ROTSE-III hardware manual

Version 0.2 08-Aug-2003 mcba, added optical collimation section.
Version 0.1 05-Aug-2003 Michael Ashley, assisted by wisdom extracted from e-mails by Alan Schier, Carl Akerlof, Eli Rykoff, Andre Phillips. and Jim Wren.

Checklist of known problems, and what you should do to protect your mount/telescope

  1. A drive motor can overheat due to excessive current over a prolonged period of time (>30 minutes?). The motor will partially demagnetise due to the temperature, and will therefore have reduced performance, making it more likely to overheat in the future. Solution: install a thermostatic cutout switch on the motor body; ensure I2t protection is enabled in the mount software; check the balance of the telescope, and look for cable snags.
  2. The drive o-rings can be shredded, and the capstan then grinds into the drive wheel. Solution: install Andre's protection widgets on both RA and DEC axes; replace o-rings; adjust the preload and servo coefficients appropriately; repair the damaged metal surfaces with gentle application of, e.g., Scotch-Brite (take great care not to damage the encoder tapes). See also this additional information.
  3. Misshapen images out of focus. Solution: consider reducing the tension on the primary mirror retaining screws.
  4. A "thump" is heard as the mount slews. Solution: it is possible that the RA and/or Dec axes are loose, consider adjusting the preloads.
  5. The mount oscillates at low frequencies, or with an audible whine. Solution: the servo coefficients may need adjusting; the o-rings may be worn; the capstan preloads may need adjusting.
  6. The Renishaw encoder shows orange or red on its status LEDs for part or all of the sky, possibly leading to a position error being flagged by the mount computer. Solution: adjust the encoder position; clean the encoder tape; inspect the RA/Dec axes for loseness.

Adjusting the drive capstan preloads

Alan: It is not too difficult to get the preload for the drive correct. One tightens the preload screw until the capstan does not slip when the axis experiences a significant disturbance (like being pushed by hand). And then maybe a half turn more.  The drive surface and capstan should be reasonably clean, but need not be spotless.

Andre comments: when tensioning the RA capstan onto the drive wheel, I find there's a narrow range of tensions between too little (capstan skids) and too much (worrying creaking noises from the o-rings).  The Dec drive seems to be less touchy about the capstan<->drive-wheel tension.

Possible causes for shredded o-rings on the RA and Dec drive capstans

How many o-rings to use?

If you use too many o-rings, you will find that they will ride-up on each other when the capstan preload is increased. Back off by one or two o-rings. On ROTSE-IIIa we have 15 o-rings on the Dec shaft and 14 on RA. See the previous section for o-ring materials and part numbers.

Preventing o-ring damage from causing capstan/drive damage

If the o-rings are sufficiently damaged, the metal capstan can make contact with the drive wheel, and eat its way into it, causing substantial damage. To prevent this, you can install Andre's protection widgets. Here are some images of them on the ROTSE-IIIa telescope (which uses a different RA motor to the other telescopes). An alternative is to use a stop-screw next to the preload spring (Alan is designing one for ROTSE-IIId, and Dmitry has made one for ROTSE-IIIb).

Setting the mount servo coefficients: Kp, Ki, Kd

The ROTSE motors are stabilised by a PID (proportional, integral, differential) feedback loop. The three coefficients Kp, Ki, and Kd, have to be set appropriately in order for this loop to be stable under all conditions. The position encoder is a Renishaw tape mounted on the axis drive wheel. If anything changes in the mechanical part of the loop (e.g., o-ring wear, bearing preload), then this can affect stability.

Here is a prescription for setting initial values for Kp, Ki, and Kd, from Alan Schier:
  1. Note the existing parameters (Kp, Ki, Kd) so that you can restore them if needed.
  2. Put 2 and 0.4 in the f and z box respectively in the "Calculate Gains from Break Frequency and Damping" box and click the calculate button.
  3. Copy the resulting Kd and Ki value to the servo parameters for the axis of concern.
  4. Lower Kp for that axis to maybe half of its current value.
  5. With the servo running (axis in the "Stop" state, probably), raise Kp as described previously.  That is, perturb the axis while raising Kp until the servo oscillates and then lower Kp until the oscillation stops.
This will leave the servo in a reasonably robust state.

If you want to push the envelope and make the servo more "aggressive", start rasing the break frequency value from 2 to 3 or 4, recalculate the values of Kd and Ki. (In this case there is nothing to gain by changing the damping, z.)  Copy the new values to the servo parameters and then readjust Kp again.  What you will find is that the servo becomes "stiffer", and the range of acceptable Kp values will narrow.  That is, the range between where Kp is too low (sub-audible low frequency servo oscillation) and too high (audible oscillation) becomes smaller. Eventually the range disappers altogether, and you will not be able to make the axis stable by adjusting Kp.
 
When you do this, you are moving the roots around in the complex plane of the chacteristic equation that describes the system dynamics.  Some root positions are better than others.  Roots in the right half of the complex plane are unstable.  Roots in the left half are stable and have a continuum of performance characteristics ranging from "twitchy" to "sluggish".
                                                                               
The break frequency and damping describe the position of a complex pair of roots over which we have direct control.  The other roots are controlled less directly since they are determined by the mechanics of the system.  In a sense, Kp is our remaining handle on those roots.
                                                                               
A further note:  beware if you start to adjust Kp, Ki, and Kd by simply observing the perfomance of the axis and changing one or another.  The stable combinations of these three exists is a space in which they are surrounded by combinations that are unstable.  In particular, if you start with all of them near zero, you will have to traverse unstable territory to get to anything that performs well at all.
 
Here are some additional notes from Alan on adjusting Kp in the event of mount oscillation:
  1. Lower the Kp gain by 60%.   Push on the RA axis with your hand and watch the resposne.  The servo should still be stable.  The axis may have a sluggish response and recover with a bit of overshoot.  This is normal.
  2. Start raising the gain with the little up arrow at the right side of the Kp box.  The axis's response to a push will become faster and the overshoot will decrease.
  3. Keep raising the gain and listen/feel for an oscillation to start as you keep perturbing the axis with your hand.
  4. Once you get an oscillation, quit pushing on the axis and start lowering the gain by clicking on teh down arrow until the oscillation stops by itself.  That's the right place to leave the gain.
Some general notes:
Andre notes that a good way of getting a feel for the system is to grab the tail of the motor shaft that protrudes from the rear of the servo motor and twist it abruptly this way and that. Andre noted some slight "springiness"  in the capstan drive shaft, deriving from the helical coils in the shaft coupler.  These are the coils that Alan has partially filled with modeling putty in order to reduce vibration.

The following table shows the coefficient values at the indicated dates (note that ROTSE-IIIa has a different RA motor from the other sites). See also the servo.ini section below.


 Kp (RA/Dec)
 Ki (RA/Dec) Kd (RA/Dec)
ROTSE-IIIa at 2003 Aug 14
 27.273 / 8.743
 15.708 / 10.472
 0.06 / 0.066
ROTSE-IIIb
 Dmitry?
 Dmitry? Dmitry?
ROTSE-IIIc Eli?
 Eli? Eli?
ROTSE-IIId ?
 ?
 ?

The server loop runs with a 2millisec period on a Windows NT computer.

Servo.ini files

The mount computer software reads a file called servo.ini when it starts. This sets various parameters. Here is servo.ini for ROTSE-IIIa at 2003 Aug 14.

Renishaw encoder

The specification for the RGH22-Y30D00 encoder head is for a maximum runout of +/-0.1 mm, which is comfortably greater than the measured runout of the ROTSE-III mounts. See the Renishaw RGH22 data sheet, the installation guide, and the DXF drawing, for more information on the encoder.

The encoder head is sensitive to the yaw angle relative to the scale, and less sensitive to the pitch and roll angles.  If the yaw angle is off, the encoder will not tolerate the full +/-0.1 mm range of runout.  If the light is turning red in places, move the axis just to the edge of the red zone (the light may be orange) and poke the encoder mount with your finger in the yaw direction.  If you can make the light turn green, that means the yaw angle could be improved.  For that matter, you can poke the mount in the other directions as well to see if other adjustments would help.

Andre remarks: most of today I've been fiddling with the RA-axis Renishaw encoder.  I was mostly adjusting yaw and could get things mostly right, but never consistently so.  Then I decided that guesswork was out and removed a lead weight and the lower aluminium 'bumper' bracket so that I could get vernier calipers around the sensor and find out what was really going on.  Thus exposed, it's relatively easy to measure the yaw by measuring for parallelism of the sensor with the overhead drive wheel.  The correct 0.8mm spacing from the graticule was achieved with a paper 'feeler gauge' (by a convenient coincidence, our standard photocopy paper thickness is 0.10mm).  Then I measured the roll (i.e. parallelism with the tangent to the drive wheel) and was surprised to find it off-by-a-country-mile (i.e. 2.3 degrees). This was corrected and... hey presto... solid green status light all the time, no red/orange flickering, no exceptions.
 
The moral of the story is that these encoders are very sensitive and have to be set up precisely to spec, and trying to 'wing it' by simultaneously adjusting three free-parameters is not the optimum way of doing it.  Getting down-and-dirty with measuring gear is the best method, even if it means stripping the telescope down a bit to get the micrometer into position. 

The RA Renishaw encoder sits on an aluminium block heater. There is a heater and thermocouple in that location.  If you look at the servo control panel, you'll see the temperature monitoring/control info toward the lower left.  The servo software monitors the temperature to keep the block above 0 C (or whatever setpoint is set in the control panel). The encoder heads are specified to 0 C.  Below that, the LED output declines. If properly aligned, they will work below 0 C (to -10 C at least).  If the alignment is marginal, the status light will turn red.

Right ascension bearing adjustment

There is a nut on the lower end of the RA axle that bears against a wavespring and the brake wheel.  These items bear against the inner race of the lower bearing and provide tension to keep the fork snug against the inner race of the upper bearing.

The preload nut is the ring at the center of the brake wheel.  First loosen the 1/4-20 set screw on the hub of the brake wheel.  1/2 to 1 turn should be enough. Then tighten the nut until it is snug using a spanner wrench or a couple of punches placed in the holes for the spanner wrench.  The bearings are hugely oversized for the load they carry, and there is little danger of overtightening.
 
When things are snug, you should notice an appreciable increase in the friction when you rotate the RA axis.  If the friction hasn't increased, the upper bearing is probably still loose.  You could monitor the position of the RA drive wheel relative to the base using a dial indicator.  If the axle was loose, you will see the drive wheel settle back toward the base when you tighten the nut. Once this is done, you can loosen it just a bit.  About 30 degrees is fine. This allows the wavespring to expand a bit.  You can then tighten the set screw on the brake wheel, and that should be it.

The bearings are tapered rollers.

What should the maximum runout be on the RA drive wheel?  Alan's recollection of the ROTSE-IIIa mount is that it had no more than +/-0.04 mm (+/-0.0015") of runout on the drive wheel. Alan measured another mount and found it is within +/-0.02mm (+/-0.0075") except for near a machining ding that is outside the encoder area.  Including that ding, it would be about about +/-0.04 mm.

Declination bearing adjustment

The bearings on the drive side of the dec axis are an opposed pair of roller bearings and are preloaded with wave springs on the inside.  The bearing on the other side is a single roller bearing on the outside portion of the axle with a wave spring under the nut (on the outside).

If the bearings are loose, a "thump" might be heard as the axis settles depending on the telescope position.

The way to adjust the bearing is:
  1. Loosen the non-drive side a bit.  Two or three turns on the nut should suffice.
  2. Tighten the drive side nut until you flatten the preloading wave spring. You'll feel the tightening torque increase suddenly when this happens.
  3. Back the drive side nut off about 1/8 turn.
  4. Tighten the non-drive side nut until the wave springs are flattened. This may take considerably more torque that for the drive side since the preload on this side is larger.
  5. Back off the non-drive side nut about 1/8 turn.
That's it.  The bearings are oversized for the task and very robust. There's no need to be particularly ginger when tightening the nuts.

Andre commented: unlike the RA bearing adjustment, I can easily see the Dec bear adjusting nuts and wave springs.  On the motor-drive-side I noted see that the wave spring was looking nearly fully squished, but on the non-drive-side, somewhat less squished.  So I progressively tightened the non-drive-side adjustment, and after at about half-a-turn, the "thump" disappeared.

Motor protection

The servo motors can be driven with quite high voltages (>10V) for a short amount of time (a minute or so) in order to produce large accelerations. However, they can only withstand 2 or 3V continuously before they will overheat (on a timescale of perhaps 30 minutes). The mount control computer will register a fault condition if either a high voltage is applied for more than a user-defined number of seconds, or the integrated power to the motor (measured by integrating the current squared) exceeds a user-defined threshold. To prevent the "High Output I^2" fault, you need to ensure that the mount is carefully balanced and that there is no excess friction from, for example, snagging cables.

Thermostatic protection of the servo motors

At ROTSE-IIIa we have installed thermostats on the RA and Dec motors to provide a last line of defense against motor overheating (which has burned out at least three motors in ROTSE-IIIa/b). The thermostat we used is a normally-closed encapsulated type, with a setpoint of 50C.  This image shows the thermostat topped with a gob of heat conductive paste, and set into a foam-plastic 'blanket'.  This image shows the thermostat strapped onto the side of the motor, and connected in series with it.  These particular thermostats have a hysteresis of around 10C, so if the thermostat is tripped, it will take some time for the motor to become reconnected.  Note that this addition doesn't interfere with the removal of the Dec drive wheel cover.

Primary mirror cell adjustments

The radial restraint is via four rather broad arcs, although as a practical matter, the contact will be localized to a region in the center of the arcs as the arcs are relatively compliant.  The arcs bear on the rear of the mirror.  The arc material and dimensions are chosen such that the system is athermal for uniform temperature--they are not spring loaded.  After the mirror is installed, the arcs are adjusted such that they are in contact with the mirror and are left there.  Should this arrangement ever put the mirror into a bind, it would be apparent as astigmatism in the image.

The following image shows the primary mirror cell, with the mirror itself removed. The four arcs bear on a machined lip on the back of the mirror, and are adjusted radially by set screws (one for each arc).

Image of the primary mirror cell

This image below shows one of the primary mirror retaining screws, which bears down on a plastic L-shaped bracket, which in turn contacts the edge of the mirror. The back of the mirror rests on the small brown plastic strip visible below the bracket. It is very important that the spring is not completely compressed (since this will allow large forces to be placed on the mirror). It is also important that the L-bracket and the brown strip are aligned, and that the mirror is resting on the three brown strips, else a torque will be applied to the mirror.

image of a primary mirror retaining screw

During the commissioning of ROTSE-IIIb the three primary mirror set screws were too tight, turning the out-of-focus doughnuts into triangles.  Adjusting the set screws 1/128 turn past contact held the mirror nicely without distorting the shape.

Metric/imperial?

Are you wondering why there is such a bewildering mix of metric and imperial fasteners on that unit?  It's because the telescope was designed before there was a mount for it.  It turns out that the telescope was designed using metric dimensions.  When we got a mount, it was imperial. So, use metric wrenches if you're working on the telescope; imperial for the mount.

Adjusting the secondary support vanes

It is possible to make the vanes tighter or looser, depending on how the secondary assembly has been installed.  To tighten the vanes:
  1. Loosen the both ends of the two vanes whose alignments do not pass through the center of the secondary assembly.  That is, there is one vane that is strictly radial and two other that are a bit off of radial. So, loosen the both ends of the two non-radial vanes.
  2. For each of those two vanes, slide the end nearest the secondary cylinder in a direction tangential to the cylinder.  There is a bit of clearance in the screw holes that will allow this.  Q: Which direction to slide them (clockwise or counter-clockwise)?  A: In the direction that would tend to pull the vane away from the outer barrel.
  3. Tighten the screws that attach the vanes to the secondary cylinder.
  4. Tighten the screws that attach the vanes to the outer barrel.  This should tension the vanes.

Collimation of the primary mirror and optics tower

The optical design of the ROTSE-III telescope employs a parabolic primary mirror, a flat secondary, and a multi-element optical tower. Aligning the secondary mirror is a trivial operation, since the fact that it is flat means that the only necessary adjustments are two tilt axes, and these can be easily done using out-of-focus star images.

Aligning the optical tower is much more difficult, since it has to be both centred and tilted correctly with respect to the parabolic surface of the primary, which is not necessarily aligned with the edges or back of the mirror. Note that it is not necessary to align the optical axis of the system with the telescope tube, or perpendicularly to the mirror cell, although in practice any misalignment will probably be less than 1 degree in tilt and 1 mm in displacement.

In the following procedure, we assume that the internal elements in the optical tower are correctly aligned.

Begin by taking the primary mirror cell and optical tower out of the telescope. Insert a wooden block to hold the shutter open, as per the following image.

image showing a piece of wood opening the shutter

Mount the mirror cell on a dividing head from a milling machine (see the image below). The dividing head we used had a central hole which enabled us to inject a laser beam via a 45 degree mirror. We used a portable hand-drill to assist with rotating the dividing head. The next task is to align the parabolic surface of the primary mirror as exactly as possible with the rotation axis of the dividing head. This can take a couple of hours of trial and error with the crude setup we used. Sheets of paper were placed under the mirror cell in order to make fine adjustments to the tilt. We bounced a laser beam off the mirror surface and looked at the reflection on the wall of the room, several metres from the mirror. When the beam does not move when the dividing head is rotated, the mirror is perfectly aligned. We were successful in aligning the tilt to < 1 arcminute, and the centering to better than 0.2mm. The optical access of the ROTSE-IIIa mirror is certainly within 0.5mm of the physical centre as defined by the edges of the mirror.

image showing the mirror on a dividing head

Now shine a laser through the optical tower, using a mirror at 45 degrees positioned immediately below the dividing head. It is relatively easy to align the tilt and decentre of the laser precisely to the optical access of the tower by looking at the reflections of the laser light back along the optical path (the laser reflections on the various lenses in the tower are visible in the image below). Now spend another hour or so adjusting the tilt of the tower so that the reflected laser light does not move when the dividing head is rotated. You may need to machine an allan key in order to fit the four adjustment points on the tower. Adjustment is performed by adding/removing plastic shims. We were able to achieve <1 arcminute alignment after some hours of work. However, we noticed that the laser beam passing through the optical tower wobbled by +/-2.5 arcminutes as the tower was rotated. We suspect that this is due to some residual misalignment of the elements within the tower.

image showing laser light passing through the optics towercloseup of laser reflections


Enclosure

The enclosure electrical diagram.