NAOJ GW Elog Logbook 3.2
What has been done:
1) The area around the work station of the new DGS has been cleaned and tidied up.
2) The cabling toward and from the new ADC and DAC has been realized.
3) The real time model has been completed to include all the suspensions (pic 1). It is mainly fine even if I could not implement some more sofisticated features (see dedicated entry #1359).
4) The damp on BS has been implemented. See TFs of pitch and yaw (pic2-3) and specra with open and closed loop (pic4-5). They can be compraed with the "labview ones", measured in november 2016, reported in entry #337.
Some comments on BS CONTROL
1) TF are both one order of magnitude smaller wrt labview ones, this seems a bit strange to me. I would say It shouldn't depend on the digital system. But note that in KAGRA DGS there is a factor 2 btw ADC and DAC count calibration (entry #1315).
2) Pitch TF is less clean wrt labview one: there is noise and lack of coherence in correspondece of the zero of the TF. Not that the maximum output of DAC is +/-5 V, while in labview was +/- 10 V. The max white noise amplitude I could send to avoid any saturation is 6500 count witch according to the count/ volt calibration corresponds to about 2 V.
3) Spectra levels look similar. A rough computation of the count -> rad calibration (using the information from entry #337 and #1315) gives roughly 1.9e-7 rad/count.
4) There are more structures at high frequency then before. I don't think this corresponds to a real motion of the mirror but rather some oplev mount resonances. It is clearly visible in the time doman where it masks the low frequency motion. This excess of noise at high frequency wrt before showed up also in PR. The gain should be low enough not to feed this noise back to the mirror.
5) There is a high Q resonance at 10 Hz which is effectively damped by the control.This was already observed in labview. Another line is present at 11.5 Hz both in pitch and yaw. I don't think it is from the mirror and I fear the gain is not low enough at that frequency. To be investigated.
In the future, we could turn off ceiling light only to avoid obvious 50/100Hz noise on PD. And keep wall light on to avoid dark.
The characterization of the DDS output was done long time ago. You can find the graph here.
As clear from the schematics there is a high order low pass filter with corner frequency around 200MHz, therefore the cut is expected.
I made the optimization of the telescope from OPO transmission to AMC. The entry is here. From my impression, I feel I move both lenses quite a lot. I just checked the position difference before and after optimization. I also compared the final implementation scheme with the initial design. The result is attached as the following form.
lens1(holes=2.5cm) | lens2(holes=2.5cm) | |
design position | 3.16 | 25.68 |
first-time implementation position | 2.2 | 25.9 |
final implementation position | 3.1 | 23.4 |
1. The position of the lens is relative to 'the dichroic mirror after OPO'
2. I assume that the beam waist inside OPO is 2 holes away from 'the dichroic mirror after OPO'
I am sorry that all the design work of Eleonora Polini should be done again because I moved this lens.
Chien-Ming, Yu-Hang, and Aritomi
We replace the flipping mirror of BAB and CC to a 2" BS today.
As for the p-pol PLL, Yu-hang saw a 0.1Hz slow fluctuation signal at the output of the monitor mixer. The input to the mixer were 252 MHz laser beat frequency and 252MHz DDS channel 2 output. Today, we try to figure it out but fail to see the fluctuation at 252MHz but at 150MHz (We adjust the beat frequency to close to 150MHz).
We use the spectrum analyzer to check the output of the DDS directly. We see a drop in DDS amplitude after 200MHz as shown. The drop after 240MHz is serious and there is another peak appearing. However, we assume that the DDS should be able to provide up to 500MHz.
We feel that the 0.1Hz slow fluctuation may be due to the phase (or frequency) difference from DDS after frequency division. (For locking to 252MHz, the PLL needs to be divided by 3).
The characterization of the DDS output was done long time ago. You can find the graph here.
As clear from the schematics there is a high order low pass filter with corner frequency around 200MHz, therefore the cut is expected.
[Aritomi, Yuhang, Chien-Ming]
We replaced a flipping mirror for BAB and CC with 95:5 BS since repeatability of the flipping mirror is not good. 95:5 BS is CVI PR1-1064-95-IF-2037-C. We put this BS so that 95% of BAB reflects and 5% of CC transmits. Instead of increasing the CC power, we removed ND1 and ND0.4 in CC path before the flipping mirror.
Then we tried to align CC and BAB to OPO. We found the resonance of CC, but the alignment of CC is not finished. We'll finish the alignment of CC and BAB tomorrow.
Chien-ming, Yu-hang, and Aritomi
Today we found power fluctuation after we lock IRMC. Then we did lots of investigation. Especially we put power meter at some position to monitor laser power coming from the main laser. However, as we found before, we didn't see power fluctuation from it.
However, later we used photodetector directly monitor the part of the beam from the main laser. The signal from PD is shown in the attached figure. We could see a clear signal fluctuation.
Then we blocked the light going to SHG, and this power fluctuation disappears.
This proves that it is really necessary to install the second FI for the main laser.
To see the effect of coherent control beam on homodyne noise, I performed the measurement of the noise spectrum with LO or LO+CC on homodyne.
The result is attached as a figure from 10Hz to 51.2kHz. Coherent control beam brings a large 280Hz peak and several small peaks. Also, it rises up the noise level between 10Hz and 40Hz.
There are also videos show spectrum without average. Please refer to LO https://drive.google.com/open?id=12jaWRG_3HjDOSp6thXOixYxd7Y17WkIx, CC+LOhttps://drive.google.com/open?id=1OjRokr1ly4zR8Tf0TltP-oGDvwoo92RT
Notice that the signal of homodyne SUB-DC low-frequency noise spectrum has fluctuation. It means there are noise sometimes large and sometimes small. If we could float bench, it will be better.
Today we used the oscilloscope to monitor the homodyne SUB-DC channel. We can see clearly that the unlock of the IR phase part is related to the jump of the homodyne signal. This jump will rise up the whole spectrum level. The video can be checked from the following link. (blue line: IR coherent control error signal. yellow line: homodyne sub dc signal)
https://drive.google.com/open?id=1RpOabqKvQRVOS4uYonU8hqClGpDq8HDu
We can see from this video that every time homodyne signal jumps when the IR coherent control error signal (blue line) becomes thin.
The thickness of the thin line is comparable with the situation when there is no coherent control signal. So I checked the PLL and the lock of OPO. All of them are fine. The strange thing is this line doesn't drift away but get thinner.
Also:
3. The spectrum is plotted with poor resolution, so another reason could be that at low frequency nearby lines are "grouped" and the baseline seems higher.
Eleonora, Irene, Yuhang, Federico
We made the same excercise as yesterday, but now with squeezing ON
fig.1 is a whole spectrum from 10Hz to 100kHz, comparing the squeezing ON with the squeezing OFF conditions
fig.2 is a zoom from 10Hz to 200Hz
fig.3 is a zoom from 100Hz to 2kHz
(note that an offset has been added to the data with no squeezing in order to have the level of noise 4dB above the squeezing curve)
We see and recognize many peaks with seismic origin, and we know for sure that there is also a large noise associated with acoustics.
13.5Hz is associated with a bench resonance; 24.62Hz is a "scroll" vacuum pump; 33.88Hz is a "moving line" that jumps from 34 to 37 Hz with a cycle of about 20 seconds; perhaps it is connected to an air conditioning machine that "modulates" an air flow to maintain the controlled temperature; 603Hz comes from all Turbo Molecular vacuum pumps in operation at that time. The large noise from 400 Hz to 1200 Hz could be of acoustic origin and could be associated with air flows. This is the region where are the resonances of the various mounts, it is not surprising that they are excited by ambient seismic and acoustic noise.
The possible solutions are the suspension of the bench (seismic isolation) and its complete seal with aluminum and rubber panels 1 cm thick; you should get a reduction from these actions.
I think there are mainly two points:
1. Turn off fan of clean room. Especially it brings noise around 20Hz. It is roughly the corner frequency where our spectrum starts to go up(from 20Hz to 0Hz, spectrum goes up). And also some other noise around low frequency region.
2. Maybe common noise rejection is also better. Everytime we make measurement, we make sure it is good.
Thank you a lot for your work.
I want to know the reason why the homodyne noise is improved from before.
Alignment of homodyne is improved? or something environment is changed?
Eleonora, Irene, Yuhang, Federico
Yesterday we took a high resolution spectrum of the homodyne signal (no squeezing): this is shown in Figure 1.
Above 1kHz the spectrum is quite flat at around -134dBV/sqrt(Hz). Instead below 1kHz some clear structures appear. Below 10Hz the spectrum is rising with no clear structure. In Figure 2 we marked some of these peaks.
Some of these peaks we recognize from May 7th investigation (logbook 1336): 13.6Hz looks the table horizontal resonance, 24.6Hz and 49.25Hz are the scroll pump, 603Hz (turbo pump). Peaks at 33.75 and 67.6 (looks its harmonics) are not known.
We then put the accelerometer on the homodyne box and measured coherence with the homodyne signal. This is in figure 3 and figure 4. In Figure 4: we see coherence with some narrow peaks (possibly originated by something like a fan?) plus the turbo pump (603Hz) and some large coherence around 640Hz and some reasonable coherence around 577Hz, however the accelerometer has no associated peak at these same frequencies, so we would conclude these are not resonant mode of the homodyne box (see also figure 10 of 1336).
The closest match is with the "no-piezo mount" we measured yesterdy (resonant mode 576Hz, see figure 8 of 1336).
NOTE: this frequency (around 576Hz) should be typical first resonant mode of all NO-PIEZO mounts - there are several. While, PIEZO mounts have first resonance mode around 470-500Hz. Then mounts seem to have higher resonant modes up to 2kHz or so.
Thank you a lot for your work.
I want to know the reason why the homodyne noise is improved from before.
Alignment of homodyne is improved? or something environment is changed?
I think there are mainly two points:
1. Turn off fan of clean room. Especially it brings noise around 20Hz. It is roughly the corner frequency where our spectrum starts to go up(from 20Hz to 0Hz, spectrum goes up). And also some other noise around low frequency region.
2. Maybe common noise rejection is also better. Everytime we make measurement, we make sure it is good.
Also:
3. The spectrum is plotted with poor resolution, so another reason could be that at low frequency nearby lines are "grouped" and the baseline seems higher.
Irene, Federico, Eleonora, Yuhang
Turbopump's 603Hz is everywhere.
Fig1-2: acoustic spectrum(fan on and off). From these two figures, we can see the clean room fan brings some narrow and broad peaks. The obvious narrow peaks are 22Hz and 386Hz. We could also see some broaden part around 60-110Hz and 330-470Hz.
Fig3: squeezing and acoustic coherence check. The acoustic coupling is everywhere. Some high and meaningful coherence peaks are 313Hz, 453Hz, 526Hz, 756Hz, 1008Hz, 1524Hz.
Fig4-5: squeezing and homodyne box vibration coherence check. Above around 2kHz, there is almost no coherence. Below 1.6kHz, at 24.5Hz(also harmonic), there is some coherence. And it should come from scroll pump. Also 134Hz, 140Hz, there is coherence. Also 281Hz, which comes from the resonance of the homodyne box. Also 516Hz, 533Hz, 591Hz, 757Hz.
Fig6: squeezing and IR phase shifter coherence check. Resonance around 468Hz gives rise to the noise level of squeezing. Other narrow peaks should come from other vibration sources.
Fig7: squeezing and GR phase shifter coherence check. The resonance of the GR phase shifter affects squeezing less. Maybe this is related to the high Finesse of GRMC.
Fig8-9: horizontal and vertical vibration on the top of the bench and coherence with squeezing noise. Horizontal contains 533Hz, 313Hz, 685Hz. Vertical coupling is not so severe.
Fig10-11: squeezing and rack coherence check. It contains 416Hz, 694Hz, 758Hz(from new spectrum analyzer), 833Hz, 878Hz.
Conclusion:
1. We could see a broad peak at around 9kHz. The peak around 20kHz is not analyzed with enough resolution.
2. The noise of the squeezing part is quite broad. But the acoustic and vibration effects cannot explain all the broad rise of squeezing noise level. There should be other noise sources.
plan:
1. We should check the acoustic spectrum at a higher frequency region.
2. For peaks related to mirror mounts and homodyne box, we should be able to remove it if we float bench. For acoustic noise, we should cover bench.
3. Check homodyne sub-DC spectrum up to 50k.
4. There are broad peaks around 800Hz, 640Hz. We should find out where is it from.
After having a better matching of homodyne, we tried to measure the squeezing level before the golden week. However, we failed because of misalignment of homodyne. Besides, we also changed green power to 40mW.
We have a better squeezing level, which is 4.4dB.
We can also see we have two main peaks in this plot. One is around 9kHz and the other is around 20kHz.
Eleonora, Eleonora, Irene, Yuhang, Federico
Day one: first measurements on "in-air" bench and on some (supposed) critical optical devices/mounts.
We used an accelerometer Wilcoxon 731 (low frequency, low noise) for bench measurements, and a PCB 352C68 (low weight - see photos) for optical mounts.
Bench characterization (accelerometer placed in horizontal and vertical over the bench; later over a top cover).
In fig.1 we see a main horizontal bench mode at 13.5Hz
In fig.2 we see some vertical mode of the bench (broad, not well defined) around 45Hz
In fig.3 we see a comparison (vertical only) between floor and bench acceleration: they are quite similar except the 45Hz region that is a bit amplified on bench as expected because of the resonance. We also notice a line at 26.75 that goes on/off, maybe is some devices into the building.
In fig.4 Bench lines at 24.75Hz and 49.5Hz are due to the vacuum Scroll pump always running near the bench.
In fig.5 we tapped ont the top panel (bench cover) and we excited a fundamental mode (drum?) at 4.5 Hz with harmonics.
Optical mounts characterization (accelerometer clamped on devices using some dedicated harware - if you need it in Italy, ask for a "fattapposta" - see photos)
In fig.6 "Infrared Phase shifter" (Coherent Control) placed horizontal, tapping on the bench we excited a resonant mode of this mount at 468Hz
In fig.7 the same but with the accelerometer placed in vertical, broad structures excited (one around 1212Hz and one around 1748Hz)
In fig.8 accelerometer horizontal on a similar mount but without the piezo: we excited a structure at 576Hz (very near to the 604Hz coming from the many turbo vacuum pumps)
In fig.9 "Green Phase shifter" (Coherent Control) placed horizontal, tapping on the bench we excited resonant modes at 500Hz, 1144Hz, 1584Hz
In fig.10 we placed the accelerometer horizontal (well, in line with the PD box) on the "homodyne" and we excited modes at 132Hz, (maybe) 281Hz, 393Hz, 508Hz
Preliminary conclusions:
It seems some optical devices are sensitive to vibrations, and moreover some eigenfrequencies are present also in quiet seismic spectra.
The bench has its own "natural" frequencies 813.5Hz horizontal, 45Hz vertical).
Mounts have their own frequencies above 100Hz (see the numbers quoted above).
Some external noise sources (Scroll Pump, Turbo Pump, etc) come to the bench via direct coupling and/or air coupling; worth to suspend it with air legs and close it with the 1cm thick panel covers.
Yuhang and Federico
We found one of the high voltage drivers is broken. It is the high voltage driver for IRMC.
Federico checked the circuit inside. The regulator part of this machine is broken.
Matteo, Takahashi, and Yuhang
We connected the translucent tubing to each leveling valves. We also connected each valve to each isolator through the grey tube(One of them is from the valve to the 'Tee' and use translucent tubing between the 'Tee' and either isolator). We also found an isolated air compressor which can continuously provide air. Although the compressor is quite noisy, we could put it far away in the future.
However, before going on, we have a to-do list:
1. To buy an air regulator filter(ARF) https://www.newport.com/p/ARF
2. To evaluate the required pressure. To evaluate this value, we need to know the weight of the part above isolators including breadboard, all the optical components, and the shield box. Up to now, I don't know how to evaluate it. The formula is ((weight(in pounds)/103.6)+10) psig.
3. To check if the center of gravity of the part above isolators is higher than B/2 or 3B/4. B is the distance between the longer leg distance.
4. To consider how to deal with the exhaust air from valves. Now all of them are closed. Since our bench is inside the clean room, we should find a way to evacuate it.