The oplev p/y sensing coupling was optimized long time ago. SInce recently we compare AA diagonalized signal with oplev signal. We decide to check again the oplev p/y coupling again.

The PR and INPUT have not obvious p/y coupling as shown in the attached figure 1 and 2. But there is coupling for end mirror (as shown in the attached figure 3), which was removed by putting PSD rotation angle to be zero (as shown in the attached figure 4).

Long time ago, we had issue of BS PSD pitch which shows large noise at high frequency (see elog1642). Today, we found BS PSD pitch noisy again. And this time, the spectrum is as shown in the attached figure 1, which is even worse.

To make sure it is the problem of PSD, I replaced it with another PSD, which has smaller gain. Then the spectrum is as shown in the attached figure 2. As we can see, the pitch spectrum is fine. Therefore, the old BS PSD really has some problems.

I also discussed with Eleonora, she suggested me to buy/make a new one.

The problem is solved by using directly a long ethernet cable between computer and picomotor box.

Today, when I align filter cavity, I found the output of Input pitch is very large. As shown in the first attached figure, the H3 coil has an output of 30,000. This is reaching the maximum of coil.

So I tried to offload this value with picomotor. However, there is an error coming out when I use labview to move picomotor. The error is shown in the attached figure 1. To fix this problem, I did following check:

1. We tried to move PR picomotor, which worked. Therefore, the Labview should work well. And the network connection of the old PC should be also fine.
2. We checked the ethernet cable connection to the picomotor box. But the cable is connected to a high cable (as shown in the attached figure 2). So it is not very easy to check if the cable is correct or not. But anyway, this cable is connected.
3. We also tried to plug in and out the power cable for picomotor box, which didn't solve the problem.
4. We checked also the IP address of picomotor box, which is consistent with the one used in Labview.

However, the picomotor still didn't work. I have already sent Takahashi-san an email to ask how to solve this problem.

The measurement is the same with the last comparison of AA/oplev except for that the PR/BS local control is off.

But even in this case, the AA signal is still higher than oplev signal.

This coherence check was done between PR/BS oplev singal and AA diagonalized signal. Note that there is no excitation sent to PR/BS, but the PR/BS local control loop was closed.

Note: blue curves are PR/BS local control open, red curves are PR/BS local control closed.

We could see a strong coherence between PR pitch/yaw oplev signal and AA.

Therefore, the PR local control seems to be introducing noise in pitch/yaw from 10 to several tens of Hz.

This coherence check was done between PR/BS oplev singal and AA diagonalized signal. Note that there is no excitation sent to PR/BS, the PR/BS local control loop was open.

It can be seen that there is almost no coherence above 10Hz. I think this is also reasonable, because PR/BS motion at high frequency should be similar.

As reported in elog2245, the AA noise level is in general higher than the noise of oplev from roughly 10Hz to 100Hz (looking at all Input/End pitch/yaw). This noise is suspected to come from PR/BS motion.

Therefore, the coherence check between PR/BS pitch/yaw excitation and AA signal was done. The result is shown in the attached figure 1-4. It can be seen that:

1. Compared with BS pitch excitation, all AA singals have high coherence between 10 and ~30Hz. (Fig 1)

2. Compared with BS yaw excitation, Input yaw shows the most coherence between several Hz to ~30Hz. (Fig 2)

3. Compared with PR pitch excitation, all AA signals have high coherence between ~30Hz to 100Hz. (Fig 3)

4. Compared with PR yaw excitation, Input yaw shows the most coherence between several Hz to ~30Hz. (Fig 4)

Besides, the AA signal with PR/BS local control loop open was also checked. The result is shown in the attached figure 5. It is clear that, the noise floor from ~10Hz to 100Hz becomes generaly lower. But there is no difference in End mirror yaw.

Later on, I found that all the filters disappear in foton.

Eleonora checked that we have archive file in the directory '/opt/rtcds/kamioka/k1/chans/filter_archive/k1fds'. The file used in foton is '/opt/rtcds/kamioka/k1/chans/K1FDS.txt'.

Then Shoda-san helped to check them. 'ls /opt/rtcds/kamioka/k1/chans/filter_archive/k1fds -all' was used to check the date of these files. The latest archived file in that directory was found to be 'K1FDS_201021_164656.txt' However, it was also shown in the terminal that this latest file has size of 3075 (usual size is 107155). We opened the file and found that the latest archived file is actually empty. In the end, we found  'K1FDS_201017_210326.txt' should be the latest useable file.

By using 'cp /opt/rtcds/kamioka/k1/chans/filter_archive/k1fds/K1FDS_201017_210326.txt /opt/rtcds/kamioka/k1/chans/K1FDS.txt', the filters are copied to foton. Now filters also work well!

When the transfer function is measured in diaggui, several choice could be used for measurement. Especially, when performing FFT measurement, there is 'Gaussian' and 'uniform' noise to be chosen. I checked today what is the difference of them.

I used Diaggui to generate these two noise as shown in the first figure. Then the spectrum of them was measured and shown in attached figure 2.

The conclusion is that both of them are white noise. But the 'Gaussian' noise is about twice the amplitude of 'uniform'.

It was figured out that there were two problems happened at the same time and caused this problem.

1. Instead of using 'make install-k1fds', I was using 'make install -k1fds'. By using correct command, the problem was solved.

2. One of the directory was full (/opt/rtcds/kamioka/k1/target/fb/log/old was as large as 18GB) . Then I couldn't make k1fds. By using 'sudo rm -r old', this problem was solved.

Tips:

To check if k1fds is running, command 'lsmod' could be used. After using it, if you see the name k1fds is listed on the left side in a certain column, it means k1fds is running.

To check the space used by different directory, 'du -h --max-depth=2 .' should be used.

Shoda and Yuhang

I found that the driving matrix of AA had large coupling last Friday. So I was thinking to make the driving matrix not seperate in pitch and yaw. It means to make the 2*2 matrix be a 2*4 matrix. For example, to control the yaw motion, there is not only signal coming from yaw but also from pitch.

Then I asked Shoda-san to modify the k1fds code. She helped me make the modification described as above.

But then, later on last Friday. By improving the driving matrix index with the separation of pitch and yaw, the pitch/yaw coupling becomes much better. So I decided by myself to modify it back.

However, after I modified it back. I found I couldn't make k1fds. The problem showed up as

gzip: stdout: Np space left on device.

make: ***[k1fds] Error 1

However, although this issue reased up, I was able to use control loops. So I didn't care about that.

Then, the day after, I found filters in foton disapeared. 

Eleonora and Yuhang

Firstly, we found that all the filters disappeared in foton (as shown in the attached figure 1).

Then we tried to restart standalone computer. However, after that, k1fds stopped working (as shown in the attached figure 2).

At the same time, we found that by using command 'df', /dev/sda1 shows 100% used (as shown in the attahed figure 3).

Eleonora tried to ssh standalone, and used command 'lsmod', and found there was no k1fds. She tried command 'startk1fds'. But it didn't work (see attached figure 4).

So probably the problem is connected with the full of /dev/sda1.

To check how AA affect input/end mirrors, I checked their oplev signals when AA loop is on/off.

From the measurement, the yaw motion of input/end mirrors is actually increased by AA loop.

0. Note that PR/BS local control is closed in this case.

1. The noise spectrum of AA is higher than oplev (from 10 to 100Hz)

2. The input pitch coupling to input yaw is visible.

To compare oplev and AA spectrum, the first step is to calibrate both of them. The calibration is done as follows:

1. Calibrate oplev signal. The oplev signal calibration method was developed by Eleonora and written in elog1874. 

2. Clibrate AA signal. The 4Hz sine wave was sent to Input/End pitch/yaw. By adjusting calibration factor of AA signal, AA's 4Hz peak was matched to oplev's peak. Then the calibration of AA signal was decided. The comparison is shown in the attached four figures.

unit: urad/counts

The first attached figure was the coupling (input pitch) situation before optimization.

The following figures show the coupling situation now.

According to the suggestion of Raffaele, I checked the diagonalized sensing signal (driving signal). I found there are quite a lot of coupling between each DOF. Therefore, I optimized the invertion matrix. After that, I also increase the gain of each loop.

New driving matrix:

New gain:

input yaw: 10

input pitch: 8

end yaw: 10

end pitch: 12

The comparison of new spectrum and old one is shown in the attached figure. We could see:

1. The coupling is bascially not visible in the spectrum.

2. The control bandwidth is increased.

3. The AA helps to reduce input mirror pitch motion to 2urad. Reduce input mirror yaw motion to 1urad.

4. The AA helps to reduce end mirror pitch motion to 1.2urad. Reduce input mirror yaw motion to 0.8urad.

As we concived that the mirror angular motion results in the difficulty of mode matching, this situation maybe improved by the AA control.

Calibration factor: The calibration was done by sending 4Hz with 5urad expected motion (driving magnitude is shown in elog2216). (I didn't consider yet the transfer function of pitch/yaw. Since 4Hz is not far from resonance frequency, the pendulum effect is temporary neglected.) Beside, looking at the noise spectrums, at 4HZ, there is not the effect of AA loop. Then I check the peak value at 4Hz. After subtracting the offset (value at 4Hz without excitation), the value is divided by 5urad. Then I get calibration factor. I use it to calibrate the spectrum measured from wavefront sensor.

The measurement result is shown in the attached figure. (REF0,1,2,3: AA loop on. REF4,5,6,7: electronic noise. Others: AA loop off. )

1. From this measurement, End mirror is moving less than Input mirror.

2. The RMS motions of Input mirror reach about 7urad for both pitch/yaw.

3. The RMS motions of End mirror reach about 2urad for both pitch/yaw.

4. The spectrum is well above the sensor noise.

5. The control bandwidth situation could also be seen from this measurement. The control in yaw direction has larger bandwidth and reach about several Hertz. The control in pitch direction has smaller bandwidth, but also reach about 1Hz.

6. It seems that the coupling from pitch to yaw is not small. The peaks in pitch (around 8-9Hz) is visible also in yaw.

The coherence in pitch direction is not very good in the last elog. Therefore, some new measurements were performed.

1. The first four figures are measured with gaussian noise excitation, resolution is 0.01Hz.

2. Figure 5 shows a measurement of END_yaw TF with uniform noise excitation. The measurement result is a bit different with guassian excitation.

3. I increase the input pitch gain from 0.3 to 2. The new TF is shown in figure 6.