DX Cell Re-verification

Off-sky Verification

Hard point setup (DONE)

  • Stall motor at both ends of travel.

For this test, we reduce the drive current to 0.5 Amps rather than the nominal 1.0 Amps. Record the absolute encoder position at each end of travel and calculate software offset (Off = (Up + Down)/2).

We tested two hard points in the cell; the others had been tested downtown (but not by stalling the motor).
All mechanical travel limits and center points are recorded in VxWorks.

  • Put offsets into VxWorks
  • Get load cell serial numbers and put correct scale factors in VxWorks
  • Reload VxWorks
  • Get load cell zero points

Each load cell zero point is determined by decoupling the load cell from the mirror interface. With the hardpoint floating, the load cell value is read and applied to the software as an offset.

All hard point load cell zero points measured. The values ("preloads") are recorded in VxWorks.

Actuator checks (DONE)

  • Bump test actuators

The actuator bump test is used to verify that the actuators are working and that they are properly coupled to the load spreaders.

  • Fix/replace bad ones

The bump test was successful and no bad actuators were found.

Hard point checks (DONE)

  • Check over pressure protection in the air box by lowering the relief pressure to below 120 psi and raising the supply pressure to above 120 psi. The relief valve should open to regulate the pressure.

We adjusted the over-pressure protection in the air box and it seems to work fine. To start with, the over-pressure is set to 60 psi while the regulator is set to 30 psi.

  • Test over-torque switch (air off)

Twisting the counterweight should trigger the over-torque switch. This should cause the red light on the HP control box to illuminate and the PMC control software to switch to the panic state.

We found that HP#2 does not have a torque-switch. We have bypassed it for today and we expect a replacement component tomorrow.

The over-torque switches do indeed cause the drives to shutdown and the control software to switch to the panic state.

  • With the air off, verify that the hardpoints operate without tripping the over-torque switch.

With the air turned off, the hardpoints rely on the torque detentes to prevent the over-torque switch from triggering.

We found that since HP#2 and HP#5 have asymmetric counterweights, the torque imbalance is excessive. We will need to attach more weight to one of the arms near the lower flexure to avoid tripping the over-torque switch.

Update: We corrected the torsional imbalance on HP#2 and HP#5 by adding a calibrated weight to one of the counterweight arms. The hardpoints now work without triggering the over-torque switch.

  • Turn HP air on low (30 psi) and run through lockup.

We had to replace the power supply since the +15V was not stable.

We also had found that loop monitor was not satisfied and would thus not allow us to turn the air on to the hardpoints. After spending a few hours debugging this problem, we found a loose terminal in the rack in the lower right treehouse. The terminal was probably disturbed when the power supply was changed earlier.

The low-pressure test forces looks fine.

  • Increase air pressure to nominal (100 psi) while monitoring hardpoint force.
  • Listen for air leaks.

We have found HP#5 has a serious leak. The problem seems to be serious enough that the air pressure is low at HP#4 and HP#5. The problem will need to be corrected, but the we should be able to proceed with the mirror raise despite the leak.

  • Verify that the maximum and minimum force is appropriate at each hardpoint and that the force vs displacement curve looks normal.

According to 725s005a, the valid force range is -2000 to 2930 N (-450 to 658 lbs). The hardpoint operating pressure should adjusted to keep the HP forces within the tolerable window.

Since the software limits are set to be symmetric (+/- 450 lbs) we are currently limiting the supply pressure to 80 psi.

We have also found that the hardpoint air pressure transducer is not scaled correctly (or at least it does not agree with the gauge). We plan to add a precision mechanical gauge to find out which one is wrong.

  • Resolve any serious issues encountered.

We found a problem with the load cell cable for HP#3. The connection appears as an open when the cable is moved. This issue must be resolved before we can perform the remaining tests. Update: The problem appeared to be a bad crimp on at least one of the wires. The responsible pins were replaced but we may need to recheck some of the other new connectors in the cell.

Update: It turns out that the wire size is incorrect for the loadcell cables. The cables will need to be replaced before the cell is released for service but we can proceed with initial tests.

First raise (DONE)

At least three people must be in the cell while the mirror is raised for the first time. Critical clearances should be monitored at each hardpoint while the mirror is being raised.

  • Manual mirror raise
  • Chase once
  • Zero the relative encoders
  • Turn on the force servos
  • Move up
  • Find (estimate) the nominal middle of travel (magic) by bumping the mirror into the static support.
  • Put absolute encoder values in VxWorks as magic
  • With the mirror at magic, look for any potential clearance issues in the cell
  • Manual mirror lower
The force chase is too slow when the hard point are a long way from lockup. This needs to be improved.

The platform position displayed on the GUI does not take into account the relative encoder positions. This should be changed to reduce confusion.

Second raise (DONE)

(This may be done before or after "Hardpoint performance".)
  • Auto mirror raise
  • Auto mirror lower
  • Repeat 2 additional times at different elevations
  • Drive around with PSF commands to verify cell software response

Hardpoint performance

  • Verify that the X, Y and Z directions are correct using dial indicators or the LVDT.

The directions appear to be correct.

  • Verify encoder scale and motion repeatability by making 10 μm moves in 'Z' while monitoring the mirror position with respect to the vacuum flange using a precision LVDT.

The scale is correct to the precision of the LVDT (1 μm).

  • Record hardpoint position at 37Hz while moving the mirror 10 μm. Show that the mirror moves in a coordinated way and that there is no significant overshoot and little of no hunting. Re-tune position control loop as necessary. Repeat using 10 arcsecond rotations in Rx and Ry.

Sample8: 10 microns in the +Z direction

The backlash is already taken up in this move, so the residual errors are very small (5-6 mas peak).

sample8.png

Sample9: 10 microns in the -Z direction

The backlash is evident in this move since a small error develops at the beginning which results in 15 mas peak error in rotation.

sample9.png

Sample10: 10 arcsecond rotation about the +X direction

sample10.png

Though the above plots look nearly perfect, a peak error of about 73 mas develops for a fraction of a second at the beginning of the move. This error is again due to the backlash and the time necessary for the servo to take it out.

sample10-zoom.png

Sample11: 10 arcsecond rotation about the -X direction

sample11.png

Sample12: 10 arcsecond rotation about the -Y direction

sample12.png

Sample13: 10 arcsecond rotation about the +Y direction

sample13.png

  • Apply a low-frequency sinusoidal force bias and record HP force vs. incremental encoder displacement (measure slope for each HP).

Sample21: A force of about 160 lbs pk-pk was used for the following data.

Hardpoint Breakaway Stiffness Friction (Hysteresis)
0 131 N/μ 42 N
1 233 N/μ 37 N
2 267 N/μ 44 N
3 182 N/μ 19 N
4 164 N/μ 45 N
5 237 N/μ 37 N

Hardpoints 1, 2 and 5 appear to be working well. Hardpoints 0, 3 and 4 are showing low stiffness. This may be resolved by the redesigned torsional breakaway.

  • Apply a low-frequency sinusoidal force bias and monitor mirror motion at mirror surface using precision LVDT with respect to vacuum flange. Compute nominal system stiffness. This measurement should be done at three locations (over HP 0/1, 2/3, 4/5).

Finish up

  • Resolve any critical issues like clearance
  • Finish up wire management
  • Reprogram HP control module loaders to resolve in-system-programming problem and verify that we can reprogram the PIC in-system.

While the counter weights on HP2 and HP5 and the detent systems on all hard points are being worked on, the electronics boxes have been labeled and removed so the PICs can be reburned and retested downtown.

On-sky Verification

Misc

I took advantage of the DX mirror still being up in calm observing conditions (20100213 2UT, calm wind and no mirror ventilation) to compare the hardpoint loadcell readings. After several minutes of eyeball collection of SX and DX readings they appear to be the same. Both sides fluctuate +- 2# p-v or about 0.5# rms.

Collimation & Pointing (DONE) 17-Feb-2010 UT

  • Prerequisite: Magic Position has been defined off-sky (see above).
  • Preset to a star, and (May be in parallel with LUCIFER, if so skip the preset. Do NOT connect LBC to TCS.)
  • Adjust the global offsets in PSFR to obtain a roughly collimated image. 20-min
  • Run LBCFPIA, /REDONLY to refine the initial collimation. 20-min
  • Preset to a star at EL=85, and adjust co-pointing with LBC-Blue. 20-min
  • Take images at 5 elevations to verify that collimation and pointing are OK. 1-hour
  • Preserve the new global offsets by updating the collimation model file DXPMLBCCollimation.dat
  • Contingency: Create a new collimation model if needed. 1-hour

On 20100217 UT JMH completed the first phase of the on-sky tests with LBC-Red and the DX Cell. A short summary of the results:

a) Both LBC-Red and the DX cell seem to be working OK.

b) I have updated the DX collimation table for LBC-Red. (Maybe it is now better than the one for LBC-Blue.)
The collimations were updated on 20100222 to reflect a new magic position.
Net collimation corrections: dX=-3.042 dY=-0.500 dZ=-0.157 dRX=0 dRY=+32.68

c) I made the first tests that small repointings with the cell are repeatable. This was limited because I didn't have the Red Rotator running. The repeatability was already known to be OK from daytime tests.

d) You are able to use Red for LBC focussing tests tomorrow night (Feb 18 UT). Just remember that it has not been extensively tested, so there could be problems that we have not discovered.

e) I have not co-pointed Red and Blue, so that adjustment is needed if you want to observe with both at high elevation.

f) More work is scheduled in the DX cell for Thursday during the day. This probably makes it work even better, but has a small chance of making DX inoperable on Thursday night.

g) If you wish, Andrea (or even John) can change the hardwired flag so that Blue is the Guiding Master and Red is the Slave. The DX cell should now be free of the jitters that were making some Blue images elongated. As discussed in a previous email, the danger to switching the flag is that you may not always have a U-band guide star to guide the mount.

As of 12-Mar-2010, Andrea has improved the LBC software so that the flag in the OB file controls which side is Master and which is Slave.

The detailed observing log can be found at PartnerObserving.20100217UT .

Position Stiffness

  • Prerequisite: Stiffness and air pressure issues in the hardpoints must have been resolved.
  • Preset to a star at EL=70-80 in TRACK (LBC-not guiding) mode 5-min (This test is improved if AGW3/LUCI is guiding the mount. This can be fully parallel if you can coordinate your bias forces and LBC images with LUCI dithering.)
  • Take an Image with LBCR. 5-min
  • Adjust the Z-bias on the hardpoints to 300 pounds. 5-min
  • Take another image and compare whether the star images moved (modulo open-loop tracking drift). 5-min
  • Clear the Z-bias. 5-min
  • Repeat procedure for Z (100), X (100), Y (300), MX (10000), MY (10000) biases. 2-hours
  • Repeat procedure for a star at EL=30. 2-hours

Guiding and Offsetting

  • Preset to a star in GUIDE (LBC-Red is guiding as MASTER) mode. 5-min
  • Play a dithered OB with 30 arcsec dithers to verify the magnitude and direction of offsets. 30-min (This verifies primary collimation performance with telescope offsets.)
  • Preset to a star in GUIDE (LBC-Red is guiding as SLAVE) mode. 5-min
  • Use exposure sequences long enough (>2 min) to verify guiding. 2-hours
  • Preset to a star in TRACK mode (LBC-Red not guiding) (This test is improved if AGW3/LUCI or LBC-Blue is guiding the mount.) 5-min
  • Use PSFGUI to issue pointing offsets to DX M1. Verify the magnitude and direction of offsets. 40-min

Vincenzo et al. spent two hours at the end of 20100218 UT guiding with LBC-Red as SLAVE (Blue is guiding mount, Red is guiding DX primary.) In 214 guide corrections, the rms X-correction was 0.115 arcsec, and the rms Y-correction was 0.136 arcsec. The X-corrections all fell within the range +-0.3 arcsec p-v. Alone, this seems completely consistent with the normal atmospheric motions. If we look at the X-corrections that Blue was sending to the mount, they tend to be about 70% as large. One of the Red Y-corrections was as large as 0.6 arcsec, but this happened at a time where Blue seemed not to have a good guide star.

Marco reports the following image analysis results. The Red images (r') have ellipticities between 3% and 5%, and there are no bad images in the group of 10 6-minute exposures. The Red images were about 0.9-1 arcsec FWHM. The Blue images (g') have ellipticities between 7% and 9 %. (I'm still trying to understand how the Blue ellipticities can be larger, while the Blue guiding disturbances are small. Does this mean that there is inherent ellipticity in the Blue camera?)
###   Seeing data for LBC, UT=20100218
###   Run time: Thu Feb 18 14:52:35 MST 2010
###   UT(hr)  FWswl   IMAGE CAM  FILT   Texp   AIRM   FWHM  Ellip  Np
      10.487   0.79  102912  B    gS  180.23  1.250   4.11  0.078  251
      10.547   0.81  103250  B    gS  180.23  1.250   4.18  0.069  267
      10.608   0.86  103627  B    gS  180.23  1.250   4.46  0.071  237
      10.668   0.87  104004  B    gS  180.23  1.250   4.51  0.074  237
      10.728   0.82  104341  B    gS  180.23  1.250   4.22  0.085  252
      10.788   0.85  104718  B    gS  180.23  1.250   4.40  0.082  245
      10.849   0.88  105055  B    gS  180.23  1.250   4.58  0.084  245
      10.969   0.92  105808  B    gS  180.23  1.250   4.79  0.093  237
      10.477   0.90  102836  R    rS  180.24  1.250   4.43  0.043  186
      10.540   0.88  103224  R    rS  180.23  1.250   4.32  0.039  191
      10.606   0.87  103622  R    rS  180.24  1.250   4.25  0.039  178
      10.665   0.86  103955  R    rS  180.23  1.250   4.22  0.046  178
      10.728   0.80  104339  R    rS  180.24  1.250   3.92  0.044  195
      10.787   0.83  104713  R    rS  180.23  1.250   4.07  0.047  171
      10.848   0.66  105052  R    rS  180.24  1.817   4.07  0.046  175
      10.907   0.67  105425  R    rS  180.23  1.817   4.12  0.047  172
      10.968   0.69  105806  R    rS  180.24  1.817   4.26  0.050  149
      11.028   0.73  110141  R    rS  180.23  1.817   4.50  0.047  151

Issues

1. +15V Supply (resolved)

The +15V supply was not producing a stable voltage. The supply was replaced, but it is not clear that it had an issue. (More detail is required here)

*1b. On 15-Feb-2010, the +15 voltage was reading +14.48V, so needs adjustment.

2. Air Monitor (resolved)

The air monitor card indicated that sector 11 (power distribution) was open. The problem turned out to be a terminal in the bottom of the rack in the lower right treehouse. The terminal would not clamp down on the wire and thus the wire was loose.

3. Torsional Imbalance (HP#2 and HP#5) (resolved)

Two sets of hardpoint counterweights were design with a dog-leg for cell clearance. This caused a torsional imbalance that resulted in the over-torque switch being tripped. A weight was added to resolved the issue.

4. Detent support structure is too flexible

5. Over-torque Switch Needs Shims (resolved)

Shims were installed and the over-torque switch is again working correctly.

6. Air leak HP#5 (resolved)

The shaft inside the breakaway was unscrewed. This apparently happened while installing the filter in the top of the unit.

7. Incomplete set of hardware for detent support structure

8. Unreliable loadcell cable connectors (resolved)

The connectors were not compatible with the gauge of wire. All loadcell cables were rebuilt and installed.

9. Load cell connector cable clearance (resolved)

The new loadcell cables were made without strain relief to increase cell clearance. Therefore, special care must be taken to externally strain relief the cables when they are dressed.

10. Install ground straps

11. Dress all hardpoint wires

12. Remove old motor control electronics

13. Remove unused cables

14. Dress air lines

15. Fix HP air pressure transducer (Resolved)

There was nothing wrong with the transducer. The pressure was low due to the air leak in HP#5.

16. Fix differential pressure transducer

17. Motor cable for HP#1 has no back shell

18. Several vent hoses are contacting counterweights on HP#1 (maybe more as well)

19. Several cables are tied to the counterweights

20. HP boots must be installed

21. Reinstall ventilation, some tubes may require modification

22. HP#2 has insufficient counterweight clearance (resolved)

A couple of the weights and part of the cell structure were modified to improve clearance.

23. Cut flexure cage on HP#2 to improve counterweight clearance

24. Plumbing for actuator #333 is contacting HP#2 counterweights

25. Increase clearance for upper counterweight arm on HP#4

26. Hoses and cables for actuator #233 contact counterweights on HP#5

27. The absolute encoder cables on several HPs are contacting the counterweight arms

28. No armor on torque switch cable for HP#3

29. Install inline air filters on all hardpoints

30. Bad load cell connector pin on HP#2 (resolved)

After a failed attempt to repair the load cell in place, we were forced to replace the unit.

31. Hard panics

We had two hard panics while testing Feb 18 and 19. One occurred spontaneous while raising the mirror and the other occurred while the mirror was down when the air was shut off. This appears to be a power glitch associated with the +7V supply. This has become Issue #2602.

-- DaveAshby - 01 Feb 2010
I Attachment Action Size Date Who Comment
sample10-zoom.pngpng sample10-zoom.png manage 10 K 16 Feb 2010 - 00:19 DaveAshby 10 arcsecond +Rx rotation rotation error detail
sample10.pngpng sample10.png manage 9 K 15 Feb 2010 - 22:37 DaveAshby 10 arcsecond +Rx rotation
sample11.pngpng sample11.png manage 9 K 15 Feb 2010 - 22:38 DaveAshby 10 arcsecond -Rx rotation
sample12.pngpng sample12.png manage 9 K 15 Feb 2010 - 22:38 DaveAshby 10 arcsecond -Ry rotation
sample13.pngpng sample13.png manage 9 K 15 Feb 2010 - 22:39 DaveAshby 10 arcsecond +Ry rotation
sample8.pngpng sample8.png manage 9 K 15 Feb 2010 - 22:35 DaveAshby 10 micron +Z move
sample9.pngpng sample9.png manage 9 K 15 Feb 2010 - 22:36 DaveAshby 10 micron -Z move
Topic revision: r25 - 17 Mar 2010, JohnHill
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