SYSTEM DESIGN NOTE

SDN0002.12 - Prototype Mechanism Testing



 
 
Prepared by Date  Approved by Date Rev. Rev Date
J Elias/B. Gregory/D. Joyce 3/29/99 N. Gaughan 3/29/99

 

1. Purpose

The purpose of the prototype mechanism testing is to verify that the general mechanism design concept will meet the specific needs of the instrument (GNIRS). It is possible, even likely, that this will require more than one test and may require modification or reconfiguration of the prototype in order to fully test all capabilities.

The idea is not to experiment with different designs but to fine tune the design while the layout and details are still flexible and before much engineering and machining have been done. (If the mechanism simply fails to work beyond repair it is good to know this early. If we find something that works we will resist the temptation make it better than necessary.)

2. General

The baseline test configuration should comprise:

a - stepper motor and driver. Software should be available that enables tests to be run, but it does not need to be EPICS-compliant or otherwise suitable for direct use in GNIRS. b - drive train. This should be similar in size, intended precision and range, and mode of operation to the drive trains in "real" GNIRS mechanisms, but it does not have to be the prototype for any specific GNIRS mechanism. The prototype should be representative of the "most demanding" case or cases, but it does NOT have to be reusable in any particular GNIRS mechanism. No time to be spent engineering or designing aspects of the mechanism that are not being tested (for example, optics mounts) and packaging in GNIRS should also not be a concern (fitting into the test dewar does matter, of course).

A judicious choice of the "most difficult" case may serve as a test bed for any and all of the mechanisms since the mechanisms are similar in design; comments on this particular issue are welcome. (Brooke proposes the grating turret as the "most difficulat mechanism".)

Note that the design concept calls for driving both strictly linear motions (slit, decker, "flip-in" mirror slides) and rotary motions (everything else). Can a single mechanism test both types of drive? If not, reconfiguration and retest will be needed. Note that adding fiducials at different locations (see below) may help.

With regard to whether a single test configuration could test both modes of operation, one could argue that a prototype of a rotational mechanism would be more comprehensive than a purely translational mechanism.

1. The rotational mechanism would already involve linear motion (of the rack) and would in addition test the rack-pinion coupling.

2. It is easier to test the repeatability of a rotating object.
 

However, since the mechanical support of the rack will almost certainly differ from that for a linearly driven mechanism, the test of a rotational mechanism will not provide a proper evaluation of a linear mechanism. In particular, measuring the effects of gravity on the positioning of a heavy linear slide may require the testing of such an item. The support of the two types of mechanism are different. The rotational mechanism will presumably rely on shafts and bearings, a technology with which we have a fair bit of experience. Larry may have some real good ideas on how to achieve a stable, low-friction mounting for a linear mechanism, but our corporate experience here is very limited.

c - driven object. This should have mass, friction, and imbalance (or lack thereof) and size (overall dimensions) similar to that of typical GNIRS mechanisms.

d - fiducials. These should be duplicated for comparison; it may be desirable to put fiducials at points in the drive train as well as on the final part of the mechanism, for use as diagnostics, or for testing the linear portion of a composite drive.

e - true fiducial. Some fiducial must be provided that of known accuracy, even if it is unsuitable for use in the real instrument. One example might be a rotating mirror that reflects a laser or auto-collimator. We would like to measure to ~0.01 mrad (=2 arcsec), which produces 0.2 mm displacement of a reflected spot at 10 m. distance. Another example might be an encoder mounted on a stiff shaft (possibly a feedthrough?) attached to the mechanism. The encoder must have a resolution of ~2 arcsec, corresponding to about 1 micron at a radius of 100 mm. It might also be possible to use two plates (one fixed and one moving) as a capacitor.

 Note that the HP encoders work cold but have limited resolution; they may still be useful.

Can a dial gauge be degreased and made to work cold?

Brooke comments: "I always find an unfocussed laser marginal in these tests. I suggest that we provide a camera lens to project the beam through the mirror onto a screen."

f - variable gravity. Ideally, one would like to vary the gravity vector while the mechanism is inside the dewar, and then to test repeatability. This really applies only to internal fiducials, since it will be impossible to preserve alignment relative to external things when the dewar or mechanism is rotated. The nature of this provision depends on exactly what tests we propose to do (see below).

g - thermal strapping and sensors. The design of the prototype should allow it to cool in a reasonable time, and there should be sensors (where possible) to measure it temperature.

 3. Tests a - Repeatability. We need to determine absolute repeatability; that is, repeatability relative to the "true fiducial". Depending on its nature, it may only be possible to measure repeatability of a single position. Next, we need to measure repeatability of the regular mechanism fiducials. Are they are repeatable as the mechanism itself? If not, how well do they repeat? Note that the ability to go between two fiducials provides some further checks.

We may need to check performance at more than one location in the drive train, for several reasons: - only the linear portion may be used for some mechanisms - it may help identify sources of error. It is clearly desirable to measure repeatability at more than one orientation.

b - Flexure. It is desirable to test the mechanism in more than one orientation. It is also desirable to change the orientation in such a way that step count is preserved. That is, if we move to step 10,000 from the "home" fiducial, rotate the mechanism, and go back, how many steps are required to "find" the fiducial? This particular test is conceptually simple but it may be impossible to carry out directly as described. If so, some substitute needs to be contrived that will answer the question.

4. Lifetime

If the mechanism performance is satisfactory, it would be desirable to run it through a large number of motions in order to check for the following:

a - wear

b - changes in performance

 A crude test of the torque required to move the mechanism can be made by reducing the current in the motor until it stalls. This characteristic can be monitored from time to time during the test, to see if it changes. One could also bring the shaft out through a ferrofluidic seal and measure it with the torkducer. (Leave it running a loop over a weekend with a chart recorder...)

Would the torkducer work cold? The design looks like it might...

c - long term repeatability. Note that the requirements for long-term repeatability are not the same as short-term repeatability. (In fact there is nothing explicit.) But it is still useful to know how good (or how bad) it is, especially since poor long-term effects are probably tied to wear or other problems. "A large number" in this context is ~10,000 although some mechanisms can be expected to approach or exceed 100,000 motions during their lifetime.
 
 
If you have any questions or suggestions regarding this website, please contact Melissa Bowersock.

 
NOAO Intranet Services
NOAO Copyright
 Statement

National Optical Astronomy Observatories, 950 North Cherry Avenue, P.O. Box 26732, Tucson, Arizona 85726,
Phone: (520) 318-8000, Fax: (520) 318-8360