The purpose of this trip was to refine the tuning of the telescope altitude and azimuth servo loops and resolve noise problems with the telescope drive encoders. We did not fully accomplish these goals due to limited time on the telescope and to time spent fixing a problem with the windbaffle servo system.
Monday, June 22
The UW group was finishing up their collimation work on the telescope,
so we were limited in what we could do on the telescope. We only
corrected the polarity of the encoder feedback signals according to the
TCC signals and rules.
Tuesday, June 23 & Wednesday, June 24
We worked on solving a problem that had arisen with the wind baffle azimuth drive system. Two of the azimuth drive rollers slipped when the velocity was higher than siderial rate. After 1-1/2 days of work we found and fixed the problem. As explained by Paul Czarapata, the windbaffle azimuth servo system is configured in a master-slave arrangement, with one motor acting as the master and the remaining two acting as slaves. The master drive is set to speed control and the other two to torque control. A signal from the PLC goes to the master (this is the error signal from the LVDT). The master's torque monitor signal goes to the torque command of the second and third drives. The original controller had a "Azimuth Feedforward" signal that actually went to the slave drives on the "speed reference" input. This signal acts as a maximum speed reference that tells the drive "give me xx amount of torque but don't go faster than yyy rpm." When running with the manual control box this signal is jumpered to the Master speed reference signal (both are in parallel). The problem was that the wind baffle PLC azimuth feedforward signal was miswired and shorted. The shorted signal kept the maximum speed reference at zero, so the slave motors were being commanded to provide torque but not to turn. The additional friction from the recently installed windbaffle centering rollers exposed this problem. After we corrected the wiring, the windbaffle tracked the movement of the telescope at high rate without problems.
Late Wednesday afternoon, we began modifying the control of the altitude
and azimuth axes to include the integral term. We found that the MEI controller
has saturation problems associated with the integral term. The controller
does not work in floating point mode and has a maximum count of 32000 for
this term. The controller performs an accumulative sum of the error
to approximate the integral but does not divide by the sampling frequency.
In this case, the number accumulated grows very fast and saturates the
integral count. By changing parameters of the loop such as increasing
the gain, it is possible to fix the problem and make it operative with
an interpolator gain of 25. If we need to move back the interpolator
gain to 100, we need to reduce the controller gain by a factor of 4 and
it will be very difficult to avoid the integral saturation. In this
case, we need to replace the MEI controller with another unit that does
not have this problem. For now, all of the altitude, azimuth, and
rotator encoder interpolators are set to a gain of 25.
Thursday, June 25
During the morning, we checked for the coupling noise between the windbaffle and the telescope that we had observed during our last trip. This time we did not see any coupling signal between these structures. During the afternoon, we adjusted the altitude axis controller to include the integral term.
Friday, June 26
We tried to address the encoder noise problem by improving the grounding between the telescope and windscreen. We fixed the noise problem on the altitude encoder and the second azimuth encoder by grounding both ends of their respective interpolator cable shields, but this is only a temporary fix. We plan to work more on the grounding problem during our next APO trip.
Last modified 08/04/98