SYSTEM DESIGN NOTE

SDN0002.10 - Mechanism Positioning

Prepared by	Date	Approved by	Date	Rev.	Rev Date
Jay Elias	2/4/99	N. Gaughan	2/5/99	a	3/4/99

There are three approaches to defining the position of a mechanism. These are:

Positive definition. That is, use of a detent or its equivalent to physically locate the mechanism at its correct position. A hard stop for a flip mirror is a similar type of definition.

Where such positioning is used, the motor must still have some level of positioning. It must get close enough to the position for the detent to activate, or there must be some means of sensing activation (such as stalling against a hard stop).

Position sensing. In this case, the motor is moved until sensors indicate that a specific position has been reached. These could be Hall-effect sensors, micro-switches, or something else. In all cases, however, the sensor itself is what determines position.

Dead reckoning. Here the mechanism is positioned solely by counting steps on the motor. There may be algorithms for backlash removal to help reduce errors, but in all cases the positioning depends on never losing a motor step.

Note that there must be some means of identifying at least one specific position on a mechanism (“home”) from which dead-reckoning is done thereafter. Ideally, this would be done by one of the two methods above, although it could also be done by viewing alignment with the detector.

The purpose of this document is to examine the requirements for use of the second and third approaches in the GNIRS mechanisms. An examination of the mechanism requirements indicates that for most of the mechanisms, a rotational positional accuracy in the range 0.03-0.07 mrad is required. This translates into a requirement of a few microns at the edge of a mechanism of ~100 mm radius. For the purposes of this discussion, we require that the error associated with either of the  two approaches be 2 microns on a 100 mm radius, since there are other sources of positional error (bearing play, for example).

For an approach using dead-reckoning, one must have 1 full step of the motor significantly less than the allowable error. The most common motor type has 1.8°/step (31.4 mrad/setp) If we set 1 full step to 1 micron at the edge (=0.01 mrad), then the reduction from the motor must be roughly 3000:1 or greater. It's also clear that the errors in the repeatability are dominated by the last interface in the gear train, namely between the gear on the mechanism wheel (e.g. at the edge) and the gear driving it directly.

Since the reduction factor is at least 3000:1, the motor must operate at >3000 rpm (for "fast" mechanisms like the slit wheel) or >1000 rpm (for "slow" mechanisms like the camera turret). This is a maximum step rate of ~10,000 Hz (full steps).

Since the reproducibility depends on the gear train, it is important to note that the gearbox must return to its starting configuration after a full rotation if the mechanism is capable of turning >360 degrees. That is, any gear in the train must make an integer number of rotations for a full rotation of the mechanism. This is not a requirement if the mechanism does not make multiple rotations.

For an approach using position sensing, the requirements on the motors are not much different, since the mechanism must still be positioned in integer motor steps. It is not, however, necessary, to never lose a step -- only to do so infrequently enough that the fine positioning of a mechanism can be done.

In this case, it is necessary to sense the position of the edge of the mechanism to ~2 microns. It is not clear that this can be achieved directly with either Hall-effect sensors or microswitches. One might deal with this by mounting a second sensor in the gear train at a location where the reduction factor is 10:1 or more. One would then do coarse location using the sensor on the main mechanism and fine positioning using the sensor on the drive gear. One caution is that the positioning error must include not only the error in sensing the position of the drive gear, but any residual effects in the drive train between the gear with the sensor on it and the mechanism itself. Since this is the dominant error for dead reckoning, as described above, it is not clear that position sensing offers particular advantages. It may be a very good way to define a "home" position from which dead reckoning is done. (A mechanism that uses this approach is the focus drive in NIRI.)

The dual-sensor approach to positioning sensing is fine for home positions, which can be essentially arbitrary, but has some difficulties where multiple, specific locations must be sensed. In particular, since drive gears tend to be small, there will be a limited number of separate sensor positions on the gear. This complicates the task of making sensor positions coincide exactly with desired positions of the mechanism.

Another issue that arises is redundancy. Microswitches are reputed to be failure-prone (there is also some concern regarding the Hall sensors), although there is also a substantial body of experience to the contrary. Although the reputation may be largely hearsay, it does not appear that difficult to duplicate the sensors, especially where only a "home" position is involved. NIRI does this and can switch between the redundant sensors in software. Note that there may be some offset between the home positions defined in this way, which should (ideally) be calibrated beforehand so that a sensor failure is nearly transparent to the user.
 

---

---

If you have any questions or suggestions regarding this website, please contact Melissa Bowersock.

NOAO Intranet Services

NOAO Copyright  Statement