Sheila. Alexa.
When we got here this morning the ETM ISI was tripped. We decided to use the chance to roughly aling the cavity (using the SUS offsets) with it tripped, and reset the CPS targets. Hopefully this will reduce the ringing up of trilliums when we isolate by reducing the amount that the ISI has to move. We only did this for stage 1, since stage 2 was already pretty cose to the target values. Before and after screen shots of what we started with is attached. The ! Store target offsets button didn't seem to do anything, so I copied the numbers by hand.
After this I tried to follow hugh's alog 9926
first I set all the blends to 750mHz, then tried to isolate stg1 level3. However, when I tried to move the blends to tcrappy it kept tripping, so now I have it at level 2 and tcrappy doesn't trip it. On stage 2, I tried level 2 (starting with 750 blends) moving to tcrappy would cause it to trip. Stage 2 level 2 tripped, 1 when I tried to change from 750 to Tcrappy, then even when I left it at 750. Now it is at level 1 with 750 blends.
Modified black glass holders worked only for some Siskiyou mounts, not all.
Some screw heads are sticking out more than the others. I could have forced it really hard and probably it would have fit, but I didn't want to risk changing the alignment. Another round of clean milling job was done, parts will go to air bake again on Monday.
TMS was balanced nicely at the level good for IAS.
We put everything except the black glass holders on the table (we put black glasses themselves directly on the table at their approximate positions).
Added two 200 grams masses to compensate for the fact that we removed Hartman reference path mirrors and mounts (480 grams together). Combined with 8 pieces of black grasses, this should be close enough.
We changed the cable bracket position to prevent cable interference.
After these, we balanced both the top mass and the table by using slide masses.
No signal from TMS BOSEMs.
BOSEM cables were connected but I don't see anything in the digital world.
Ideally, we'd like IAS to work while TMS is damped.
This morning Richard powered up the TMSy AA chassis and I plugged in the satellite box end of the test cables.
TMS Cable Installed
Unfortunately, we installed the Lower Structure before installing a cable, so an easy job was made much harder, but it was done. Our damaged D1000223 cable for the Beam Diverter was replaced with a D1000225 s/nS1106887. This is a variance from the Cable Harness Routing Config (D1200111).
Stefan, Lisa It is still not great, but we could stay locked on 3f for more than half an hour. We intentionally unlocked by letting Sheila aligning the arm. The sideband power dropped significantly during the lock, by about 30% (neither dither nor alignment loops running yet).
Kiwamu, Lisa, Stefan We increased the 9MHz modulation depth by 3dB. This was done by increasing the IFR drive from 10dBm to +13dBm, and putting 3dB attenuators in all path except the one for the 9MHz EOM. This should raise the 9MHz modulation index from 0.1 (alog 9395) to 0.14. We expect a 9dB stronger signal for the 27MHz. Indeed with this setting we were able to get a pretty stable lock on the 3f signal (its locked as I type this.) Some of the settings changes were: - PRCL gain -0.4 - MICH gain 40 - BS M2 LOCK L: added the comb60 (FM4) and ELP90 (FM9) in addition to the f^2 (FM6) and z30:p100 (FM7) - We didn't need any additional filters for PRM M2 LOCK L (We only have FM3 FM4 FM8 and FM10). - We notices the occasional saturation in the REFL27 input, so we lowered the whitening gain to 21dB (val=7), and increased the digital gains from 3 to 4.243. Whitening stages 1 and 2 are also on. - We also noticed that it prefers slightly hight MICH and PRCL gain for acquisition (we used -0.6 and 60) - Also, we noticed that the REFLAIR45 (which currently has 0dB whitening gain and only whitening stage 2) tends to saturate during lock acquisition. It seemed to lock quicker without any whitening filters on. - Even though CDS doesn't report any saturations, we occasionally see the BS optical levers get kicked and ring down. (The BS has optical lever damping.) The attached plot shows the raw signals of REFLAIR 45 and REFLAIR 27, as well as the drive to the BS M2. (The signals are scaled to gain-match.) PS: it is still locked on 3f - 16min and counting.
Last night before leaving we realized that some of the changes made to the locking script by Kiwamu during the day would make the guardian unhappy. Today Stefan and I put back the version which was working yesterday morning, so the changes that Kiwamu made are not running at the moment (but we save the script as LSC_PRMIsb_kiwamu.py).
Evan, Lisa, Kiwamu,Stefan Tonight we fixed our PRC oscillation problem: it was due to a demod phase that was 11deg off. Combines with the fact that MICH is only acting on the BS, this messed up our loop shape. We carefully phased REFL_45 to 144.6deg, and lowered the MICH gain to 50 and PRCL GAIN TO -0.6. (All these gains are in guardian now.) Next we went on to 3f locking. We phased the 27MHz by simply maximizing the signal in I (we needed 90deg). Switching PRCL to 27_I was straight forward. For MICH we wanted to user the 135MHz signals, but the signal seemed dead. Thus we tired 27_I. The transition worked fine, but we are still saturating the BS around 30-40Hz, and typically loose lock after 20sec or so. As a result Lisa was disappointed: despite cooking dinner for us, be didn't quite achieve a stable lock on 3f tonight. But we promised her that we'll increase the modulation depth and the light of the 3f diode.
Also, we implemented a new state in the PRMI guardian for the 3f locking. It transitions the sensor from the 1f signals to the 3f signals by ramping the sensing matrix elements. This works fine although the way it does currently is a rough discrete gain step.
Another thing we would have to mention is that we could not bring HAM3 ISI back to its level 3 isolation. We tripped the ISI when we accidentally tripped PR2. We tried the isolation script from the medm screen but it seems that it tries to enable a couple of blank filters and stops at some point before ramping up the isolation gains. This needs to be revisited. Currently it is only damping.
Lisa, Kiwamu
Today, we looked at free swinging wave form of the 3f signals. The signals in REFLAIR_B_RF135 was visible and we confirmed that the signal size made sense by comparing it with the RF27 signals.
Here are our back of envelope calcuation:
Therefore the RF27 and RF135 should be almost the same signal level. Indeed, we see almost the same size (i.e. peak-to-peak) of the signals in both RF27_I and RF135_I which showed 700 counts p-p and 800 counts p-p respectively without any whitening gains or whitening stages. We didn't check the absolute value at this point.
Evan, Stefan, Kiwamu, Lisa There has been some confusion about factors of two in previous entries, so here are our "final" numbers: - Pin = 8.8 W entering the IMC, measured on the PSL table - with PRM aligned, R = 98.6% - from Paul's calibration of the MC2Trans diode (we measured 645 uW), it looks like there is T = 88% from the PSL to the transmission of the input mode cleaner; - given the HAM1 path, we expect 1.25% of the REFL power to reach the LSC RF diode P_exp = 8.8 * 0.986 * 0.88 * 1.25e-2 = 95 mW; LSC in vac REFL A (LSC RF PD) (1.25% of REFL power) = 15900 counts (maximum found by moving RM2) ==> 15900 [counts] * (1/1638) [V/counts] * 2 [diff to single input] * (1/200) [1/Ohm] * 0.8 [W/A] = 78 mW LSC REFL AIR A (LSC RF PD) (0.625% of REFL power) = 7955 counts ==> 32.5 mW measured directly in front of the diode on table ==> 7955 [counts] * (1/1638) [V/counts] * 2 [diff to single input] * (1/200) [1/Ohm] * 0.8 [W/A] = 39 mW LSC REFL AIR B (BBPD) (0.625% of REFL power) = 21770 counts ==> 34 mW measured directly in front of the diode on table ==> Conclusion : The discrepancy between what we see and what we expect is not a factor of 2, but more like 20%-30%, and in vac and in air diodes gives us consistent numbers.
Nope..we didn't get the right factors this time either: W/A = 1/0.8, and the factor of 2 diff to single is not there (and, even if it doesn't matter, R for PRM is 97.3%, not 98.6%). I will post another entry with the right numbers..somehow I can't edit this one.. P.S: Still need to calibrate these numbers, measurements were taken all at the same time. It might be also worth to check if the T = 88% transmission from PSL to IMC throughput is correct. LSC REFL AIR B (BBPD) (0.625% of REFL power) = 21770 counts ==> 34 mW measured directly in front of the diode on table WFS in vac REFL A (0.625% of REFL power) = 34200 counts WFS in vac REFL B (0.625% of REFL power) = 36400 counts
Left these ISIs in following State
HEPI position loops on
Stage2 ISI in Lvl3 control with 750mHz blends for BS & ITMY and 250mHz for all dofs sans 100mHz for X & Y on the ITMX.
Stage1 ISI in Lvl3 control with T250 blends everywhere except: T100mHz.44NO on BS & T40mHz.44NO on ITMY or the X & Y dofs. Stage1 of the ITMX blends w/ TCrappy everywhere.
For the ETMX, the blends are all TCrappy, Stage1 is in Lvl3 control while Stage2 is in Lvl2 control. We are still getting some peaks slowly ringing up things out past the UUG which require notches for the Control3
Just found the BS Stage2 tripped back to damping only. Sounds like the Michelson Feedback forces our Actuators to their limit.
Alexa, Daniel, Sheila
Today we were able to lock COMM stably. (Thanks to some extra time from the red team and SEI people). By COMM I mean that we can lock the PSL to the green arm transmitted power, using the mode cleaner as an actuator. I don't want to call it the CARM handoff as we called it durring HIFOY, to make a distinction between this ALS handoff and when we do the handoff of the common arm degree of freedom to corner IR signals from the ALS signals.
We have 2 scripts that do the handoff in the new way, userapps/als/h1/scripts COMM_handoff and COMM_down_2
COMM_down_2 brings COMM down in a nice way that does not unlock the mode cleaner. We will work on getting this into guardian at some point. We will try simplifying the COMM handoff script in the future.
We are now using a crossover at 15 Hz, 23 degrees of phase margin (plot attached). This allows us to engage the two notches in the MC2ISCINF (at 29 and 40 Hz) that prevent us from ringing up suspension modes. We struggled to get the crossover this low, because the slow path losses phase at low frequencies much faster than we understand. We added into LSC CARM a z5p20 to get some phase back, and also lowered our compensation filter (6) zero to 6Hz. We are also no longer using the z40:p150 which is used by the mode cleaner locking in MC2M3.
The ugf is at 10s of kHz, Alexa has plots she will post. The open loop gain shows a peak at around 27kHZ. When we looked into why this was, we saw that the PLL has almost no phase margin.
Anyway, the good news is we can reliably do the handoff, and it seems to stay locked at least as long as we left it alone, which wasn't more than 10 minutes or so.
Settings after the handoff:
COMM PLL
IMC Servo Board
CM Servo Board
LSC:REFL_SERVO_SLOW Filter
LSC-CARM Filter
MC2_M3_ISCINF Filter
MC2_M3_LOCK_L Filter
MC2_M2_LOCK_L Filter
Note: no gains present in any of the input matrices
Measurements:
SCRN0110.TXT, SCRN0111.TXT, SCRN0112.GIF are the magnitude, phase, and pic of the COMM PLL open loop transfer function with the above settings.
SCRN0102.TXT, SCRN0103.TXT, SCRN0101.GIF are the magnitude, phase, and pic of the CM common path open loop transfer function with the handoff (case of MC2_M3 filter FM3 On)
SCRN0104.TXT, SCRN0105.TXT, SCRN0106.GIF are the magnitude, phase, and pic of the IMC open loop transfer function without the handoff (i.e. above settings w/ input 2 enable: OFF)
SCRN0108.TXT, SCRN0109.TXT, SCRN0107.GIF are the magnitude, phase, and pic of the IMC open loop transfer function with the handoff (i.e. above settings, case of MC2_M3 filter FM3 Off)
See attached for test progress and final result (Red Traces). The values for these corrections are 0.0015 & 0.00075 for XtoRX & XtoRY respectively.
The values for ITMY Y-Tilt are -0.0007 & 0.0005 for YtoRX & YtoRY. The second plot is for this drive axis. The plot isn't as busy or interesting but the results are slightly better in the cross term.
I'll attach the MEDM to just cause I can.
I'll try to get measurements to show the results that Rich & I have gotten on the other ISIs over the this frenzied week.
for future reference, attached are the new offsets of MC corresponding to the new calibration calculated by Paul, cf
https://alog.ligo-wa.caltech.edu/aLOG/index.php?callRep=9870
[Fil Arnaud]
The upper left osem of the ETMX PUM remains noisy : its noise floor is higher than LR by a factor of ~5 @10hz, see red curve vs green of 1st pdf.
To make sure it didn't come from some electronics noise, Fil unplugged the cable between the vacuum flange and the PUM. The noise floor went down by a factor of 15 @10hz, meaning it's definitely not coming from any of the electronics. cf blue of 2nd pdf before unplugging vs green after. The flag might be not centered as well as the other osems
8:50-9:31 Mirror Mounts work in H2 LAE (LVEA)– R. Savage 8:57-12:00 Going to End X to do documentation work – Luis 9:20 PSL Check List: All the parameters were OK except the FSS Threshold on transmitted photo-detector PD = 0.49V (should be 0.9V) 9:25-9:50 Repair person on site to fix the tractor 9:30- Going to END Y to do TMS work – Corey/Keita 9:49-12:00 Work on SUS electronics (PUM Chassis) at END X – Filiberto 10:07-10:33 Searching for parts in LVEA – Thomas
I finally got a script working to step the alignment offsets on the IMC mirrors and record the transmitted power drop with MC2trans and IM4trans QPDs.
The idea behind this was to compare the quadratic function for power drop with misalignment with the theoretical function, giving us a means of accurately calibrating the alignment offsets.
The reason I'm interested in calibrating these offsets accurately is for beam jitter measurements using the coupling from jitter to RIN in transmission of a misaligned IMC [see e.g. LHO aLOG entry 8190]. The coupling factor is determined by the slope of the quadratic function, so we can't calibrate jitter measurements made in this way any better than we can calibrate the DC alignment offset.
The first attached plot shows the normalized transmitted power obtained for each individual mirror DOF, from both IM4trans QPD and MC2trans QPD, over "intended" alignment offset. Also included is a plot of the normalized transmitted power from a Finesse model of the IMC over "real" misalignment offset. From these plots we can see that in general the alignment offsets actually applied to the MC mirrors are larger than the intended alignment offsets. However, the measured data is not always symmetric (especially for IM4). This could be due to clipping at the QPDs. The centering on IM4trans is not as good as the centering on MC2trans, so I would be more confident in the numbers from MC2trans.
I fitted a quadratic function P=A(x-h)^2+k to each of the curves. The calibration is then done by scaling the alignment offsets applied to the actual suspensions by sqrt(Amodel/Adata). The second attached plot shows each DOF again, but this time with the x-axis scaled for the measured data to fit the model. I used the scaling factors calculated from MC2trans data since the centering on this QPD was better. For the most part I'd say the data matches the model well after this scaling.
These scaling factors are:
DOF | Scaling factor |
MC1 Pitch | 0.7043 |
MC1 Yaw | 0.8223 |
MC2 Pitch | 0.8572 |
MC2 Yaw | 0.8326 |
MC3 Pitch | 0.7823 |
MC3 Yaw | 0.8588 |
I would propose to include these scaling factors in the calibration of the MC mirror offsets.
The script can be run again at any time to check for any possible changes in e.g. the OSEM coil driver gains over time. It might be beneficial to take more data points at some point too, but the script takes 15mins or so to run as it is (mainly due to the time given for optics to settle between changes of offset). Another improvement would be to step MC1 and MC3 pitch over a larger range, since the transmitted power is actually fairly insensitive to these DOFs. Both these things can be edited in the top few lines of the script.
In case anyone is interested in running this script in future, it is located at opt/rtcds/userapps/release/ioo/h1/scripts/imc/pfulda/IMC_align_calibrate.py
Be sure to run the mcWFSrelieve script located in opt/rtcds/userapps/release/ioo/h1/scripts/imc/ first though!
I attach the analysis scripts here too, including the measured data, Finesse model and results, and other functions used.
In preparation for beam jitter measurements, I applied these calculated gains to the MC mirror M1 OPTICALIGN filter banks. Rather than just edit the gain directly, I made a new filter in each DOF's filter bank with the calculated gain. These filters are all called "alog9870" to point anyone towards the above entry for explanation. I then calculated the required new input offset values to retain the current alignment, and adjusted these while switching the filters on. The IMC is still aligned, but now the alignment offsets are calibrated to um using the data gathered from the IMC power drop measurement.
Last night I had a bit of time to run the script again for a more detailed measurement, with 21 alignment steps, waiting 15 seconds for alignment to stabilize between each, and also averaging PD data for longer than previously. I also increased the misalignment range from ±30urad to ±50urad for MC1 and MC3 pitch.
Since the alignment offsets are now calibrated based on the previous measurement, I was curious to see if the model now fits well to the new data without the step of adjusting the x-axis.
The attached plot shows the new measured data and the model, without any adjustment of x-axis scaling. I think they all agree pretty well, though the one thing that concerns me slightly is the apparent offset in MC2 pitch. Is it possible there is an offset somewhere in the WFS loop that causes this? If so, maybe we could try adjusting this to maximize transmitted power / minimize reflected power.
Just for completeness. the new MC opticalign offsets with this calibration included are:
DOF | Old offset | New offset |
MC1 P | 883.3 | 1254.2 |
MC1 Y | -1945.7 | -2366.2 |
MC2 P | 470.4 | 548.8 |
MC2 Y | 257.2 | 308.9 |
MC3 P | -430.6 | -550.4 |
MC3 Y | -2119 | -2467 |
The MC2 roll mode notch (FM6 in ISCINF) was left off overnight, resulting in a rung up 28Hz peak in MC_L, which was producing a huge signal in PRCL. Once on the peak slowly disapeard. PRMI locking is now quite reliable. However we still have relatively big power fluctuations. I recorded today's settings in a setup script: sballmer/tempfixes/setPRMI Really, this should be fed to a guardian.
I left the PRMI locked tonight.
PRMI stayed locked for 9h with only 1 lock loss in the middle. 1st lock: 2014/02/03 05:03:00 to 08:39:00 (duration 3h36min) 2nd lock: 2014/02/03 08:41:00 to 14:31:00 (duration 5h50min) All times UTC.
The coherence analysis using STAMP-PEM revealed strong correlation (0.5 Hz) between H1_SUS-ITMX_L3_OPLEV (YAW and PIT) and H1:ALS-X_ARM_IN1_DQ. The coherence plots produced by Patrick can be accessed from here. Josh did follow up studies confirming the presence of 0.5 Hz strucutre in ITMX oplev signals and also saw strong correlation between ITMX oplev and POPAIR signals. The coherence plots Josh made can be accessed from here.