MODS1 AGw Commissioning at LDG: 26 September 2010 UTC

Observers: J. Hill (LBTO), R. Pogge (OSU), R. Stoll (OSU), M. Pedani (LBTO)

Telescope Operator: DHuerta (LBTO)

Software: Support: TLeibold (Tucson)

Telescope Support: JUrban (LBTO)

Instrument Support: JMorris (LBTO)

Summary

The offset problem has been solved.

Active Optics with IDL on-axis is working fine. We won't get the corresponding GCS changes in this run. We measured the rotator center on the on-axis guider, and the rotation angle of the slit relative to the sky to good precision (~0.2 pixels for rotation, and 2 parts in 1000 for the rotation angle).

A problem was encountered with the mask insert/retract mechanism that is being analyzed and we will consult with OSU engineering personnel tomorrow about possible remedies.

The last part of the night was lost to high humidity and high winds as the summit of Mt. Graham became wrapped in blowing clouds.

New Pointing Model derived by Marco from yesterday data (new IDL procedure) but not implemented today. The file is ready to be loaded.

Preparations

GCS

Torsten made a patch to GCS to fix the offset problem from last night. See Issue 2881.

New Pointing Model

The NEW Pointing Model for MODS is as follows:
*** Pointing Model Terms ***
IA      +399.66
IE        -36.95
NPAE    +5.26
CA        11.58
AW       -0.93
AN        21.04
TX        11.54
HACA2  -1.95

With these parameters (Note that I dropped TF and A1A) we get a Standard Deviation of 3.74" across the Sky.
We'll test it tomorrow Sept27 UT
IMPORTANT: for some reason, the value of IA is Positive now!!! It could be something related to the RA,Dec format used in the Tpoint Input file. Anyway, I put IA=-399.66 in the MODSSX.20100926.ptmod file to test it

Active Optics Rotation

John (with a bit of IDL help from Fernando) hacked recon_process_image.pro to rotate the subimage by -1.75 degrees before analysis.
; HACK #3 FOR MODS
; JMH WANTS TO ROTATE THE SUBIMAGE HERE
; http://143.117.54.5/idl/Rotating_Images.html#wp1032754
;
; rotate the subimage
  rot_image = ROT(*wfsc.sub_image, -1.75, MISSING=0)
; copy the rotated image back into the 
  (*wfsc.sub_image) = rot_image
;
  print, 'WARNING: The subimage has been rotated by 1.75 degrees for MODS1!
;
; END OF HACK #3

See the attached JPEG below which shows the analysis results with the rotated subimage. Compare to the similar images on the previous night's log. I've adopted a rotation of -1.67 degrees to start the night, although Rick's measurement of the lab wfs images give a smaller -0.78 degrees. We changed to Rick's value later after the wavefront was flattened.

Active Optics pupil centering

I've subtracted 26.6 pixels (2 subaps) from the y lenslet_edge parameter to try to improve the centering for pupil shifts in different y directions. In agw_set_values.pro:

    agw[4].lenslet_edge[0,0] = 354.3 ; x
    agw[4].lenslet_edge[1,0] = 221.4 ; y  JMH 20100926  (was 248.0)

That didn't seem to help when analyzing wfsc000010 from last night, so we reverse the sign:

    agw[4].lenslet_edge[0,0] = 354.3 ; x
    agw[4].lenslet_edge[1,0] = 274.6 ; y  JMH 20100926  (was 248.0)

But I had my X's and Y's screwed up, so we try:
    agw[4].lenslet_edge[0,0] = 327.7 ; x JMH 20100926  (was 354.3)
    agw[4].lenslet_edge[1,0] = 248.0 ; y  

Now we are getting somewhere, but went the wrong way (right):

    agw[4].lenslet_edge[0,0] = 380.9 ; x JMH 20100926  (was 354.3)
    agw[4].lenslet_edge[1,0] = 248.0 ; y  

No, that moved too far the other way (left), so I must need a fractional subap correction in this same direction:

    agw[4].lenslet_edge[0,0] = 360.9 ; x JMH 20100926  (was 354.3)
    agw[4].lenslet_edge[1,0] = 248.0 ; y  

This looks excellent on wfsc000010, and on wfsc000055, so I think I've conquered step b) from last night. However, wfsc000058 looks to be too far right by one subap. Added these coefficients to OSU_L.cfg:

WFS_lenslet_edge_x 360.9  double # defines the edge of the virtual lenslet grid in x
WFS_lenslet_edge_y 248.0  double # defines the edge of the virtual lenslet grid in y

Comparing with last night's applied Zernikes:

These remeasurements from last night use the above parameters:

imagename Total measured Z5 Manual Applied Z5
wfsc000058 -243 +1000
wfsc000061 -1260 +3000
wfsc000064 -96 -4000
wfsc000067 -297 +1000

MODS

R & R installed the 1.2 arcsec wide long slit mask, and some other software fixes from the night before that fixed a few minor nuisances in the target/guide-star selection parts of the MODS control panel GUI. We are now operating MODS from one of the observing workstations instead of the MODS laptop, and routinely archiving data.

Details

Open

02:04 Open T=11.9 D=2.9 Wind 8 m/s from NE, sky is mostly clear

02:08 Active preset (not really intended to be active)

02:11 Acquire Preset - adjust IE=-17 CA=+17

02:14 Guide Preset WT10_302F

02:16 collimate_once

02:25 Adjust the lenslet edge, since obscuration is in column 8.

    agw[4].lenslet_edge[0,0] = 347.6 ; x JMH 20100926  (was 354.3)
    agw[4].lenslet_edge[1,0] = 248.0 ; y  

We hit the M1 collimation limits twice here. I'm suspicious that we're going the wrong direction on the sign of Zernikes.

Offset Tests to Test GCS

The good news: GCS now adds in the hotspot offset, but we don't recover the star on the guider. The telescope is dropping to none mode. 3rd instance at 02:35.

02:36 David restarts M2.

Torsten reports that the telemetry database is not working. This explains why we didn't get any temperature plots from last night.

02:38 Setting none mode on left side again.

02:44 The offset command drops telescope into none mode for even a relative offset of 0,0.

02:47 Revert back to previous GCS.

Here the offset fails as you expect, but the telescope doesn't drop into none mode. The preset is preserved intact, and only the offset and guiding fails.

Back to active optics

agw[4].rot_offset = 30.0

But IDL doesn't seem to have recognized the change. ????

02:58 collimate_multiple wfsc000007-09

The first corrections fixed Z7, but then we started diverging in Z8. By eye collimation gives 0.9 arcsec image with -2000 Z7 and -500 Z6. IDL clearly converged into the wrong value of Z7. John is working on figuring out Active Optics while the others measure rotator center. Michele and Torsten are on the phone trying to figure out the offset problem. It seems that GCS sent a guide correction of 0.3, 13 mm which was large enough to put the telescope off-source.

agw[4].lenslet_edge[0,0] = 374.2 ; x JMH 20100926 (was 354.3) agw[4].lenslet_edge[1,0] = 248.0 ; y

We are getting garbage from the M2 offload calculation even though we aren't using M2. Maybe this is because of long intervals without active updates.

03:42 Restarted IDL, and now the 30 angle has been recognized and the signs of Z7, Z8 really have flipped as expected. Probably we are over-rotating the spots and

03:44 preset guide on-axis

collimate_multiple, 3, /send wfsc000010-12

03:51 collimate_multiple, 3 wfsc000012-14 - reference

Center of Rotation Measurement

03:15 - Acquired WT10-302F as a test star with a guided preset. John cleaned up the image by hand, and then we moved the star back to the nominal pointing reference (guide probe back to field center position). Took a 5s exposure to verify we had the star, using the ND1.0 filter to avoid saturation. Took a 70s exposure rotating MODS from 0 to 360 (raw rotator values through the MCS). Repeated after the first image was dirty and John cleaned up again.

Reference image is /OldRepository/20100926_MODS/guider000007.fits taken at 03:37:28UTC.

* Rotation center image for MODS1 (guider000007.fits):
RotationCenter.png

Center of rotation was measured by two methods: eye-ball a circle over the track in ds9 and measure marginal sums in X and Y measuring the centroids with spectral line fitting programs (IRAF splot and LINER). Both methods and independent measurements by MP, RS, and RP gave consistent results:

X_CoR = 315 Y_CoR = 220

This is our adopted center of rotation for the GCS on MODS guider. Precison of the quantitative measures using marginal-sum profile fits is about 0.3 pixels (33 mas), but we only need to know this to 1 pixel or so (110mas).

More active optics.

collimate_multiple 13-15 Z7 wants -1000 Z8 wants +1100 send Z7 -1000 image gets worse

03:56 collimate_multiple 16-18 Z7 wants -800 Z8 wants +1700 (we seem off by 90 deg) agw[4].rot_offset = 120.0 ; JMH 20100926

03:59 collimate_multiple, 3, /send 19-20 - Z7 improves but Z8 gets worse (Aaargh - transposed) Clear Active Optics - Why is there large residual coma now when there wasn't at the start? reverse the sign on Z8 in agw[4].zernike_factor -1.4 changes to +1.4

04:06 collimate_multiple, 3, /send 21-23,24-26

04:12 New GCS build with offsetting really fixed. The answer is that Torsten has to do double bookkeeping of where the probe really is, and where he tells Michele (PCS) that the probe is. Michele called to say that the rotator also went off-source during Torsten's erroneous 21 arcsec guide offsets, but we don't understand why. Change my spot rotation to -0.78 now that we have clean images.

04:24 collimate_multiple, 3, /send 27-29 - Z5 diverges

04:26 Clear Active Optics Change Z5 factor from +1.7 to -1.7

04:29 collimate_multiple, 3, /send 30-32,33-35 Send +500 Z9

04:36 collimate_multiple 36-39 (the corrections are correctly -310 Z9)

04:39 collimate_multiple, 3, /send, 40-42

Adopted MODS parameters in agw_set_values.pro that work on-axis

    agw[4].focal_station = "MODS"
    agw[4].side = "left"
    agw[4].station = "directGregorian"
    agw[4].instrument = "MODS"
    agw[4].rot_offset = 120.0    ; JMH 20100926
    agw[4].pupil_x_on_axis = 361.0
    agw[4].pupil_y_on_axis = 254.5
    agw[4].x_probe_flip = 1.0
    agw[4].pupil_direction = -1.0
    agw[4].lenslet_edge[0,0] = 374.2 ; x JMH 20100926  (was 354.3)
    agw[4].lenslet_edge[1,0] = 248.0 ; y  
    agw[4].zernike_factor = [0.0, 0.0, 0.0, 0.0, -1.5, -1.7, -1.6, 1.4, 1.4, -1.7, 1.7, -1.3, $
                             0.0, 0.0, 0.0, 0.0, 0.0, 0.0,  0.0, 0.0, 0.0, 0.0, -0.2]
Present Active Correction are: X=-0.931 Y=+1.924 Z=+0.381 RX=20.67 RY=10.00 at EL=57.2 I will add these to the global offsets. These will include +1 X and +1 Y parallel translations.

New global offsets for M1

M1 Global X=-0.514 Y=+0.291 Z=+0.381 RX=-18.570 RY=+14.870 applied in SXPMMODSCollimation.dat matching with: M2 Global = -6.010 -0.317 0.000 148.650 181.800 0.000

Measuring Rotator angle

Torsten's offset fix meant we could proceed to measure the instrument rotation zero point. We did this as follows:

An image of the 0.6-arcsec long-slit mask is used to establish our vertical reference. The goal is to determine the LEFTZEROPOINT parameter for the PCS that will align this slit with the celestial PA=0-degrees. Our target field was the Stone-O astrometric field, centered on the target is Stone-O star 1477 with the offset guide star is Stone-O 1136. This fainter Stone star allows us to use 30-second integrations to average out short-term seeing variations, and gives us a slightly brigter off-axis star that will give good guiding and is placed to allow a good range of N-S offsets that won't run out of guide probe travel. John Hill adjusted the collimation using the IDL program and we achieved nice, round 1-arcsec diameter images. We then executed the "slitscan" script for MODS that took 6 integrations of the star after the following absolute offsets sent to the IIF:

dRA (") dDec (")
0 0
0 +30
0 +15
0 -15
0 -30
0 0

At each location a 30-second integration was used in the SDSS r filter with the MODS red camera. The closed-loop image motion compensation system (IMCS) of MODS was locked on solidly for the entire sequence. The last repeat offset back to our nominal (0,0) location was to verify (to first order) the repeatability of the offset after Torsten's fixes. We found that the two (0,0) pointings differ only by (dx,dy)=(0.2,0.4) pixels, corresponding to (24,48) milliarcsec for our estimated pixel scale of 0.122 arcsec/pixel. We note that because we were not actively correcting collimation as per John's notes in this report, the second image was out of round (1.17x1.02 arcsec compared to 1.02x1.02 arcsec). This is good agreement.

We then measured the (x,y) centroids of the star using the XVista markstar commands. A line fit to the X displacement in pixels as a function of Y position gives a slope of 0.00687+/-0.00014, corresponding to an angle of 0.394+/-0.008 degrees of rotation of the N-S offset scan relative to the Red CCD camera. Using the slit image, we took cuts along the Y-axis positions of the 5 star offsets and measured the X-centroid of the slit cuts. A line fit to these gives the slit position angle of 0.237+/-0.002 degrees. This gives a net rotational offset of 0.157 degrees. If we have the sign right, this suggests that LEFTZEROPOINT should be 30.157 from our initial guess of 30.0.

05:30 Change LEFTZEROPOINT from 30.0 to 30.157, stop and restart PCS. Verification of our rotation change was aborted due to a mask extraction fault and increasing wind/humidity after correcting that fault.

Mask Extraction Failure

05:35 Stop for a MODS mask problem. On trying to extract the long-slit mask to repeat the instrument rotation zero-point measurement and verify our rotation, we got a very odd fault that indicated that the mask insert/extract "grabber" was in the mask-stowed position, but that there was not mask in the cassette. A quick image of the sky showed the mask there, but slightly off center and out of focus. A few diagnostic commands showed that the grabber was in the stow position, but the two sensors that indicate either a mask in the cassette stow position or the science-field position were both showing a mask in neither location. We then drove the telescope to Zenith and set the stow pins on the telescope and rotated the instrument around to the mask load position for a visual inspection.

What we found was that the mask was sitting below the focal plane, resting on the science field baffle plate that runs the length of the slit mask insert/retract run. We removed the tail cover and installed the collimator mirror covers, and then closed the telescope enclosure doors and brought up the lights. John Hill and Rick fashioned a tool to extract the mask, and with careful placement of the storage cassette and a whole lot of care, we successfully retrieved the mask. Subsequent tests and inspection showed a problem with the mask grabber unit. We will work without masks until we can consult with OSU engineers back home, and put the masks in a safe configuration that will not lead to a repeat of the problem.

Update: 2010 Sept 26: consulted with Tom O'Brien at OSU who inspected the MODS2 slit mask mechanism currently in our lab. In summary the problem we see is plausibly related to the distortion of the MODS1 structure by the rotator that is causing other optical and mechanical alignment problems in the instrument. A plan to correct this is being formulated, but will not be implemented during the remaining time in this observing run. For the time being, we have reloaded the masks so that there is an empty cassette position above and below each mask, which avoids the problem we encountered.

05:38 Install the new M1 and M2 collimation tables.

05:45 Close with telescope at zenith so we can turn on lights and extract the lost mask.

Remain Closed for weather

Clouds, Humidity and Wind have arrived. -- JohnHill - 25 Sep 2010
  • wfsc000055 analyzed with subimage rotated -1.75 degrees:
    rotn000055.jpg
Topic attachments
I Attachment Action Size Date Who Comment
RotationCenter.pngpng RotationCenter.png manage 176 K 26 Sep 2010 - 04:03 RichardPogge Rotation center image for MODS1 (guider000007.fits)
rotn000055.jpgjpg rotn000055.jpg manage 234 K 25 Sep 2010 - 22:06 JohnHill wfsc000055 analyzed with subimage rotated -1.75 degrees
Topic revision: r11 - 26 Sep 2010, RichardPogge
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback