james.batch@LIGO.ORG - posted 08:59, Monday 15 December 2014 (15615)
Restarted cds wiki computer
Restarted cds wiki computer, it was unresponsive.
Restarted cds wiki computer, it was unresponsive.
model restarts logged for Sun 14/Dec/2014
2014_12_14 14:27 h1fw1
2014_12_14 20:42 h1fw0
both unexpected restarts. Conlog frequently changing channels report attached.
In preparation for venting in the morning the Y end is laser safe.
As Richard requested I first removed the interlock jumper, the laser turned itself off. I've replace the jumper now, and left the key on top of the controller.
The end Y picomotors are software disabled.
Stefan, Sheila
We burt restored ETMX and ETMY sus models to december 3rd at 19:00 UTC. The difficulties with DIFF yesterday were caused by wrong filter settings probably due to a bad burt after the model restarts. We then had no problem locking ALS both DIFF and COMM, with the green WFS and tidal feedback on. The guardian is not yet controlling the tidal feedback correctly once ALS COMM and DIFF are locked. We need to think about how to get the common signal (from MC-L) to the tidal servo, possibly adding it in the front end.
Once we turned of the sensor correction DRMI locked well. We saw that the AS 36 signals were saturated so we changed the whitening gain (1 stage of whitening filters, gain of 30dB). We spent some time trying to close some loops on the SRC, we closed both AS_A 36I-> SRM in pitch and yaw. We were not convinced that either loop really works; when we misaling SRM, the error signal doesn't change much before we start mode hopping and the error signal becomes unusable. We also closed the AS_C -> SR2 pitch loop, while it did bring the beam to the center the SRM loops had truoble keeping up. We think the next step is to try using SRY and set the output matrix to a combination of SRM and SR2.
Stefan, Sheila
Today we had difficulty locking DRMI, we tried PRMI and found we had difficulty locking that as well. Then I found that I couldn't even keep MICH DARK locked. Stefan was able to keep it locked by turning the gain up to -1000, removing the limiter from the BS M3 ISCINF once it locks, and adding a boost (FM4 in LSC_MICH). I compared the control signal to what we saw on decmber 3rd at a time when DRMI was locked, and see that we have about a facotr of ten higher RMS, all due to motion below 0.2 Hz. I then tried turning off different parts of sensor correction, first just the ISI XY, then also the HEPI Z. turning off HEPI Z brought the RMS back down to the old value. For now I'm leaving it all off.
BS M2 DRIVEALIGN L2Y gain changed from -1 to -0.8. This significantly improved the BS DC balancing.
K. Venkateswara
Looks like Z sensor correction is producing excessive Rz (Yaw) motion of Stage 1 at all three CS BSCs, but particularly so at the Beamsplitter. Attached plot shows the ASD of the CPS_RZ from the same period. The REFs were taken when Z sensor correction (to HEPI) was ON for all three, and the normal traces were with it OFF. While the sensor correction reduces motion at ~0.15 Hz, it is unexpectedly inducing excess RZ in all three, but much more so in BS. The coherence plot on Page 2 shows that RZ of the BS was the reason the MICH was difficult to control.
I don't have an immediate answer but will investigate some more.
I think one solution to the above problem is to turn on Z to Rz subtraction and enable the Rz loop. Jim and I tried this last week briefly and the result appeared agreeable. So I tried it again at ITMX and the results are shown in the attached file. First page shows the CPS_RZ, second shows the T240_RZ and the third shows the OpLev_Yaw. Green curve is with no sensor correction, blue is with Z sensor correction to HEPI and the red is with Z sensor correction to HEPI and Z to Rz subtraction at Stage 1.
I will also try this at BS and ITMY chambers.
no restarts reported. Conlog frequently changing channels report attached.
I just discovered that the PLC2 ecat systems at both end stations were restarted. Last restart reported:
X1PLC2 16:08 12/13 2014
Y1PLC2 16:32 12/13 2014
With both WFS and tidal running on both arm, we quickly tried DIFF. It failed. We took out the clearing of the drivealign filters, which was the first culprit. But it still failed. Some more work is needed on DIFF. COMM however worked just fine.
The green slow and tidal servos have been engaged. This is done through the arm guardian which in turn select a corresponding state from the ALS auto-locker. These states are "End Locked", "Slow Engaged", "Transition" and "Red Locked". With "End Locked" neither the slow nor tidal servo are engaged. With "Slow Engaged" they are both on, tidal feedback only in "Transition" and tidal and common mode feedback in "Red Locked". These three new feedback paths use the new integrator filter module. When the green PDH is not locked, the integrators are bled off.
The feedback topology is as follows:
The "Red Locked" state triggeres on the transmitted red power. Once in this state, it no longer matters whether green is locked. It can only transition back, when the red lock is lost.
Here is a 1.33 hour trend plot. Both the arm and tidal error signals stay within a few µm of zero, whereas the HEPI drive slow ramps upwards to 8 µm. With the WFS enagaged the arm power is solid. (The jump towards the end is due to some stray light from the x arm.)
Here is a 21h trend plot of both green arms. WFS & tidal get a thumbs up.
Evan, Stefan Did some more WFS/camera loop work on both arms: - Added a 0.3Hz LP with a notch at the main pendulum modes into the WFS output filter modules. This allows for a ~3-5 times higher gain. - Carefully measured the output matrix for the camera loops (X and Y arm, yaw and pitch) by dragging the ETM around and letting the WFS follow. The new output matrix values are in the snaphots #3 below. With them the camera loop no longer relies on gain hierarchy - it's gain was increased ~x10. - Some gains were redistributed - see the attached snaps. - Cleaned up the arm Guardian. Removed all old slow feed-back stages, and instead added a state commanding Daniel's new setup. - Added a WFS relief step to the Guardian.
We noticed that the higher Camera gain can cause a too big kick during the WF engaging process. Thus we used the existing FM triggering to set up a 10 second delay. During those 10 seconds the DoF3 (P&Y) , i.e. the camera loops, have 20 times lower gain. This seems to take care of the engaging problem. The attached snapshot contains all elements that changed.
The experimentally measured output matrix to move TMS, ETM and ITM simultaneously to center the ITMX green camera (without misaligning the beam) is
X PIT: X YAW:
ITMX: 0.68 -0.43
ETMX: 1.00 1.00
TMSX: 1.43 1.00
Y PIT: Y YAW:
ITMY: 1.43 -1.23
ETMY: 1.00 1.00
TMSY: 1.54 1.15
The expected matrix would be
PIT:
ITM: 1.06 = -g1
ETM: 1.00
TMS: 1.00
YAW:
ITM: -1.06 = g1
ETM: 1.00
TMS: 1.00
This means that our angular actuation calibration is somewhat fishy...
For reference, the H1 installed optics radius of curvature are (from https://galaxy.ligo.caltech.edu/optics/, after coating measurements):
ETMX: ETM-08: RoC=2242m ± 2m, (LIGO-C1103233), g2= -0.7817
ITMX: ITM-03: RoC=1940m, (LIGO-C1103237), g1= -1.0590
X Arm length: L=3994.4704 m ± .3mm (alog 9626)
ETMY: ETM-12: RoC=2240 ± 2mm, (LIGO-C1103257), g2= -0.7832
ITMY: ITM-11: RoC=1939.4m, (LIGO-C1103255), g1= -1.0596
Y Arm length: L=3994.4691m ± .7mm (alog 11611)
model restarts logged for Fri 12/Dec/2014
2014_12_12 11:11 h1iscex
2014_12_12 11:14 h1iscex
2014_12_12 11:26 h1iscey
2014_12_12 11:28 h1fw1
2014_12_12 13:41 h1fw0
two unexpected restarts. ISC code changes at end stations (no daq restart needed). Conlog frequently changing channels report attached.
AlexaS, TravisS, RickS Yesterday (Friday) afternoon, we re-focused the Y-end Pcal beam localization camera for green light in preparation for next week's efforts to clean the ETM surface. The attached image was taken after re-focusing the camera, with the green light resonating in the Y-arm.
Dan, Evan
Some housekeeping for DRMI locking:
We wanted to reimplement the SRC WFS loops, but so far we are unable to get DRMI locks that are long enough to do anything useful with.
[Mackenzie, Evan, Paul]
Today we improved the alignment of the aux laser into HAM2, and after some searching around with the aux laser temperature control were able to observe the beat signal on the RFPD. The beat signal amplitude was around -45dBm at best, when we had 1.5V DC on the PD. I didn't check the difference between DC and AC gains for that pd, but we expect we should be able to get a significantly larger beat signal. We adjusted the aux laser alignment trying to improve the beat signal amplitude, but weren't able to get a big improvement. At this point we think that we are limited by the mode matching of the two beams: it's clear just looking at the card that there's at least a factor 2 in beam sizes at the IM4 forward trans beam path. Next step will be to make some quick mode measurements on the table and either adjust the lens currently in place or add some more to the path to get a better overlap.
Evan, Daniel, Stefan We duplicated the full WFS + camera auto-alignment scheme on the y-arm. In particular we - Restarted h1iscey.mdl to include the CAM.mdl change (SVN revision 9393) - Set the Y Green WFS autocentering UGF's to 10Hz. - Re-phased Y Green WFS using a 55Hz length dither. - Closed all loops with the same scheme as the x-arm (WFSA-->ITMY, WFSB-->TMSY, ITMY camera-->ETMY). The camera loop relies on gain hierarchy through the WFS loops. - The overall gain is slightly lower on the y-arm, but all seems to work. - Did a from-scratch baffle-PD alignment of TMSY, aligned the y-arm to it, updated the camera config file and took a camera reference. - All this again seemed to work fairly reliably. - All medm setting are shown in the attached snapshot (the PIT and YAW output matrices are identical). Still TBD: Same tweaking as for the x-arm: - Fine-tune the output matrix - in particular the DoF3 (camera path) should be set to move all 3 optic in the right way - this would avoid having to go through a gain hierarchy. - Increase the loop UGF: main pendulum notches are needed to avoid a 2nd unity gain crossing.
