[RXA, KK]
As shown in the gigE camera image earlier in this thread, the MCO spot position is largely off in YAW (cm scale).
We observed the MCL control signal coupling from mirror YAW (PIT) motion (angle to length coupling) while changing the coil balance of YAW (PIT) from [H1 H2 H3 H4] = [1 1 -1 -1], [1+1dB, 1+1dB, 1-1dBm, 1-1dB] (let's call it 1 dB unbalance), ... to 3dB unbalance. When the spot position and the P/Y tilting center are matched on the mirror, no angular signal should be seen the MCL ctrl signal. We used the lockin osc matrix to apply the balances and actuations.
The result is attached. The x axis is the applied unbalance, which is equivalent to the beam spot distance from the center of ther mirror. The x axis should be calibrated to the spot position from the center, but I couldn't find the distance between the magnets. It will be updated later. The Y axis is the signal in MCL control therefore showing the coupling level from the angular motions. As shown in the plot, MCO YAW is the most mis-centered one.
Looking back the klog, MCO was repaired from the fallen-off magnet and re-installed in May 2018, then the in-air alignment and spot position check was done in June with Stefan. Also, as posted earlier, checked again in September 2018 (klog 6317). I am puzzled how this position change on MCO could have happened. Removal of the MZM path (May 2019) on the PSL table was very careful (checking the MC flash every time we removed a mirror) I don't think that could cause the big input beam misalignment.
I have added velocity damping of the MN oplev to the MN stage of ETMY, as well as for the TM stage of ETMY to the TM stage.
I have turned off the outputs of the P## filtes in the OLSERVO screen, since most of them seem to not do much, or ring up modes. I'm not happy with it yet though, so we need to revisit this more tomorrow - just wanted to write a note explaining why ETMY is in a different state for now.
I know that people have told me this, but to show myself, I did a rough calibration of the ETMY photosensors to the ETMY test mass staget optical lever (the marionette oplev has a different calibration by a factor of almost 2!) Anyhow, indeed the photosensors are very, very noisy, and it's not clear that using them even just for damping will be very useful.
It looks like we accidentally turned off the upper stage (BF & IM) damping on the PRC suspensions. That's why we were getting so much problem at 0.1 & 0.15 Hz..
We turned on all the BF & IM damping again. The RMS at the mirror stage is ~1 urad now, so we need to still do some tuning.
right on, man
As far as I remember, those whitening filters for the QPDs of oplevs have not yet been fully well activated, right?? I just wondered if the timing has come to activate them. The noise floor above 2Hz would be due to ADC noise (or... the limited amount of the power got by the QPDs... shot noise is far away) according to my rough check (1602), though the vertical reads seemed "digital counts" there at that time (I did not remember well, sorry).
right on, man
I again diagonalized the PR2 with free-swinging condition. L and Y seems to be diagonalized well, however, L and P could not diagonalized. I also checked
- I turned all damping filters off.
-
Take a ratio of the tilt oplev signal to the length oplev signal at the resonant frequency as follows:
L: 0.453, P: 0.85, Y:1.0 - Pitch cannot be diagonalized, since the coupling between the length oplev and the tilt oplev changes every measurement by factor of 2-3.
- PRM diagonalization matrix introduce the noise above 1 Hz. I'm not sure what happen from yesterday, but I turned off it.
The PR3 oplev damping had some problems and so we have modified and simplified it. Pitch is OK, but the low frequency yaw motion needs more work.
- the previous design had DC gain, so it was used to hold the mirror alignment to the OL QPDs. I removed this DC part so we can DC couple some upper stage and just use the TM stage for the reduction of motion in the 0.1 - 3 Hz band. This makes the filter design easier.
- The DC coupling was making the suspension trip the watchdog since the applied torque was too strong when the Guardian tried to align the mirror.
- The boosts in the previous design made it conditionally stable.
- The new design is just velocity damping and a 13 Hz low pass filter.
The pitch loop is working well, but the 0.15 Hz yaw mode (which probably comes from the upper stages of the Type Bp suspension) is not dampable from the bottom stage, so we'll have to figure out what to do above. Similarly, the 0.1 Hz mode is not simple to damp.
The 2nd attachment shows the plant models that I got from this very nice Model Plotter webpage. I wish we had this kind of thing in LIGO.
It looks like we accidentally turned off the upper stage (BF & IM) damping on the PRC suspensions. That's why we were getting so much problem at 0.1 & 0.15 Hz..
We turned on all the BF & IM damping again. The RMS at the mirror stage is ~1 urad now, so we need to still do some tuning.
this plot shows the comparison of PR2 & PR3 w/ oplevs on/off. There are problems...
- The PR2 & PR3 loops are suppressing the error signals below the level of the sensor noise (which seems to have a 1/f noise of ~5 nrad/rHz @ 1 Hz). i.e these loops are injecting noise in the 1-10 Hz band.
- The PR2 Yaw loop is making 20 dB of gain peaking at 5 Hz.
- There is a lot of acoustic noise. We need more serious low pass filters for the 30-100 Hz band. There is not much BBH SNR below 30 Hz, so lets not worry about 10-30 Hz for now.
I attached the spectrum for the K1-ASL_X/Y_FIB_SLOW_DAQ_OUT_DQ.
The low frequency part seemed to be suppressed a lot or fake ?
The 10x unbalance must be from somewhere else but the coils since all four coils on MCI, MCO, MCE are all healty. Diaggui data is on /users/VIS/190916/.
MCE
H1 @DC -45 @10Hz -74
H2 @DC -45, @10Hz -74
H3 @DC -45, @10HZ -74
H4 @DC -45, @10Hz -75
MCO
H1 @DC -50, @10Hz -81
H2 @DC -49, @10Hz -79
H3 @DC -49, @10Hz -75
H4 @DC -50, @10Hz -77
MCI
H1 @DC -45, @10Hz -77
H2 @DC -47, @10Hz -77
H3 @DC -47, @10Hz -77
H4 @DC -46, @10Hz -77
unit is [urad/cts]; each coil to oplev PIT channel.
And yes, the beam doesn't seem to be at the center on MCO. It seems to move since this time. MCE is better, and we have no camera for MCI. See the attached photo.
Lucia, Luca
Today we investigated the noise at 238Hz visible when the coil drivers are used for the feedback control of the the vertical filters. Indeed at closed-loop this line appears in LDT-F1, LVDT-F2, LVDT-F3. We applied a lowpass at 1Hz to the digital filter of the output coil of F1, but it doesn't have any effect. We also note that its value is close to be a multiple of the powerline frequency 60Hz, so we guess that it can be due to a mass loop entering through the hardware. We suggest to check the coil driver boards when it will be possible.
Then we measured the open loop transfer functon at each vertical filter (see attached plots):
- We observe that the phase is shifted of about 180° with respect to the measurement made in August. If we want to increase the gain of the F1 loop we should have to change the shape of the control but it would be not so easy due to the 238Hz noise line.
- we found that in F2 the gain was too low, and for this we were not able to damp the 0.2Hz peak. First we set the agin to 18dB, as shown in the attached picture. Then we increased the gain to about 26dB, keeping the loop stable without injecting too much noise. We note that also here the 238Hz noise line appears, but with a reduced amplitude (about a factor 10).
- we tried to increase also the gain for the F3 control loop, setting it to about 56dB. The filter is stable but it is not possible to increase more the gain.
- when we close the control loop of the BF it saturates due to the 238Hz noise line, then it is still open.
The attached spectra before (blue line) and after (red line) the tuning of the GAS filters are attached here. The 0.2Hz peak is damped.
Thank you so much, Rana and Jenne; actually as you mentioned, it would be difficult to work for covering the QPD side. For example, some pillars for supporting the cover would be required later...
Rana, Jenne
Saturday, during the AS WFS investigation and REFL beam clipping investigation, we constructed a wind shield enclosure for PRM. It has reducethe 1 Hz noise for the PRM optical lever by a factor of ~8.
In the attached PDF, you can see the spectra of the pit/yaw spectra. Nakano-san tells me that they are all calibrated into micro-radians, so I have labeled them as such.
For PR2 and PR3, the loops are closed and so the noise below 10 Hz is suppressed. However, it seems like to me that the loops are mostly sensor noise limited (increasing the gain will not reduce the RMS motion of the mirror).
For PRM, the BLACK trace shows the open loop spectra before shield and the Blue-Gray trace shows after wind shield. For this, we're using the QPD labeled as TILT. Later, we will mix in the QPD of the other Gouy phase to cancel the large longitudinal-to-yaw coupling (otherwise, these cannot be used for feedback stabilization).
By the fact that the noise reduction was so much means that most of the wind-susceptibility comes from the laser side (not surprising). The QPD side is difficult to work on.
also, I forgot to reconnect the PRM OL beam position sensor cables
, so that will have to be done next time we enter the mine.
QPD Comments:
- The QPDs should be mounted in a box, not on translation stages. We want to have EMI shielding for sensitive low frequency circuits. Some opamps have a large RFI downconversion problem.
- I recommend not to use ANY translation stages on these QPDs. They are not long term stable. Instead we should use the stable mounts (e.g. Thorlabs Polaris) for steering onto the QPDs.
- The optics should be moved away from the edge of the breadboard so that a proper box can be attached the breadboard.
- The box used for the MMT trans is too thick. For the OL, we can just use aluminum framing (as is used on the PRM OL laser side) and some steel sheet metal. Soft rubber gaskets should be used to attach the sheet metal to the frames. The sheet metal should have a few small holes for the cables to come out and the holes should include rubber grommets so that the cable holes don't leak so much air.
- All cables from the Optical lever laser and QPD should be strain relieved by attaching them strongly to the optical lever pier so that the vibration from the cable is minimized.
- The current 'round steel pier' approach is probably not very mechanically stable, since the pier's foot is so small relative to the moment of inertia of the pier. Ideally, the whole system should have a 'flagpole' mode resonance above 30 Hz.
Thank you so much, Rana and Jenne; actually as you mentioned, it would be difficult to work for covering the QPD side. For example, some pillars for supporting the cover would be required later...
this plot shows the comparison of PR2 & PR3 w/ oplevs on/off. There are problems...
- The PR2 & PR3 loops are suppressing the error signals below the level of the sensor noise (which seems to have a 1/f noise of ~5 nrad/rHz @ 1 Hz). i.e these loops are injecting noise in the 1-10 Hz band.
- The PR2 Yaw loop is making 20 dB of gain peaking at 5 Hz.
- There is a lot of acoustic noise. We need more serious low pass filters for the 30-100 Hz band. There is not much BBH SNR below 30 Hz, so lets not worry about 10-30 Hz for now.
EPICS gateway was stopped.
== What we did ==
- Changed EPICS_CA_ADDR_LIST
- Requested safe state for all suspensions
- Stopped EPICS gateway process
- Restarted MEDM and StripTool processes on k1ctr* and k1mon*
== Remaining problem ==
- IMC lock doesn't come back. (too large MCE length offset)
I forgot how we should do...
tasukete Naka-emon :_(
- Put excitation signal at the longitudinal resonance, and esimated the coupling raito between oplev yaw signal and len signal.
- No oplev control was engaged for this measurement
- resonant frequency is 0.66 Hz for PR3 and PR2
- The oplev signal with excitation after the diagonalization is shown in attached plot.
- PRM has been diagonalized by Shoda-san. I measered the spectrum with the excitation in L with/without non-diagonal component, and attached it.
- Even after the diagonaization of the sensor, the PR2 has a large peak in yaw oplev even I excited in longitudinal direction. It might be because of the actuator, but I'm not sure.
I again diagonalized the PR2 with free-swinging condition. L and Y seems to be diagonalized well, however, L and P could not diagonalized. I also checked
- I turned all damping filters off.
-
Take a ratio of the tilt oplev signal to the length oplev signal at the resonant frequency as follows:
L: 0.453, P: 0.85, Y:1.0 - Pitch cannot be diagonalized, since the coupling between the length oplev and the tilt oplev changes every measurement by factor of 2-3.
- PRM diagonalization matrix introduce the noise above 1 Hz. I'm not sure what happen from yesterday, but I turned off it.
Miyoki-san
I’m sorry, the legends of 9796 fig1 and fig2 are wrong.
Their correct channel is K1:ALS-Y_FIB_SLOW_DAQ_OUT_DQ .
The measurement result of the PLL channels is in 8685 (only high frequency side), and the noises level looks roughly same.
I put a camera on the IMMT TRANS breadboard to watch the beam shape of the transmitted light from the IMMT1. You can see the image of the beam in IMM camera. This is for the beam walking to fix the beam clipping on the IFI.
Rana pointed out that I shoud no use the ND to see the beam shape. However, I need to reduce the beam power, so for the first step, we will try the beam walking with using this image and the image on the IFO REFL. If we would have a problem, I will replace it by the 99 % reflection mirror.
------------------
Due to this camera, ISS PD don't have any light right now. Sorrry for inconvinience.
