrier and CC AOM frequency are as follows. Frequency difference between carrier and CC is 60Hz which is good. We scanned AOM around CC resonance. The CCFC error signal seems I phase (Pic. 2). Theoretical CCFC error signal with 56mW of green is Pic. 3.
We tried to change the CCFC demodulation phase by changing 14MHz DDS phase, but CCFC error signal didn't change. Since changing 14MHz DDS phase changes both CC1 LO and CCFC LO, this may cancel out the phase change.
So we tried to change the CCFC LO demodulation phase only by adding a 2.5 m cable to CCFC LO. Then CCFC error signal became Q phase (Pic. 4) although it is not crossing 0. Theoretically 90deg phase delay of 14MHz signal should be 3.5 m (lambda/4 = c/sqrt(2.3)/14MHz/4) where 2.3 is the relative permittivity of polyethylene. There is a bit difference between theory and experiment. We need a phase shifter for fine tuning of CCFC demodulation phase.
Anyway it seems we can use this error signal to lock FC.
