NAOJ GW Elog Logbook 3.2
To see the possibility of using HAMAMATSU PSD for INPUT mirror oplev, I did the comparison between Thorlabs PSD (with amplification) and HAMAMATSU PSD (without amplification) for INPUT mirror oplev.
The comparison result is shown in the attached figure.
Upper figure: (for INPUT pitch) REF0 is INPUT oplev spectrum with Thorlabs PSD. REF 2 Thorlabs PSD electronic noise after amplification. REF 4 INPUT oplev spectrum with HAMAMATSU 04 PSD. Red line: HAMAMATSU 04PSD electronic noise.
Lower figure: (for INPUT yaw)REF0 is INPUT oplev spectrum with Thorlabs PSD. REF 2 Thorlabs PSD electronic noise after amplification. REF 4 INPUT oplev spectrum with HAMAMATSU 04 PSD. Red line: HAMAMATSU 04PSD electronic noise.
It can be seen that Thorlabs PSD has a bit better SNR. This better SNR is proven to come from the amplification (proved in elog2114).
As reported in elog2281, the HAMAMATSU PSD electronic noise level is similar with DGS ADC noise. In this entry, I report a futher investigation of them.
1. I had a closer look into these two noise spectrum. As shown in the attached figure 1 (RED is elec noise, BLUE is ADC noise, GREEN/BROWN are integrated noise), elec noise is a bit higher than ADC noise but not a factor of 2. Therefore, to investigate further the electronic noise, we should amplify signal from PSD before it goes inside DGS system.
2. Before amplification of PSD individual signal, I checked the time series of individual signal from PSD. This is shown in the attached figures 2. We could see that it has very large noise. Therefore, it is very diffcult to amplify such large noisy signal. Indead, I found SR560 could only give a factor of 2 amplification for it before saturation. But the good thing is that, after the combination inside DGS, this noise is cancelled and resulted in a clean pitch/yaw electronic noise.
So if we really want to check better the electronic noise, we should combine PSD individual signal before they go inside DGS system.
By replacing 03PSD with 04PSD for BS oplev, I compared their performance.
The measured oplev spectrums are shown in the attached figure 1 (RED 04PSD, BROWN 03PSD), while the electronic noise comparison with ADC noise is shown in the attached figure 2 (RED is elec noise, BLUE is ADC noise, GREEN/BROWN are integrated noise).
We could conclude that both 03PSD and 04PSD have almost the same performance. The difference in position resolution is not affecting their performance now. So I guess this position resolution maybe related with beam size.
I tried to measure the residual gas molecules by a mass spectrometer.
The Q-mass could be operated by a front panel, and the pressure was 3.0*10-5 Pa, 7.8*10-7 Pa for H2o and N2 respectively.
Hoever, I could not connect to PC, and could not do degas.
This Q-mass needs degas but in order to do that, the connection to PC is needed...
I will ask the company.
Since we have problem of TAMA PSD, we considered to buy new PSD. Matteo asked two PSD from HAMAMATSU company for test. They are C10443-03 (we call it 03 later) and C10443-04 (we call it 04 later). The information of them can be found from this link.
From the datasheet, 03 and 04 PSD only have difference in position. I think the integrated noise spectrum can represent a position resolution. If so, the one with better position resolution should have a lower noise spectrum. Therefore, I firstly tested 03 which has a better position resolution. If it is better than the old PSD, we should use it in the future.
Together with Matteo, we made a customized circuit based on Mammoth connectors, bananna connectors and lemo connectors. In this way, the eight channels from HAMAMATSU PSD can be connected to power supply and ADC of DGS system. Four channels from PSD are used for power supply, which contain one plus(12V), one minus(-12V) ,and two grounds. Another four channels are called x1, x2, y1, and y2 separatly. The experimental set-up is shown in attached figure 1.
To convert four signal channels into pitch and yaw, I realized a matrix inside simulink file. This matrix is enclosed in a block called 'BS_test', as shown in the attached figure2.
Since old measurements have been saved in DGS system, I just plot new PSD data with old ones. As shown in the attached figure 3, there is comparison between old low gain TAMA PSD and 03PSD. We can see
1. REF0/1 (BS angular motion sensed by low gain TAMA PSD) is comparable with REF16/17 (BS angular motion sensed by 03PSD). This means that two PSD have the same gain.
2. REF8/9 (electronic noise of low gain TAMA PSD) is lower than REF24/25 (electronic noise of 03PSD).
3. Red lines in this figure shows the ADC noise. Actually this measurement is a bit strange. It shows that ADC noise is even higher than the electronic noise of low gain TAMA PSD. This can be correct if the ADC noise really becomes worse by itself.
4. Since the Red lines overlap with REF 24/25, it means the measured electronic noise maybe just ADC noise. So if we want to check better mirror angular motion, we need to amplify the signals coming from PSD.
I also compared the electronic noise between high gain TAMA PSD with 03PSD. The result is shown in the attached figure 4. Red lines are 03PSD electronic noise, which is lower than REF8/9 (high gain TAMA PSD electronic noise).
We will also test 04 PSD to justify the relationship between PSD position resolution and noise spectrum.
I pumped down the cryostat and turned on the Q-mass.
The LED lamp next to "POWER" turned on, though that of "Pa" did not.
It seems that this Q-mass may not be used anymore.
Furthermore, in order to degas the Q-mass, we need to pump down below 10-4 Pa, though it cannot reach below 6*10-4 Pa...
I gave up the measurement.
1) Two vacuum sensors were found not working. They are END station (arm side) and MID station (TAMA central side). The controller seems to work properly but it provide a "FAIL" message when reading the sensors.
2) The rotary pump in MID station has what it seems an oil leakage (see picture 1 and 2). I did not noticed this leakage when shutting it down on Friday but I did not payed particular attention to it, so I cannot tell if the problem was already present.
3) I could not find where to switch on the air dryer of South Arm. The West Arm air dryer are working.
Eleonora, Matteo and Yuhang
All facilities and devices were recovered after the Mitaka power outrage.
The Mitaka power outage was done on 14th Nov., while the recovery work was done on 16th Nov.. Although everything was recovered, there were some issues we encountered during the recovery work.
1. The main power switch for the filter cavity arm lights were not found at the beginning. The reason is that this switch is located in the middle of filter cavity arm (I didn't know this at the beginning).
2. The sound issue of DDS. The solution is provided in elog1794.
3. To power on DAC, there is a DC voltage supply. At the beginning, I forgot to provide negative 18V. Then, while I increased the positive voltage, the maximum can reach only around 6.5V.
4. This time, none DDS channels flipped phase by 90deg.
5. The optimal temperature of SHG has been changed from 3.074 to 3.102. On the second day, the value changed back to 3.081.
6. The position of suspended mirror didn't change a lot. We don't need to change largly voltage offset sent to mirrors to recover FC flash.
7. PDH signal of FC lock has large 5kHz resonance even when FC is unlocked. It was figured out that this oscillation comes from SHG. And the problem is solved by reduce the gain of SHG to 2.1.
8. It was found that the input value of DGS doesn't change on medm. We checked medm sitemap/CDS. The situation is shown in the attached figure 1. It was figured out later that this is because I was using sine wave instead of square wave. After correcting to the good waveform, DGS worked again. The good sitemap/DGS is shown in the attached figure 2.
9. It was found that ratory pump in the middle arm has oil leakage.
1) Two vacuum sensors were found not working. They are END station (arm side) and MID station (TAMA central side). The controller seems to work properly but it provide a "FAIL" message when reading the sensors.
2) The rotary pump in MID station has what it seems an oil leakage (see picture 1 and 2). I did not noticed this leakage when shutting it down on Friday but I did not payed particular attention to it, so I cannot tell if the problem was already present.
3) I could not find where to switch on the air dryer of South Arm. The West Arm air dryer are working.
Oil under the rotary pump was there from the old days. I replaced the rotary pump to new dry pump (ACP15). The TMP with the dry pump is working now at the mid point.
I measured SHG cavity scan again. This time, I put SHG temperature 2.8kOhm to avoid green conversion while nominal temperature is 3.1kOhm. Peak shape is still not completely symmetric possibly due to high IR injection power, but measured finesse is 70 which is reasonable value.
To project PR/BS angular motion to AA signals, we should use TF measured with excitation. But what is the difference between with and without excitation is an interesting thing to check. So I did this check. The result is shown in the attached figure.
It can be seen that magnitude of TF can be similar or very different when the coherence becomes good.
Matteo and Yuhang
In elog2236, before CCFC demodulated, its SNR was reported to be about 40dB with taking ~50% FC_ref . And the signal was ~-25dBm. So it was clear that SNR is large enough. Nevertheless, when we check the demodulated CCFC on the oscilloscope, the signal seems not to have an SNR of 40dB. So we thought the noise couples through the demodulation process, which could be solved by amplifying the CCFC signal.
The TAMA RF PD was used to acquire CCFC. And it was investigated when we used it to acquire CC1 from OPO reflection. That investigation proved that amplification was not effective to improve SNR. However, recently Matteo found that the resistor was chosen inappropriately during that time. In the first attached figure, the resistors used to amplify RF signals are marked. As shown in the attached figure 2, these resistors need to be 10Ohm and 100Ohm to have the best performance. However, it was chosen to be 1kOhm and 10kOhm when it was used for CC1. So we put the recommended 10Ohm and 100Ohm in TAMA RF PD.
Then the modified TAMA PD was placed back to the reflection of FC with ~50% FC_ref going inside. We checked again amplifier noise, TAMA PD noise ,and CCFC sideband. The test result is shown in the attached figure 3. There is also a zoom-in of the CCFC sideband shown in the attached figure 4. We can see that the signal is amplified by ~10dB while noise is amplified by ~20dB. A factor of 10 of voltage amplification should corresponds to a factor 100 power amplification. So the 10dB signal amplification seems to be limited somehow.
But anyway, we demodulated this signal and checked it on oscilloscope. The signal didn't become much cleaner.
The contribution of each DOF of PRBS angular motion is shown in the attached figures.
Today, I checked again the balance, demodulation phase, and rotation angle. The results are shown in the attached figure.
The balance between each channel is not optimized now. The demodulation phase is also not optimized. The pitch/yaw coupling becomes a bit better.
This result means the balance problem cannot be solved by a stable alignment.
To understand the noise contribution from PR/BS to the AA signal, we did several measurements and calculations. They are reported in elog2266 and elog2270. However, the sum of contribution was not done in the way of quadrature sum. So the results shown in elog2266 and elog2270 are not correct. Besides, elog 2270 used TFs which were measured without excitation. This should be replaced with TFs which are measured with excitation. This is due to that excitation can make the measurement of TF have more coherence. So we follow the same way of calculation with elog2270, except for the TFs and the noise sum method.
The measurement of TFs was done by exciting PR pitch, PR yaw, BS pitch, and BS yaw one by one. We call the measured TF with the style of TF_PRP_IP, which stands for the transfer function from PR pitch to INPUT pitch.
The quadrature sum of noise projection to INPUT pitch follows the equation: IP = PRP*TF_PRP_IP + PRY*TF_PRY_IP +BSP*TF_BSP_IP +BSY*TF_BSY_IP. Here mirrors angular motion use name-style PRP to stand for PR pitch motion.
The results of noise projection and measurement is attached. Note that the data of PR/BS oplev signals and AA signals are directly from DGS. So the value itself doesn't correspond to a meaningful unit.
From this result, PR and BS angular motion are not the limiting noise of AA signals, except for INPUT pitch.
The contribution of each DOF of PRBS angular motion is shown in the attached figures.
At the beginning of AA implementation, we checked balance (between each segment of QPD), demodulation phase (this is equavilent demodulation phase set in DGS, which can rotate the input I/Q signals by a matrix. In our case, we rotate signals to I phase.) and rotation angle (this is the quavilent rotation of QPD). All of these were done without the lock of AA loop. And we found some change from day by day.
Since now we have AA locked. I checked them again and optimized accordingly. The old work gave already quite good gain/phase/angle values. For example, in the first two attached figures, we can see the balance situation of QPD1 and 2. The balance is already quite good. Since we could get exact values from figrue 1 and 2. I calculated the required gain to balance them. After applying these gain, the balance situation is shown in the attached figure 3 and 4. We can see that the balance becomes much better. But if this will change again from day by day should be examined later.
(Old gain is 1 for each segments.) New gain is shown in the following table
|
|
I1 |
Q1 |
I2 |
Q2 |
I3 |
Q3 |
I4 |
Q4 |
|
QPD1 |
1.000 |
1.000 |
0.802 |
1.661 |
0.673 |
1.059 |
1.204 |
1.557 |
|
QPD2 |
1.000 |
1.094 |
0.850 |
1.040 |
1.055 |
0.900 |
0.789 |
1.130 |
After that, I also optimized the demodulation phase. The result is shown in the attached figure 5 and 6.We could see that only negligible signals go to Q phase. This should also be checked again later.
New phase is shown in the following table
|
|
Segment1 |
Segment2 |
Segment3 |
Segment4 |
|
QPD1 |
97 |
100 |
97 |
99 |
|
QPD2 |
135 |
125 |
138 |
128 |
Then, the rotation angle was also optimized. The optimization result is shown in the attached figure 7 and 8. However, we could see that pitch to yaw coupling has become already around 15% in WFS1/2. This should also be checked later if they will change or not.
Old rotation angles are all zero. New angles are shown in the following table
|
QPD1 |
QPD2 |
|
13 |
5 |
In the end, a new matrix was developed to close AA loop. The loop filters/gains are the same with the old case. A comparison of diagonalized signals are shown in the attached figure 9. It can be seen that they don't have large difference.
|
pitch |
WFS1 |
WFS2 |
|
Input |
1 |
-0.5 |
|
End |
1 |
2 |
|
yaw |
WFS1 |
WFS2 |
|
Input |
-1 |
0.8 |
|
End |
-1 |
-3 |
Today, I checked again the balance, demodulation phase, and rotation angle. The results are shown in the attached figure.
The balance between each channel is not optimized now. The demodulation phase is also not optimized. The pitch/yaw coupling becomes a bit better.
This result means the balance problem cannot be solved by a stable alignment.
By measuring the transfer function from PR/BS p/y motion to Input/End diagonalized signals, we could calibrate PR/BS p/y motion to Input/End diagonalized signals. As a result, we could know how much PR and BS angular motion are contributing to the diagonalized AA signals (Input/End diagonalized signals).
Some measurement situation: The transfer function measurement was done without excitation. PR/BS local control was closed.
Reconstruction method: Let's take reconstruction of Input_p as an example. Input_p = PR_p*TF + PR_y*TF + BS_p*TF + BS_y*TF. This means that we sum PR/BS p/y motion's contribution to AA diagonalized signal. (Here PR_p/y and BS_p/y signal is directly the oplev signal without any modification. The Input_p signal is also directly from DGS system)
The results are shown in attached figures. We could see that AA diagonalized signals are almost totally limited by PR/BS angular motions.
I measured SHG cavity scan to check finesse of SHG (Pic. 1,2). The fitting result doesn't match the measured data very well possibly due to the non symmetric peak shape. From the fitting, I obtained SHG finesse of 43, but this is not consistent with design value of 72.
I measured SHG cavity scan again. This time, I put SHG temperature 2.8kOhm to avoid green conversion while nominal temperature is 3.1kOhm. Peak shape is still not completely symmetric possibly due to high IR injection power, but measured finesse is 70 which is reasonable value.
Takahashi-san, Tanioka
We installed a quadrupole mass spectrometer (Q-mass) to the cryostat as shown in the attached picture.
I will try to take data from next Monday.
After the AA loop could be closed with high UGF, the locking noise was not checked again. So I did this check today.
The PR and BS local control was closed all the time. Then I measured locking noise when AA loop is closed or open. The result is shown in the attached figure.
The locking noise level is much higher than the one reported in elog2231, which has lower AA UGF.
If we compare the shape of the locking noise spectrum, it is very similar with the diagonalized AA_INPUT_PIT.
The transfer function from PR/BS motion to AA diagonalized signal was measured and reported in elog2265. We were thinking to use this TF and PR/BS motion to reconstruct the AA diagonalized signal. Then we would like to compare this reconstructed signal with the measured AA diagonalized signal. We want to do this comparison because we found the AA diagonalized signal is higher than the corresponding INPUT/END oplev signal. This was reported in elog2245, which was strange for us because AA singal is usually less noisy.
Therefore, we measured the TF from PR/BS p/y to AA diagonalized signals when excitation was sent to PR/BS. As reported in elog2265, the coherence is only large enough between around 10 to 40Hz(the spectrum below 10Hz is not considered because the higher AA noise is mainly found above 10Hz). This was later figured out that, as shown in the attached figure 1, the excitation send to PR/BS make their motion higher than PSD noise level above 10Hz. In the usual case (no excitation), as shown in the attached figure 2, the spectrum above 10Hz is bascially PSD electronic noise.
Therefore, to calibrate PR/BS motion to AA diagonalized signal, I took one point above the PSD noise level while avoiding peaks or large deviation. After that, I assume the spectrum has 1/f2 slope because the measured AA diagonalized signal has also 1/f2 slope. And I got the attached figure 3 as the spectrum of PR/BS p/y motion above 10Hz.
To combine TF and PR/BS motion, I checked the coherence of TF. Since we have p/y coupling in AA, we found the PR/BS motion has the following coupling contribution for AA diangonalized signal.
|
|
PRP |
PRY |
BSP |
BSY |
|
AA_EP |
1 |
|
1 |
|
|
AA_EY |
|
1 |
|
1 |
|
AA_IP |
1 |
|
1 |
|
|
AA_IY |
1 |
1 |
1 |
1 |
Here '1' means PR/BS motion will contribute to AA. For example, AA_EY (reconstructed) = PRY*TF(PRY to AA_EY) + BSY*TF(BSY to AA_EY)
In this way, I got the reconstructed AA signal as shown in the attached last four figures. (measured AA signal is also shown for comparison). Note that the spectrum above 30Hz is not compared because there was no coherence.
The reconstruction fit well for END mirror. But the reconstruction is higher than the measurement for INPUT mirror.
