SDN 0002.16 – Mechanism Position Calibration and Re-calibration
1. Introduction
This note discusses the issues associated with calibration of mechanism positions, and consequences of the proposed approach to mechanism service. The main conclusion is that almost any work on the mechanisms will require re-determination of the position look-up tables.
2. Mechanism Position Calibration
GNIRS mechanisms are positioned using a stepper motor driving a gear train. The gearing is designed so that positioning to the nearest motor step meets requirements of mechanism reproducibility (see SDN002.13 for specifics). In order for this scheme to work, the drive train itself must be highly repeatable – there cannot be wind-up, uncorrected backlash, or other sources of uncertainty. The mechanism designs and prototype testing are intended to ensure that this is the case.
The absolute positioning of the mechanisms is, however, a somewhat different issue. There is no particular requirement on positioning by dead reckoning calculated from the basic design parameters. Instead, upon initial assembly dead reckoning should be good enough to get mechanisms positioned approximately, and precise positions will be determined empirically.
The requirements and procedures are outlined below.
2.1 Initial Calibration
GNIRS contains 9 internal mechanism and one external (warm) mechanism. Of these, the external cover and the acquisition mirror will be positioned using hard stops and a spring-loaded limit switch, and do not require precise calibration of the stepper drive. Of the remainder, most but not all have a certain number of fixed positions to which they would be moved. These are the two filter wheels, the slit and decker slides, and the prism and camera turrets.
The remaining two mechanisms – the grating turret and the focus – have a range (or ranges) of position rather than a set of discrete positions.
For all of these mechanisms, a rough calibration will be possible based on the mechanism design. Refinement of this calibration will be done during final integration of the instrument, using observations with the engineering array to establish the “correct” step count for each discrete position, which would then replace the provisional value in the look-up table for that mechanism.
There are three exceptions to this: the filter wheels, the grating turret, and the focus drive.
For the filter wheels, positioning requirements are fairly loose and there is no easy way to determine “best” positions cold. Instead, the mechanism will be calibrated warm and no further refinement will be done.
For the grating turret, the data (wavelength at array center vs. tilt) will be used to produce an adjusted grating calibration curve. This fit would probably use the tilt zero-point and grating constant as adjustable parameters, but leave all higher-order terms untouched. Note that this approach will only remove low-order errors in the grating drive, and will not deal with small-scale effects (see next section). This means that the wavelength calibration will be approximate. Any precise calibration will have to rely on the actual data (telluric lines or arc spectra from the Gemini calibration unit). This is the only practical approach, as no calibration based on look-up tables will ever have the accuracy required for many scientific programs. The nominal calibration will be good enough to ensure proper spectral coverage and allow basic line identifications: it is good enough for operational but not analytic purposes. Also, the grating constant derived from tilt vs. wavelength could differ slightly from that derived from position on the detector vs. wavelength (by, perhaps, a part in 104).
For the focus drive, operation consists of determining best focus for a given configuration. For this mechanism, it is desirable that the step vs. position relation be fairly linear, since one would like to determine focus by adding different contributions rather than having an enormous look-up table for all possible configurations. Since allowable error is actually a relatively large fraction of total mechanism travel (total travel ~2500 steps), this should be possible.
2.2 Sources of Absolute Error
Errors in absolute position occur whenever a certain number of steps move the mechanism by differing amounts at different points in the mechanism travel. As a example, the nominal relation of tilt to step count for the grating turret is 1000 steps equal 4 degrees, so 90 degrees equal 22,500 steps. But if the worm wheel used to drive the turret is slightly out of round, by one or two thousandths, the number of steps to rotate exactly 90 degrees will be perhaps 10 or 20 different from 22,500. This difference will be highly repeatable, and one would simply adjust the look-up table from the initial
values by the observed difference in step count.
Such errors in the final drive gear are likely to be the largest, and will also tend to be on the largest scale.
Another comparable source of absolute error is the position of the home (datum) switch, which could also be off by a few thousandths of an inch. This affects the zero point for all the positions of the particular mechanism in the same way.
Other sources of error show up as higher order effects – periodic errors in the worm itself, error in gears preceding the final reduction, etc. All of these will also be repeatable once the mechanism is assembled provided there is no way the mechanism can return to the same nominal position in a different configuration. For example, if the filter wheel is not driven by a gear where the reduction ratio is an integer, any run-out in the drive pinion will have a different effect depending on how many times the filter wheel is rotated (though for the filters the effect may not matter).
A similar problem arises if the mechanism is partially disassembled – the position calibration may not longer be good enough.
2.3 Quantification of Errors
2.3.1 Final Drive
The final drive in any of the mechanisms is the portion most sensitive to run-out and equivalent effects. Furthermore, since it is usually the largest part (and often the most extensively machined), run-out will be larger. For this reason, it should be assumed that any disassembly of the mechanisms at this point will require recalibration. (Doing anything with the home switch will obviously have an equivalent effect.)
2.3.2. Motor Drive Removal
At the other extreme, what happens if you remove the first stage of the drive train – basically the motor with no reduction or a modest reduction? If nothing is moved, the design will allow the drive to be attached again with essentially the same clearances, and effects will be small.
But what if someone moves something, so that when things are put together, the gears are not meshing in the same way?
The repeatability requirement is effectively ±1/2 step, or 0.9 degrees of rotation of the motor. For a gear of 1 inch diameter, this amounts to 0.008 in. of error at the edge of the gear. This is somewhat larger than typical run-out or tooth-to-tooth errors for gears of the size and type we will use, but not by an enormous factor. There may be several gears preceding the point of separation, including some modest reduction (which effectively amplifies the errors). Also, this cannot be the dominant contribution to the overall repeatability. These errors can be removed by re-calibration after the mechanism is re-assembled.
2.4 Implications
2.4.1 Prototype Testing
The prototypes should provide some information on the magnitude of the departures from a strictly linear calibration (with the caveat that the encoders may contribute some as well). These data can be used to decide whether the recommendations below are in fact required for all mechanisms (with the mostly likely exceptions being the focus drive and the filter wheels).
2.4.2 Procedures after Mechanism Re-Assembly
Unless the test data establish otherwise, it should be assumed that any time a mechanism is taken apart at any point, it will have to be re-calibrated after the instrument is cooled down, and the appropriate look-up table will have to be updated. For the case where only one mechanism is affected, the procedure would be to set up the simplest configuration that includes that mechanism in the optical path, and then go to each position in sequence, using the nominal calibration. The correct position would then be determined and the new values would be entered in the look-up table. One would then want to check that no errors were made in the process. For the slit slide, which has the largest number of positions and is also fairly slow, the entire process could take 1-2 hours. For the grating turret, one would probably measure a small number of tilts for each grating (3-4?) and then fit the zero-point and grating constant to that.
The effort, while not negligible, is still quite small compared to the work required to open up the instrument, do something to a mechanism, and then put everything back together and prepare the instrument for use.
It’s worth noting that for almost all the mechanisms (the filter wheels are the exception), the only reason for removing a mechanism is to change or adjust some component – in which case some re-calibration would be needed anyhow.
It’s also worth noting that even if the mechanism is removed as a unit, without touching the drive, it may require re-calibration of the overall zero-point (one would certainly check it).
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 NOAO
Intranet Services |
 |
 |
Statement |
National Optical Astronomy Observatories, 950 North Cherry Avenue, P.O.
Box 26732, Tucson, Arizona 85726,
Phone: (520) 318-8000, Fax: (520) 318-8360