patrick.thomas@LIGO.ORG - posted 18:30, Monday 25 June 2012 (3262)
plots of dust counts
Attached are plots of dust counts > .5 microns.
Non-image files attached to this report
Attached are plots of dust counts > .5 microns.
We're still aligning the arm, it's difficult because we cannot see the beam at all at this stage on the mirror.
Anyway, the beam goes to the ITM, and then to the ETM, then to the ITM and comes back to the ETM. We're seeing some signal from the demod board, which goes zero when the ITM is misaligned. So this is the real interference signal!
Unfortunately it's not like the arm is flashing or anything, I suspect that our alignment is off mainly in yaw because I can tilt the ETM in yaw and see the beam on both righ and left side of the ETM at about the same height.
Tomorrow we'll look at the interference pattern in the reflection from the arm, DC level of the reflection, and the demod signal amplitude to help our alignment effort. We'll also optimize the demod phase from Beckhoff world.
The alignment jumped twice by a large amount and it took us some time to recover.
Once it was supposedly Kyle hitting something, and ITM didn't come back exactly to the same position.
Another time I don't know what happened.
Kyle, Mark L. (Apollo)
- Installation of IO components on H1 laser table started today. Currently laser beam is dumped and we get no video signal, also some MEDM screens are red, for more information see work permit #3303.
- Apollo, work around HAM03 area to remove septum plate.
- Cable tray removal from around BSC02 area, Filiberto.
(corey, greg, jim)
HAM7ISI Springs have been installed, pulled-down, and Spring Safety hardware is in place. Did a first run through on the Optics Table to pull out any broken/loose tangs. The Optics Table can be installed tomorrow. At this point, I was able to note quite a few serial numbers (s/n) while they were easily viewable. Here are some of them:
Springs
(where the Tip is associated with each Corner; i.e. a Spring which is bolted down to a Spring Post of Corner-3 actually has its Tip in the Corner-1 area)
Spring Posts
Flexures
Stage0: 014
Stage1: 010
Optics Table: 009
More Serial Numbers:
Actuators (installed 6/27)
CPS Probes (as slotted in satelite boxes)
The common mode board for the ALS cavity locking, S/N S1102637, has been modified to account for the fact that the green cavity pole is expected to be around 600Hz for the one arm mirrors we have (up from 200 Hz). The common compensation is now a 40Hz/600Hz pole/zero pair. We also modified the first boost filter, so it can be engaged. It is a 100Hz/1kHz pole/zero pair. In detail: C121 -> 220u (from 33n) R69 -> 18.0k (from 120k) C118 -> 1u (from 100n) R77 -> 165 (from 82) We left R70 at 1.2k which reduces the HF gain by 7%.
After enabling the ITMY HEPI + ISI, I scanned the ITMY bias sliders and looked at the ETMY face camera. The ITMY EUL2OSEM matrix has gains of ~5-10 (unlike the TMSY ?) and so the +/-15000 cts of the slider screen is enough to go full scale...and consequently, I tripped the watchdogs a couple times.
The yaw alignment was really good - the return beam to the ETM was nearly centered horizontally, but needed ~1/2 of the ful pitch range of the ITMY TOP stage.
With the illuminator off, it was easy to see the beam flashing around after the watchdog tripped. With the illuminator ON, but turned down to 0.5 Amps, I could see the stuff in the chamber as well as the green beam. Then I used the usual method of putting the beam on the 4 compass points of the ETMY cage and centering it on the mirror by averaging.
Moving the yaw slider to the right moves the beam on the cage to the right. Moving the pitch slider to the right moves the beam down. It takes +1000 slider units of torque to move the beam from top to bottom of ETMY.
The values of ITMY bias that center the beam on ETMY are P = +6300 & Y = +900.
So the ITMY alignment was off by ~6 mirror diameters => (0.210 m / 4000 m) / 2 = 25 micro radians.
Now that the beam goes both ways, the next thing to do is to align the ETMY to center the beam on ITMY using the baffle PDs as before. Then perhaps we'll be able to hear the cavity flashes if we can route the audio from the baffle PDs or the ALS green PDs into the control room.
And...lots of YAW bias gets the ETMY aligned to get the beam back on to the ITMY baffle PDs. The ETMY bias for rough alignment is:
P = -2700
Y = +14200
I haven't seen any flashing yet that looks like a Fabry-Perot resonance, but at least all of the suspensions are close enough in alignment to proceed. Since the TMSY and ETMY both have large yaw biases, we can still hope that they can be mostly relieved by yawing the HEPI (assuming they need the same direction of yaw). Because of the double-loop nature of the quads, we should also get some ITMY pitch relief from pitching its HEPI.
From the flashing that can be seen from the ITMY spool cam, it seems like the return green beam is hitting the ITMY baffle at the edge around the mirror, at the 2 o'clock position look at the ITM from the spool.
While trying to get the video rerouted to the control room as well as EY I will look at hooking up a FiBox to the green light signal. Will talk to Keita about this.
There is something funny about the numbers here. The claim is that the ITM alignment was off by about 25 micro-radians, and that correcting this uses about 1/2 of the pitch control range (I assume this is at the TOP stage). However, according to T1100595-v1, the control range of the quad, from the TOP stage, is 1.14 mrad in pitch; it is a little unclear as to whether this is +/- 1.14 mrad or a 1.14 mrad total swing; either way, it is much larger than 25 urad -- it should take less than 1/10 of the range to correct it.
I think its funny too. Using ~50000 cts (as measured at the DAC outputs for the face OSEMs on TOP) seems like its close to the 217 limit for the DAC.
Should have checked the coil driver readbacks to find out how much current/voltage was being used to make sure there is no attenuation happening downstream.
Jeff K and I found the basic problem with the ITMY alignment numbers. There is a factor of 10 error in the observed pitch misalignment number. The 0.21 m above should be 2.1 m (6 * 0.34 m). The pitch offset is thus 250 micro-radians. We also noted that there seems to be a factor of 2 discrepancy between the designed input range for the TOP driver and the DAC output range, so that we can only access half the current range for each BOSEM: the TOP driver was designed to give +/- 200 ma for each coil, but we can only go to +/- 100 ma. So, the 250 micro-rad of offset is about half the pitch bias range we have available, which is about half of what we should have.
P. Fritschel, J. Kissel The confusion lies in Rana's calculation of the total distance for beam travel: 1000 cts of torque = beam moves top to bottom of ETMY the ETMY is 34cm in diameter. So, a 6300 cts offset, is 6300 [ct]* 34/1000 [cm/ct] = 214 [cm] = 2.14 [m] not ~0.210 [m] as Rana uses in his calculation (i.e. he's off by a factor of 10). So, correcting the number in his calculation, we get pitch offset = torque [ct] / (2 L) % factor of two from "optical lever" effect see G1200698 = (2.14 m / 4000 m) / 2 = 250 micro radians. This jives with the predicted range of the QUAD TOP driver and the current monitors: Predicted Range T0900232 says, in section 6.1: "As the input voltage is specified as 10V* and the voltage gain of the circuit is 1.2** the voltage swing of each output terminal is +/-12[V]. As each output resistor is 40 Ohms, and the [B]OSEM coil resistance is 40 Ohms, the total output resistance is (40 [Ohms] + 40 [Ohms] + 40 [Ohms]) = 120 [Ohms]. The voltage across this resistance will be 24[V] maximim, so the [maxmimum] current will be 24 [V] / 120 [V/A] = 200 [mA] " However, * I've confirmed via direct measurement that the 18-bit DACs can only do +/- 10 [V] differentially NOT per leg. So the max possible voltage is DAC_MaxV = 10 [V] (differential output) ** The DC gain of the circuit is CD_VoltageGain = (1+R3/R10) = 1.22 [Vin/Vout] (OK, 2% error, but just wanted to make sure we understood from where Ron's numbers came) And just in case, to remove all suspicion: From the QUAD Top Driver board schematic, D0902747: - Rout = (R1+R5)*(R90+R91)/(R1+R5+R90+R91) = (22+22)*(220+220) / (22+22+220+220) [Ohms] = 40 [Ohms], so that's correct, and Stuart just measured the BOSEMs to have a resistance of - Rcoil = 42.7 [Ohms] , so that's correct as well. So, the maximum output current of the circuit is actually DAC_MaxV * CD_VoltageGain / (2*Rout+Rcoil) = 10 [V] * 1.22 [V/V] / (40+40+42.7) [V/A] = 0.09943 [A] = 99 [mA] which if all other numbers from T1100595 are correct, then this would produce a maximum displacement of Pitch: 560 micro rad Yaw: 580 micro rad Current Monitors The current monitors have a calibration of Amps across the BOSEM coil = ADC [V/ct] * Fast_I Mon Circuit Gain [V/V] / BOSEM Coil [V/A] = 40/2^16 [V/ct] * 3/2 [(single ended Vout) / (differential Vin)] * / 42.7 [V/A] = 2.1441e-05 [A/ct] The current monitors for the primary Pitch OSEM (F1) on the ITM, with a 6400 [ct] torque offset from the sliders, show -2736 [ct] of current. (I use F1 and not either F2 or F3, because F2 and F3 also contain the Yaw offset). That means the current across the coil is Icoil = 2736 [ct] * 2.1441e-05 [A/ct] = 0.059 [A] = 60 [mA] which is indeed about half the range of the coil driver, which jives with both the [corrected] calculation of 250 urads and the observation that Rana saw ~50000 cts (or half the range of the 2^17 ct DAC max). Note that the original design intent was to have ~1 mrad of range from the TOP stage, so if we want to restore that intent, we should reduce the output resistance of each leg by a factor of 4, such that the range is DAC_Max * CD_VoltageGain / (2*Rout+Rcoil) = 10 [V] * 1.22 [V/V] / (10+10+42.7) [V/A] = 0.19458 [A] = 194 [mA] which could be done by changing the values of R1, R5, R90, and R91 to 10 Ohm resistors.
The attached plot shows the control filter, the existing boost filter, and a proposed filter addition to the HEPI controls.
As we had at LLO, I think we should have a true integrator on the HEPI loops to compensate for long term drifts introduced by hydraulic pressure, thermal expansion, etc. so that the HEPI platforms can be considered to be stable alignment-wise.
Also, why is there so much high frequency gain? It seems like the actuator watchdogs will trip due to acoustic noise or EMI with this kind of design. Perhaps a little low passing can be done without too much harm to the 10 Hz phase margin.
The SR570 (not 560) current preamp that we had been using since yesterday afternoon for channel 27 of the baffling diodes, ran out of batteries today at around 1330 today (so we know that a fully charged one will last just under 24 hours).
At that point it started producing glitches of ~30000 cts in channel 27 with a steady ~4 second period. Because of some grounding issue somewhere, it also produced smaller glitches in the other 3 channels. These glitches looked somewhat like a widly swinging green beam, so I spent a little while investigating the health of the various seismic filtering servos for the ETMY/ITMY before guessing that it was too regular to be mechanical.
Preamp has been switched back to the Melles Griot 13AMP003 "Large Dynamic Range" Amplifiers that the other channels are read out with. The SR570 on batteries was nicer than these; with a little bit of thought, I guess we could figure out how to adapt the ISC QPD readout board into one that can be used for this application (FET preamps, high transimpedance, some low pass fitlering) and then switched out to a lower gain one once we get into real interferometer locking.
After swapping, I used ezcaservo to remove the offsets on all the channels using commands like this one:
ezcaservo -r H2:PEM-CS_CHAN_27_OUT16 -g -0.21 H2:PEM-CS_CHAN_27_OFFSET -t 20
Daniel, Keita, Rana, Aidan, Dale, Dick
Using the ITMY baffle PDs, we were able to get the green beam from EY to ITMY today! The aLIGO beam tubes are able to transmit light.
We spent hours today trying out various ideas about how the beam might be missing, but in the end we found it by moving the EY Transmon Suspension.
The MEDM screen for the TMS had a range of +/- 15000 cts. In fact, the TMS has an 18-bit DAC and so its full range is +/- 2^17 = +/- 131072 cts.
With the current ALS table alignment, the TMS biases which give the maximum voltage on each baffle PD channel are:
chan name |
P |
Y |
CS_CHAN_27 |
-10000 |
+2900 |
CS_CHAN_28 |
-69000 |
+1200 |
CS_CHAN_29 |
-48000 |
+1000 |
CS_CHAN_30 |
-46700 |
+17800 |
We used these numbers to "triangulate". Then we used the ITMY spool cam to further center the beam on the ITM. It is not at all visible on the ITM surface, but can barely be seen on the gate valve behind the ITM as well as a little glint on the edge of the hole in the baffle where the cavity axis is. Our random search pattern initially just gave flashes of ~50 cts, so the new digital filters helped after all. Eventually, the signal on each PD can separately be made to saturate at 32768 cts as we align the beam onto it.
The final TMS settings for "good alignment" to the ITM are P = -16000 and Y = +9250. Correction: there were some digital offsets in the 'test' filter bank. After zeroing those out, the new numbers are P = -42800 & Y = +59200.
Vincent has turned on the ETMY HEPI with low frequency boosts (although no integrators yet) so the pointing should be more immune to the HEPI hydraulic pressure fluctuations. Still need to fix up the ITMY HEPI. There's some seismic measurements going on overnight, but we can next align the ITMY to get the beam back to EY and then lock the green laser to the arm.
Angle calibration of TMS OSEM:
When tilting TMS without tilting the ALS beam on the table, the beam leaving the TMS telescope would tilt by almost the same amount (19/20 or 21/20) as the TMS itself. Here I just ignore 5% error and assume that the beam tilts as much as the TMS.
Using the baffle diode (https://dcc.ligo.org/cgi-bin/private/DocDB/ShowDocument?.submit=Number&docid=d0901376) we can calibrate the TMS OSEM. Note that the PD locations are not the same for H1 and H2. We're looking at H2 ITMY (left hole on page 3 of the above document).
There are three PDs horizontally aligned. The top one (PD3) is about 6.25 in away from the middle (PD2). The bottom one (PD4) is 11in lower than the middle. We know (from experience) that increasing P offset moves the beam down. And there's a fourth diode (PD1) that is at the same height as the middle one and horizontally off to the left by 11 in. PD1-PD4 naming convention is explained in E1100867, page 15.
|
PD3=CHAN_28 (P.Y)=(-69000, +1200) |
|
|
PD1=CHAN_30, (P,Y)=(-46700, +17800) |
PD2=CHAN_29, (P,Y)=(-48000, +1000) |
|
PD4=CHAN_27, (P,Y)=(-10000, +2900) |
For PIT, PD3-PD2 is about 6.25 in for 21000 counts, so it's 6.25*25.4e-3/21000/4000 = 1.9 nrad/count.
Also for PIT, PD2-PD4 is about 11 in for 38000 counts, 11*25.4e-3/38000/4000 = 1.8 nrad/cout.
For YAW, PD1-PD2 is about 11 inches for 16800 counts, so 11*25.4e-3/16800/400 = 4.2 nrad/count.
There's no funny thing in the output matrix, PIT to coil matrix elements are just +-1, same for YAW, so you can multiply 131072 to get the range.
We can also get the initial alignment number using the offsets as of now (-42800 for PIT and +59200 for YAW)
| PIT | YAW | |
| Calibration |
1.9E-9 rad/count Positive offset moves the beam down. |
4.2E-9 rad/count Positive offset moves the beam towards the inside of L. |
| Range | +-0.25E-3 rad | +-0.55E-3 rad |
| Initial alignment |
Off by +81E-6 rad The beam was pointing down. |
Off by -250E-6 rad The beam was pointing outside the L. |
Note that the order of the channel numbers (27, 28, 29, 30) is reversed. Right now PD1, 2, 3 and 4 = Chan30, 29, 28 and 17.
[Stuart A, Betsy B, Deepak K, Andres R, G2] After taking an initial set of M1-M1 transfer functions earlier in the week on PR2 (HSTS), it was observed by Jeff B that somehow a magnet had become detached from the M2 mass (see LHO aLog entry 3157). Thanks to the efforts of the assembly team, this magnet was rapidly re-attached and left to cure for a period of 24 hrs. PR2 was re-suspended yesterday, and a small pitch offset corrected. AOSEMs were then adjusted to their final operating positions and aligned. The canopy/cover was fitted over PR2. Following this re-work, it was prudent to take another M1-M1 transfer function with damping loops OFF for all degrees of freedom, which can be found below (see 2012-06-21_1700_X1SUSPR2_M1_ALL_TFs.pdf). Damping loops were then turned ON and a further complete set of M1-M1 transfer functions taken overnight. All the transfer functions obtained have now been plotted against all other HSTS suspensions previously measured on both LLO and LHO test-stands (see allhstss_2012-06-22_AllHSTS_ALL_ZOOMED_TFs.pdf). n.b. Yellow trace = X1 PR2 M1 (2012−06−21_1700) with damping loops OFF Purple trace = X1 PR2 M1 (2012−06−21_2120) with damping loops ON The transfers functions obtained again demonstrate good agreement with the model and the spread of all HSTS measurements obtained thus far. Power spectra have been taken with damping loops both ON and OFF for each stage (012-06-22_0800_X1SUSPR2_M*_ALL_Spectra.pdf). Power spectra plots, with both damping ON and OFF have been produced, which compare LHO PR2 and LHO MC2 measurements (allhstss_2012-06-22_ALL_Spectra_Don.pdf and allhstss_2012-06-22_ALL_Spectra_Doff.pdf). In addition, power spectra for specific degrees of freedom (L, P and Y) can be more conveniently compared across multiple stages (M1, M2 and M3) of the same suspension in the final plots found below (allhstss_2012-06-22_X1SUSPR2_M1M2M3_Spectra_ALL_Don.pdf). A BURT snapshot has been taken of the current functioning environment "20120622_x1sushxts27_PR2.snap", which has been stored in the following directory:- opt/rtcds3/tst/x1/cds_user_apps/trunk/sus/x1/burtfiles. This BURT snapshot directory has also been tidied to remove old or incomplete snapshots. All of the above data, plots, scripts, and snapshots have been committed to the SUS svn as of this entry. This should now provide sufficient measurements to complete Phase 1b testing of the PR2 suspension and allow it to progress to Phase 2 (chamber-side) testing.
The M3 stage watchdog was observed to be perpetually tripping and could not be reset. This is due to watchdogs being triggered for the OPELV_RMS and OPLEV_SUM, which are not visible/settable in the medm screens. To rectify this I have manually set the following:- caput X1:SUS-HXTS_M3_WD_OPLEV_RMS_MAX 5 caput X1:SUS-HXTS_M3_WD_OPLEV_SUM_MIN -10 For now, this prevents the watchdog from tripping.
These results look excellent. The only thing that concerns me (where the level of concern (from 1 = "it's awesome. no worries" to 10 = "OMG take it apart and rebuild it") is a 5.5) is the cross-coupling I see in the individual comparison with the model, i.e. in 2012-06-21_1700_X1SUSPR2_M1_ALL_TFs.pdf. In there, I see cross-coupling between degrees of freedom which we don't normally expect to be there: 1st Yaw Mode (@ 1.09 Hz) into T and L 2nd Yaw Mode (@ 2.04 Hz) into T and L 2nd Roll/Pitch Mode (@ 1.51 Hz) into Y 2nd Vert or 3rd Long Mode (@ 2.80 Hz) into Y I'm not terribly concerned, because -- as usual -- the cross coupling is reduced to barely visible with damping loops ON. But, it's something to watch out for with this guy as he progresses through the testing (i.e. we'll look with more scrutiny after the optic is swapped). To refresh one's memory -- the reason why we care: (1) When the model doesn't match measurements, we can't trust the model to accurately predict other transfer functions which we can't normally measure (2) If the transfer functions (which are representative of the "plant" upon which we design control loops) are *different* between suspensions by some (as-yet-to-be-quantified) significant amount, then it will make copying and pasting control systems more difficult and time consuming to design. I think with this suspension, and all others, we're doing fine in both these departments -- but again, we won't really know until we start locking some cavities and really try to push the SUS's to their limit.