This week, I have finished assembling everything I need to begin shaking. I built an intermediary mounting stage to mount the TT suspension base to the horizontal sliding platform, finished assembling the second photodiode, finished assembling the photosensor circuit box, and calibrated the two photosensors. Today I built a platform/stage to mount the photodiodes so that they are located close enough to the mirror/suspension that they can operate in the linear range. Below is an image of the set-up.
The amplifer that Koji fixed is acting a bit strange again...It is sometimes shutting off (Apparently, it can only manage to do short runs ~ 1minute? That should be enough time?).
The set-up is ready to begin taking measurements.
Last night, I attached a metal plate to the Vout faceplate of my photosensor circuit box because the BNC connection terminals were loose. This was Jamie's suggestion to establish a more secure connection (I had originally drilled holes for the BNCs that were much too large).
I have also fixed the mechancial set-up of my shaking experiment so that the horizontal sliding platform does not interfere with the photodiode mounting stage. Koji pointed out last night that in the full range of motion, the photodiode mounting stage interferes with the movement of the sliding platform when the platform is at its full range.
I have began shaking. I am getting a problem, as my voltage outputs are just appearing a high-frequency noise.
Thanks to Koji's help, the second photosensor, which was not working, has been fixed. I have re-calibrated the photosensor after fixing a problem with the circuit. I have determined the new linear region to lie between 7.6 mm and 19.8mm. The slope defining the linear region is -0.26 V/mm (no longer the same as the first photosensor, which is -0.32 V/mm).
Here is the calibration plot.
Koji and I have finished shaking the table for the first round of measurements (horizontal shaking). We have cleaned up the lab space used.
The FFT Analyzer has been put back to its position at the back side of the rack (near the seismometers).
I will calibrate the photosensor for the suspension frame and piece together/analyze/produce graphs of the data today. If everything is fine (the measurements are fine) and if there is a chance, we hope to shake the TT suspension vertically.
I have re-calibrated the photosensor I used to measure the displacements of the TT frame (what I call "Photosensor 2").
As before, the linear region is about 15.2mm to 25.4mm. It is characterized by the slope -0.0996 V/mm (-0.1 V/mm). Recall that photosensor 1 (used to measure mirror displacements) has a calibration slope of -3.2V/mm. The ratio of the two slopes (3.2/0.1 = 32). We should thus expect the DC coupling level to be 32? This is not what we have for the DC coupling levels in our data (2.5 for flexibly-supported, fully-assembled TT (with EDC, with bar), 4.2 for EDC without bar, 3.2 for rigid EDC without bar, 3.2 for no EDC, with bar, 3.2 for no EDC without bar) . I think I may need to do my calibration plot for the photosensor at the frame?
I have redone the voltage versus displacement measurements for calibrating "Photosensor 2" (the photosensor measuring the motions of the TT frame). This time, I calibrated the photosensor in the exact position it was in during the experimental excitation ( with respect to the frame ). I have determined the linear region to be 15.2mm to 22.9mm (in my earlier post today, when I calibrated the photosensor for another location on the frame, I determined the linear region to be 15.2mm to 25.4mm). This time, the slope was -0.92 V/mm (instead of -0.1 V/mm).
This means that the calibration ratio for photosensor 1 (measuring mirror displacements) and photoensor 2 (measuring frame displacements) is 34.86.
Since this "unity" value should be 34.86 for my transfer function magnitude plots (instead of the ~3 value I have), do I need to scale my data? It is strange that it differs by an order of magnitude...
All of my plots have already taken into account the calibration of the photosensor (V/mm ratio)
Here is a bode plot generated for the transfer function measurements we obtained last night/this morning. This is a bode plot for the fully-assembled T.T. (with flexibly-supported dampers and bottom bar). I will continue to upload bode plots (editing this post) as I finish them but for now I will go to sleep and come back later on today.
Here is a bode plot comparing the no eddy-current damper case with and without the bar that we suspected to induce some non-uniform damping. We have limited data on the NO EDC, no bar measurements (sine swept data from 7 Hz to 50 Hz) and FFT data from 0 Hz to 12.5 Hz because we did not want to induce too much movement in the mirror (didn't want to break the mirror). This plot shows that there is not much difference in the transfer functions of the TT (no EDC) with and without the bar.
From FFT measurements of the no eddy-current damper case without the bar (800 data points, integrated 10 times) we can define the resonance peak of the TT mirror (although there are still damping effects from the cantilever blades).
The largest resonance peak occurs at about 1.94 Hz. The response (magnitude) is 230.
The second-largest resonance peak occurs at about 1.67 Hz. The response (magnitude) is 153. This second resonance peak may be due to pitch motion coupling (this is caused by the fact that the clamping attaching the mirror to the wires occurs above the mirror's center of mass, leading to inevitable linear and pitch coupling).
Here is a bode plot of the EDC without the bar. It seems very similar to the bode plot with the bar
Here is a bode plot of the rigidly-supported EDC, without bar. I need to do a comparison plot of the rigid and flexibly-supported EDCs (without bar)
This morning (about 10am to 11am), I have collected additional transfer function measurements for the T.T. suspension. I have finished taking my measurements. The SR785 has been returned to its place next the the seismometer racks.
The data has been backed up onto the cit40m computer
Here is my bode plot comparing the flexibly-supported and rigidly-supported EDCs (both with no bar)
It seems as if the rigidly-supported EDC has better isolation below 10 Hz (the mathematically-determined Matlab model predicted this...that for the same magnet strength, the rigid system would have a lower Q than the flexible system). Above 10 Hz (the resonance for the flexibly-supported EDCs seem to be at 9.8 Hz) , we can see that the flexibly-supported EDC has slightly better isolation? I may need to take additional measurements of the transfer function of the flexibly-supported EDC (20 Hz to 100 Hz?) to hopefully get a less-noisy transfer function at higher frequencies. The isolation does not appear to be that much better in the noisy region (above 20Hz). This may be because of the noise (possibly from the electromagnetic field from the shaker interfering with the magnets in the TT?). There is a 3rd resonance peak at about 22 Hz. I'm not sure what causes this peak...I want to confirm it with an FFT measurement of the flexibly-supported EDC (20 Hz to 40 Hz?)
Since the last post, I have found from the Characterization of TT data (from Jenne) that the resonant frequency of the cantilever springs for TT #4 (the model I am using) have a resonant frequency at 22 Hz. They are in fact inducing the 3rd resonance peak.
Here is a bode plot (CORRECTLY SCALED) comparing the rigidly-supported EDCs (model and experimental transfer functions)
Here is a bode plot comparing the flexibly-supported EDCs (model and experimental transfer functions). I have been working on this graph for FOREVER and with the set parameters, this is is close as I can get it (I've been mixing and matching parameters for well over an hour > <). I think that experimentally, the TTs have better isolation than the model because they have additional damping properties (i.e. cantilever blades that cause resonance peak at 22 Hz). Also, there may be a slight deviation because my model assumes that all four EDCs are a single EDC.
As reported in my previous entry of TT supsension bode plots, I found that my experimental data had what appears to be very noise peaks above 20 Hz (as mentioned earlier, the peak at 22 Hz is likely due to vertical coupling, as 22 Hz is the resonant frequency of the cantilever blades). This is very unusual and needs to be explored further. I would like to vertically-shake the TTs to obtain more data on possible coupling. However, I am leaving on Monday and will not return until Thursday (day of SURF talks). I am leaving campus Friday afternoon or so. I would may need some help coming up with an assembly plan/assembling set-up for vertical shaking (if it is possible to do so in such a limited time frame).
Today I wanted to see if the "noisy peaks" above 30 Hz were due to EM noise coupling. I tested this hypothesis today, seeing if EM fields generated by the coil at higher frequencies were injecting noise into my transfer function measurements. I found that the "noisy peaks" above 30 Hz are NOT DUE TO EM NOISE COUPLING. I am very curious as to what is causing the high peaks (possibly coupling from other degrees of freedom)?
Using my Matlab model of the flexibly-supported eddy current damping system, I have changed parameters to see if/how the TTs can be optimized in isolation. As I found earlier, posted in my bode plot entry, there is only a limited region where the flexibly-supported system provides better isolation than the rigidly-supported system.
Here is what I have found, where \gamma is the scale factor of the magnetic strength (proportional to magnetic strength), \beta is the scale factor of the current damper mass (estimated by attempting to fit my model to the experimental data), and \alpha is the scale factor of the current resonant frequency of the dampers.
Here are my commentaries on these plots. If you have any commentaries, it would be very helpful, as I would like to incorporate this information in my powerpoint presentation.
It seems as if the TT suspensions are already optimized?
It may be difficult to lower the resonant frequency of the dampers because that would mean changing the lengths of the EDC suspensions). Also, it appears that a rather drastic reduction (at most 0.6*current EDC resonant frequency --> reduction from about 10 Hz to 6 Hz or less) is required . Using the calculation that the resonant frequency is sqrt(g/length), for my single-suspended EDC model, this means increasing the wire length to nearly 3 x its current value. I'm not sure how this would translate to four EDCs...
The amplification at resonance caused by increasing the magnet strength almost offsets the isolation benefits of increasing magnet strength. From my modeling, it appears that the magnet strength may be very close (if not already at) isolation optimization.
Lowering the mass to 0.2 the current mass may be impractical. It seems as if the benefits of lowering the mass only occur when the mass is reduced by a factor of 0.2 (maybe 0.4)
What are the parameters you are using? As you have the drawings of the components, you can calculate the masses of the objects.
Reducing the ECD resonance from 10Hz->6Hz looks nice.
The resonant freq of the ECDs are not (fully) determined by the gravitational energy but have the contribution of the elastic energy of the wire.
Q1: How much is the res freq of the ECDs if the freq is completely determined by the grav energy? (i.e. the case of using much thinner wires)
Q2: How thin should the wires be?
The drawings do not have the masses of the objects.
For the resonant frequency:
Instead of sqrt (g/l) would the numerator in the square root be[ g + (energy stored in wire)/(mass of damper)] ?
1) Drawing has the dimensions => You can calculate the volume => You can calculate the mass
Complicated structure can be ignored. We need a rough estimation.
2) Your restoring force can have two terms:
- one comes from the spring constant k
- the other from the gravity
The wire used to suspend the EDCs is tungsten?
To verify, for my model, the EDC will be the mass of all four dampers or a single damper? The length of the wire used to suspend the EDC will be the combined length of 4 wires or length of a single wire?
Taking into account the densities for each material (specific material of each component was listed, so I looked up the densities), and trying my best to approximate the volumes of each component, I have determined
the mass of the mirror + mirror holder to be ~100 g and the mass of a single EDC to be ~19 g
1) Drawing has the dimensions => You can calculate the volume => You can calculate the mass
Complicated structure can be ignored. We need a rough estimation.
2) Your restoring force can have two terms:
- one comes from the spring constant k
- the other from the gravity
The wire used to suspend the EDCs is tungsten?
I am thinking that perhaps my mass estimations were off? The model that I have used fits the data better than the model that I have made (changing the masses to fit my estimations of the values)
I have been redoing the noise test multiple times today. Here is the best plot that I got
A copy of my summer progress report 1 has been uploaded to ligodcc 7/711 and I have just added a copy to the TTsuspension wiki
PDF copy of Summer Progress Report
Calibration of Guralp Seismometers
Procedure & Results
Sinusoidal current of known frequency and amplitude was injected to the Seismometer calibration coil using signal generator and handheld control unit & corresponding Magnitude and Phase response were recorded. For Guralp B, system response was also estimated with a FFT Spectrum Analyzer.
Frequnecy Range: 0.1 Hz to 45 Hz.
Equivalent Input Velocity was derived from the Input Voltage measurements using the relation: v = V/ (2*pi*f*R*K) , V is the peak to peak Calibration Signal voltage, f is the calibration signal frequency, R is the calibration resistor and K is the feedback coil constant. [See Appendix for R & K values]
Velocity Sensitity at the required frequency is obtained by dividing the Output Response Voltage by the Equivalent Input Velocity.
The obtained Velocity Sensitivity is used to convert the recorded Volatge PSD to Velocity PSD as shown below. The obtained results are compared to gloabl high noise model [NHNM] and USGS New Low Noise Model [NLNM,Peterson 1993] which gives the lowest observed vertical seismic noise levels across the seismic frequency band. Plot legend NLNM shows both the high & low levels.
Guralp A [X Arm] Low Velocity Output
Guralp B [Y Arm] Low Velocity Output
DTT Power Spectrum
Both the Seismometers were connected to the 40 M Control and Data Acquisition System (CDS) and Power Spectrum was estimated for the Vertical, North/South & East/West Channels using Diagnostic Test Tool (DTT) software.
CMG-40T Guralp A Calibration Sheet
Calibration Resistor: 51000
CMG-40T Guralp B Calibration Sheet
Calibration Resistor: 51000
Temporary fix for the switch: give a bit of oil to the button
Permanent fix: buy better switches.
Kiwamu and I brought 2 SUPER MICRO PCs from Willson house into 40m.
Both PCs are hooked up into Martian network. One is named as bscteststand for BSC which has been set up by Cds people and another one is named kami1 for temporary use for CLIO which is a bland new, no operating installed PC. This bland new PC will be returned Cds or 40m once another new PC which we will order within several days arrives.
IP address for each machine is 22.214.171.124 and 126.96.36.199 respectively.
We have installed CentOS5.2 into the new PC.
This morning there was a confliction of tpman running on fb40m and kami1. Alex fixed it temporary but Rana suggested it was better to move both PCs outside martian. We moved both PCs physically to the control room and connected to general network with a local router. I believe it won't conflict anymore but if you guess these PC might have trouble please feel free to shutdown.
Today's work summary:
*connected expansion chassis to bscteststand
*obtained signals on dataviewer, dtt for both realtime and past data on bscteststand with 64kHz timing signal
Excitation channels are not shown, only "other" is shown.
qts.mdl should run with 16kHz but 16kHz timing causes a slow speed on dataviewer and failing data aquisition on dtt. We are using 64kHz timing but is it really correct?
I borrowed SR785 to measure AA, AI noise and TF.
We measured Open loop TF for oplev pitch on ITMX.
All feed back filter of oplev was on as same as before. Original notch filters which notches above 10Hz resonance should be modified with some measurements of present resonant frequency. Up to 10Hz, a simple f^2 filter is used, so the notch should not affect this measurement.
Measured upper UGF is about 2Hz with gain slider 1, and lower UGF is 1.3Hz. Phase margin is 40 degree, so it is not a good idea to increase the gain drastically.
I have measured the coherence also but I could not find a way to put it on this picture. Anyway coherence below 0.6Hz was not so good like ~0.95. This can be improved if larger excitation is used next time.
During this measurement around 0.2-0.3Hz, small earthquake happened but seemed OK for the control.
We will measure the other TF, yaw, ETMX or somthing, maybe tomorrow, due to free swinging ITMX and ETMX tonight.
After Kiwamu had set the free swinging mode for ITMX and ETMX, I found a big jump of ITMX pitch and yaw. This jump is shown on oplev and OSEM plots.
I talked with Kiwamu on the phone that a shutdown of suspensions does not add a big offset, and so that it should not make a big jump.
We were not sure that this jump was due to the shutdown or drift or something else. Anyway I put ITMY oplev on center again at 0;57am.
In this morning, the same thing happened but to opposit direction when Kiwamu activated ITMX and ETMX. Then it turned out that 1000ct offset was existing on pit of ITMX. Erasing the offset fixed ITMX to normal position.
However a big drift exists in 11hours plot on ITMX 0.1->-0.25 at OLPIT, -580->-605 at SUSPIT and 0.08->0.15 at OLYAW, no significan drift at SUSYAW. On the other hand, ETMX has no big drift but has 10-30minites order fluctuations.
After 6am both the drift and the fluctuation became, roughly saying, 10 times larger, probably due to the human activity.
This plot shows ETM oplev and OSEM trend for 10 hours on day before yesterday as almost the same as plot shown this entry. I reported the 10-30minites fluctuations were seen, but I noticed it comes from not suspension but from oplev power fluctuation.
After Kiwamu fixed the ETM OSEM touch yesterday afternoon, still the same trend was seen, so we had thought what we fixed was not enough. This morning I looked at the yesterday's and day before yesterday's trend and noticed the simila trend both the pit and yaw in ETM oplev but not on the OSEM trend. Kiwamu suggested me to put the oplev sum on the same plot. It was!
So, ETMX is not bad, but in fact, still alignment fluctuation exist on the cavity. ITM?
This graph shows 5 hours data in minute trend for ITMX and ETMX from 5am to 10 am today. ITM pitch drift is 3 times lager than ETM pitch if the OSEM sensitivity is assumed to be the same.
This graph is last 1 hour data of above graph in second trend.
It is clealy seen that ITM yaw is jumping between two stages. I guess ITM is something wrong, touching magnets or earthquake stops?
According to c1scy.mdl, OL signals should be connected to adc_0_24 to adc_0_27 but they were connected to adc_0_16 to adc_0_19 which are assigned to QPD signals.
Actually cable connections were messed up. One ribbon cable was connected from QPD driver and ADC ports assigned for OL, and another ribbon cable was connected from the board combining the signals of oplev and QPD to ADC port assigned for QPD.
Now ETMY oplev is working well and aligned to center.
I got a new adapter board for expansion chassis from CDS and exchanged the existing adapter board which was laid on the floor around ETMX to new one.
Then I connected the chassis to c1lcs, c1lsc seems to be running now. I will return the old board to CDS since Rolf says he wants to return it to manufacture.
I found an interface box from ADC to D-SUB37pin and a cable to connect them.
I needed to make cables to connect the interface box to existing LSC whitening filters that has a 37 pin female D-SUB connector on one end and a 40pin female flat connector on the other end. We should use shielded cables for them, but unfortunately CDS did not have right one. Temporarily I made one cable for 1-8ch using a ribbon twist cable like Joe did.
I found a saturation at ch5 of ADC0 on c1lsc. I did not check carefully but it seemed to come from the LSC whitening board. Input of ch5 of the whitening board was not terminated and had a huge output voltage, but also ch6 was not terminated and had no big output. I guess something wrong on the LSC whitening board. Needs to be checked. Anyway I unplugged the small ribbon cable between the whitening board and the next LSC AA filter board.
Finally I realized that fiber connection of RFM did not exist. What I saw was the fiber cable of Dolphin. We need a RFM PCIe interface board, and a long fiber cable between c1lsc and RFM hub.
I'll be gone to Hanford site next week and come back to Caltech on 24th's week.
I setup a standalone RT system at the desk around circuit stock in the 40m.
Please leave this setup until I come back. I'll keep working when I come back.
I implemented a slow servo for green laser thermal control on c1scx.mdl. Ch6,7 of ADC and ch6 of DAC are assigned for this servo as below;
Ch6 of ADC: PDH error signal
CH7 of ADC: PZT feedback signal
CH6 of DAC: feedback signal to thermal of green laser
Note that old EPICS themal control cable is not hooked anymore.
I made a simple MEDM screen(...medm/c1scx/master/C1SCX_BCX_SLOW.adl) linked from GREEN medm screen (C1GCV.adl) on sitemap.
During this work, I noticed that some of the epics switch is not recovered by autoburt. What I noticed is filter switch of SUSPOS, SUSPIT, SUSYAW, SDSEN, and all coil output for ETMX.
I had no idea to fix them, probably Joe knows. I guess other suspensitons has the same problems.
I calibrated noise spectrum of green lock.
1. Measurement of conversion factor of ADC input from V to ct:
As a preparation, first I measured a conversion factor at ADC input of C1;GCX1SLOW_SERVO1.
It was measured while the output of AI ch6 as the output of C1;GCX1SLOW_SERVO2 with 1Hz, 1000ct(2000ct_pp) was directly connected into AA ch7 as the input of C1;GCX1SLOW_SERVO1. Amplitude at the output at AI ch6 was 616mVpp measured by oscilloscope, and C1;GCX1SLOW_SERVO1_IN1 read as 971.9ct_pp. So the conversion factor is calculated as 6.338e-4[V/ct].
2. Injection of a calibration signal:
When Green laser was locked to cavity with fast PZT and slow thermal, I injected 100Hz, 1000ct EXC at ETMX ASL. The signal was measured at C1:GCX1SLOW_SERVO1_IN1 as 5.314ct_rms. It can be converted into 3.368e-3Vrms using above result, and then converted into 3368Hz_rms using PZT efficiency as 1MHz/V. This efficiency was obtained from Koji's knowledge, but he says that it might have 30% or higher error. If somebody get more accurate value, put it into the conversion process from V to Hz here.
Frequency of green f=c/532nm=5.635e14[Hz] is fluctuating with above 3368Hz_rms,so the fluctuation ratio is 3368/5.635=5.977e-12, and it corresponds to length fluctuation of 37.5m. So, cavity fluctuation will be 5.977e-12*37.5=2.241e-10m_rms by 100Hz, 1000ct EXC at ETMX ASL.
Finally, we knew 5.314ct corresponds to 3368Hz and 2.241e-10m, so conversion factor from ct to Hz and ct to m are ;
633.8[Hz/ct] @ C1:GCX1SLOW_SERVO1
4.217e-11[m/ct] @ C1:GCX1SLOW_SERVO1
You can measure green noise spectrum at C1;GCX1SLOW_SERVO1_IN1 during lock, and mutiply above result to convert Hz or m.
This calibration is effective above corner frequency of slow and fast servo around 0.5Hz and UGF of fast servo around 4kHz.
I show an example of calibrated green noise.
Each color show different band-width. Of course this results of calibration cactor does not depend on band-width. Noise around 1.2Hz is 6e-8Hz/rHz. It sounds a bit too good by factor ~2. The VCO efficiency might be too small.
Note that there are several assumptions in this calibration;
1. TF from actual PZT voltage to PZT mon is assumed to be 1 in all frequency. Probably this is not a bad assumption because circuit diagram shows monitor point is extracted PZT voltage directly.
2. However above assumption is not correct if the input impedance of AI is low.
3. As I said, PZT efficiency of 1MHz/V might be wrong.
I also measured a TF from C1:SUS-ETMX_ALS_EXC to C1:GCX1SLOW_SERVO1_IN1. It is similar as calibration injection above but for wide frequency. This shows a clear line of f^-2 of suspension.
Files are located in /users/osamu/:20110127_Green_calibration.
Hi 40m people,
As Rana is saying, the bounce mode does not matter, or we cannot do anything. Generally speaking, the bounce mode cannot be damped by the setting of 40m SUS. Some tweak techniques may damp a bounce mode by res-gain or something, but it is not a proper way, I think.
This is also that Rana is already saying that the important thing is to find a good direction of OSEM to hit the LED beam to the magnet. Even if the magnet is not located at the center of OSEM hole, still you can find the optimal orientation of OSEM to hit the LED beam to the center of magnet by rotating the OSEM.
I know only an old document of T040054 that Shihori summarized how to adjust the matrix at the 40m. Too bad input/output matrix may introduce some troubles, but even roughly adjusted matrix should be still fine.
I will be at Caltech on 12-14 of September. If I can help something, I am willing to work with you!
I found the PSL table left open, and unattended again.
As far as I know, Jamie and Jenne (working on the LSC rack, so no lasers / optics work involved) have been the only ones in the IFO room for several hours now.
I'm going to start taking laser keys, or finding other suitable punishments. Like a day of lab cleanup chores or something. Seriously, don't leave the PSL table open if you're not actively working on it.
Given the similarities between the MDT694B (single channel piezo controller) and TC200 (temperature controller) serial interfaces, I added the pyserial driver here.
*Warning* this first version of the driver remains untested
We went into 40m to identify where XARM PDH loop control elements are. We didn't touch anything, but this is to note we went in there twice at 10 AM and 11:10 AM.
Updated IOO.strip on Zita to show WFS2 pitch and yaw trends (C1:IOO-WFS2_PIY_OUT16 and C1:IOO-WFS2_YAW_OUT16) and changed the colors slightly to have all pitch trends in the yellow/brown band and all yaw trends in the pink/purple band.
No one says, "Here I am attaching a cool screenshot, becuz else where's the proof? Am I right or am I right?"
Mon May 24 18:10:07 2021 [Update]
After waiting for some traces to fill the screen, here is a cool screenshot (Attachment 1). At around 2:30 PM the MC unlocked, and the BS_Z (vertical) seismometer readout jumped. It has stayed like this for the whole afternoon... The MC eventually caught its lock and we even locked XARM without any issue, but something happened in the 10-30 Hz band. We will keep an eye on it during the evening...
Tue May 25 08:45:33 2021 [Update]
At approximately 02:30 UTC (so 07:30 PM yesterday) the 10-30 Hz seismic step dropped back... It lasted 5 hours, mostly causing BS motion along Z (vertical) as seen by the minute trend data in Attachment 2. Could the MM library have been shaking? Was the IFO snoring during its afternoon nap?
I borrowed the little red cart 🛒 to help clear the path for new optical tables in B252 West Bridge. Will return once I am done with it.
After sliding the alignment bias around and browsing through elog while searching for "stuck" we concluded the ITMX osems needed to be freed. To do this, the procedure is to slide the alignment bias back and forth ("shaking") and then as the OSEMs start to vary, enable the damping. We did just this, and then restored the alignment bias sliders slowly into their original positions. Attachment 1 shows the ITMX OSEM sensor input monitors throughout this procedure.
At the end, since MC has trouble catching lock after opening PSL shutter, I tried running burt restore the ioo to 2021/Jun/17/06:19/c1iooepics.snap but the problem persists
Physically rebooted c0rga workstation after failing to ssh into it (even as it was able to ping into it...) the RGA seems to be off though. The last log with data on it appears to date back to 2020 Nov 10, but reasonable spectra don't appear until before 11-05 logs. Gautam verified that the RGA was intentionally turned off then.
We tested the CM board by implementing the high bandwidth IR lock (single arm). In preparation for this test we temporarily connected the POY11_Q_MON output to the CM board IN1 input and checked the YARM POY transfer function by running the AA_YARM_TEMPLATE under users/Templates/LSC/LSC_loops/YARM_POY/. We made sure the YARM dither optimized TRY so as to maximize the optical gain stage. Then we proceeded as follows:
Ultimately, our ability to progressively increase the control bandwidth of the YARM is a proxy that the CM board is working properly. Attachment 1 shows the OLTF progression as we increased the loop's UGF. Note how as we approached the maximum measured UGF of ~ 22 kHz, our phase margin decreased signifying poor stability.
At the end of this measurement, at about ~ 15:45 I restored the CM board IN1 input and disconnected the POY11_Q_MON
gautam: the conclusion here is that the CM board seems to work as advertised, and it's not solely responsible for not being able to achieve the IR handoff.
[yehonathan, anchal, paco, gautam]
We concluded estimating the XARM and YARM losses. The hardware configuration from yesterday remains, but we repeated the measurements because we realized our REFL55_I_ERR and REFL55_Q_ERR signals representing the PD520 and MC_TRANS were scaled, offset, and rotated in a way that wasn't trivially undone by our postprocessing scripts... Another caveat that we encountered today was the need to add a "macroscopic" misalignment to the ITMs when doing the measurement to avoid any accidental resonances.
The final measurements were done with 16 repetitions, 30 second duration, and the logfiles are under scripts/lossmap_scripts/armLoss/logs/20210722_1423.txt and scripts/lossmap_scripts/armLoss/logs/20210722_1513.txt
Finally, the estimated YARM loss is 397 ppm, while the estimated XARM loss is 388 ppm. This is consistent with the inferred PRC gain from Monday and a PRM loss of ~ 2%.
Future measurements may want to look into slow drift of the locked vs misaligned traces (systematic errors?) and a better way of estimating the statistical uncertainty (e.g. by splitting the raw time traces into short segments)
[gautam, yehonathan, paco]
We went back to the loss data from last week and more carefully estimated the ARM loss uncertainties.
Before we simply stitched all N=16 repetitions into a single time-series and computed the loss: e.g. see Attachment 1 for such a YARM loss data. The mean and stdev for this long time series give the quoted loss from last time. We knew that the uncertainty was most certainly overestimated, as different realizations need not sample similar alignment conditions and are sensitive to different imperfections (e.g. beam angular motion, unnormalizable power fluctuations, etc...).
Today we analyzed the individual locked/misaligned cycles individually. From each cycle, it is possible to obtain a mean value of the loss as well as a std dev *across the duration of the trace*, but because we have a measurement ensemble, it is also possible to obtain an ensemble averaged mean and a statistical uncertainty estimate *across the independent cycle realizations*. While the mean values don't change much, in the latter estimate we find a much smaller statistical uncertainty. We obtain an XARM loss of 37.6 2.6 ppm and a YARM loss of 38.9 0.6 ppm. To make the distinction more clear, Attachment 2 and Attachment 3 the YARM and XARM loss measurement ensembles respectively with single realization (time-series) standard deviations as vertical error bars, and the 1 sigma statistical uncertainty estimate filled color band. Note that the XARM loss drifts across different realizations (which happen to be ordered in time), which we think arise from inconsistent ASS dither alignment convergence. This is yet to be tested.
For budgeting the excessive uncertainties from a single locked/misaligned cycle, we could look at beam pointing, angular drift, power, and systematic differences in the paths from both reflection signals. We should be able to estimate the power fluctuations by looking at the recorded arm transmissions, the recorded MC transmission, PD technical noise, etc... and we might be able to correlate recorded oplev signals with the reflection data to identify angular drift. We have not done this yet.
[yehonathan, anchal, paco]
Yesterday around 9:30 pm, we centered the BS, ITMY, ETMY, ITMX and ETMX oplevs (in that order) in their respective QPDs by turning the last mirror before the QPDs. We did this after running the ASS dither for the XARM/YARM configurations to use as the alignment reference. We did this in preparation for PRFPMI lock acquisition which we had to stop due to an earthquake around midnight
We picked up AS WFS comissioning for daytime work as suggested by gautam. In the end we want to comission this for the PRFPMI, but also for PRMI, and MICH for completeness. MICH is the simplest so we are starting here.
We started by restoromg the MICH configuration and aligning the AS DC QPD (on the AS table) by zeroing the C1:ASC-AS_DC_YAW_OUT and C1:ASC-AS_DC_PIT_OUT. Since the AS WFS gets the AS beam in transmission through a beamsplitter, we had to correct such a beamsplitters's aligment to recenter the AS beam onto the AS110 PD (for this we looked at the signal on a scope).
We then checked the rotation (R) C1:ASC-AS_RF55_SEGX_PHASE_R and delay (D) angles C1:ASC-AS_RF55_SEGX_PHASE_D (where X = 1, 2, 3, 4 for segment) to rotate all the signal into the I quadrature. We found that this optimized the PIT content on C1:ASC-AS_RF55_I_PIT_OUT and YAW content on C1:ASC-AS_RF55_I_YAW_OUTMON which is what we want anyways.
Finally, we set up some simple integrators for these WFS on the C1ASC-DHARD_PIT and C1ASC-DHARD_YAW filter banks with a pole at 0 Hz, a zero at 0.8 Hz, and a gain of -60 dB (similar to MC WFS). Nevertheless, when we closed the loop by actuating on the BS ASC PIT and ASC YAW inputs, it seemed like the ASC model outputs are not connected to the BS SUS model ASC inputs, so we might need to edit accordingly and restart the model.
[anchal, yehonatan, paco]
For whatever reason (i.e. we don't really know) the MC unlocked into a weird state at ~ 10:40 AM today. We first tried to find a likely cause as we saw it couldn't recover itself after ~ 40 min... so we decided to try a few things. First we verified that no suspensions were acting weird by looking at the OSEMs on MC1, MC2, and MC3. After validating that the sensors were acting normally, we moved on to the WFS. The WFS loops were disabled the moment the IMC unlocked, as they should. We then proceeded to the last resort of tweaking the MC alignment a bit, first with MC2 and then MC1 and MC3 in that order to see if we could help the MC catch its lock. This didn't help much initially and we paused at about noon.
At about 5 pm, we resumed since the IMC had remained locked to some higher order mode (TEM-01 by the looks of it). While looking at C1:IOO-MC_TRANS_SUMFILT_OUT on ndscope, we kept on shifting the MC2 Yaw alignment slider (steps = +-0.01 counts) slowly to help the right mode "hop". Once the right mode caught on, the WFS loops triggered and the IMC was restored. The transmission during this last stage is shown in Attachment #1.
Yesterday we discussed a bit about working on the PRMI sensing matrix.
In particular we will start with the "issue" of non-orthogonality in the MICH actuated by BS + PRM. Yesterday afternoon we played a little with the oscillators and ran sensing lines in MICH and PRCL (gains of 50 and 5 respectively) in the times spanning [1312671582 -> 1312672300], [1312673242 -> 1312677350] for PRMI carrier and [1312673832 -> 1312674104] for PRMI sideband. Today we realize that we could have enabled the notchSensMat filter, which is a notch filter exactly at the oscillator's frequency, in FM10 and run a lower gain to get a similar SNR. We anyways want to investigate this in more depth, so here is our tentative plan of action which implies redoing these measurements:
Task: investigate orthogonality (or lack thereof) in the MICH when actuated by BS & PRM.
1) Run sensing MICH and PRCL oscillators with PRMI Carrier locked (remember to turn NotchSensMat filter on).
2) Analyze data and establish the reference sensing matrix.
3) Write a script that performs steps 2 and 3 in a robust and safe way.
4) Scan the C1:LSC-LOCKIN_OUTMTRX, MICH to BS and PRM elements around their nominal values.
5) Scan the MICH and PRCL RFPD rotation angles around their nominal values.
We also talked about the possibility that the sensing matrix is strongly frequnecy dependant such that measuring it at 311Hz doesn't give us accurate estimation of it. Is it worthwhile to try and measure it at lower frequencies using an appropriate notch filter?
Wed Aug 11 15:28:32 2021 Updated plan after group meeting
- The problem may be in the actuators since the orthogonality seems fine when actuating on the ITMX/ITMY, so we should instead focus on measuring the actuator transfer functions using OpLevs for example (same high freq. excitation so no OSEM will work > 10 Hz).
Thu Aug 12 11:04:42 2021 Arrived to find the PSL shutter closed. Why? Who? When? How? No elog, no fun. I opened it, IMC is now locked, and the arms were restored and aligned.