TR-98020
SDSS Trip Report 
Telescope Servo System Commissioning
May 12-16, 1998

Claudio Rivetta and Charlie Briegel
Fermi National Accelerator Laboratory
Batavia, IL   USA    60510

 
The purpose of this trip to APO was to 1) resolve problems encountered with the telescope servo system encoders; 2) improve the tuning on the telescope altitude axis servo loop; and 3) refine the performance of the rotator and azimuth axis servo loops.
 

Wednesday, May 13

On Wednesday, we started working on the rotator servo system. Two problems had been identified with the rotator servo: 1) missed counts on the encoder at velocities greater than 10x sidereal rate; and 2) slippage between the drive motor capstan and the rotator disk. The first problem was the result of switch settings inside the encoder interpolator. At the given switch settings, the gain of the interpolator was set at 100, the maximum input frequency at 22KHz, and the output pulse edge separation at 80 ns. This last parameter is below the minimum pulse separation of 200 ns specified for the MEI controller and appears to be the reason why the MEI controller was missing counts. Table 1 lists the possible interpolator configurations based on internal hardware switch settings.

We changed the interpolator switch settings to obtain a gain of 25, a maximum input frequency of 22 KHz, and an output edge separation of 380 ns. With these interpolator settings, we tested position repeatability of the rotator by watching encoder output as we moved the rotator back and forth between two indicator marks we placed on the axis. We had good agreement between the actual position and that indicated on the CRT display.

To convert encoder counts to engineering units at the drive disk, we used the nominal diameter values for the rotator drive disk, drive capstan, and encoder capstan. We realize this may introduce some error in the position readings if the actual diameters vary from these nominal values. Therefore, we need to verify the actual diameters and then recalibrate the system.



Table 1. Heidenhain EXE 600 Interpolator Switch Settings
 
Gain 
Max. Input Frequency 
Output Clock Frequency 
Minimum Output Edge Separation 
Switch 
(1) 
(kHz) 
(MHz) 
(ns) 
S5 
S6 
S7 
S8 
100 
22 
10 
80 
o
x
x
x
11 
5 
180 
x
o
x
x
6 
2.5 
380 
x
x
x
x
 50 
43 
10 
80 
o
o
o
o
22 
5 
180 
o
x
o
o
11 
2.5 
380 
x
o
o
o
6 
1.25 
780 
x
x
o
o
25 
43 
5 
180 
 o o o
x
22 
2.5 
380 
o
x
o
x
11 
1.25 
780 
x 
 o o
x
6 
0.625 
1580 
x
x
 o
x
  

The second problem with the rotator axis involved slippage between the drive capstan and drive disk. We tried to address this problem by re-adjusting the preload on the rotator drive capstan. We worried that the preload had not been properly readjusted after the last set of friction measurements was made on the rotator harmonic drive coupling. However, after verifying that the motor was properly preloaded, we saw that the capstan still slipped on the drive. Further inspection revealed that the slippage was caused by the rotator cable wrap. At certain positions, the rotator wrap tends to hang up on the rotator, which causes the rotator to stall. There is little we can do about this until the rotator cable wrap is reworked or a new cable wrap installed. Jon Davis is working on an interim fix until a permanent solution is found. We did verify that no damage was occurring to the drive disk as a result of the slippage.
 
 

Thursday, May 14

On Thursday, we worked on the altitude servo system. We set the interpolators for both encoders to a gain of 25, maximum input frequency of 22 KHz, and output edge separation of 380 ns. We also installed low-pass filters, with a bandwidth of approximately 200 Hz, in each power amplifier. This allowed us to increase the gain of the altitude PID loop, which resulted in a much stiffer axis. It should be noted that the integral term in the PID loop is currently set to zero. Additional work is necessary to fine-tune the integral term.

With the new interpolator settings and the low-pass filters, we tested the ability of the control system to hold altitude position by pulling on the end of the secondary truss with a spring scale. We exerted a maximum downward force of 30 pounds and saw that the telescope held its altitude position. This test indicates that the control loop will hold the altitude axis in position against reasonably high forces.

While we worked on the telescope controls, Paul Czarapata worked on the wind-baffle altitude axis servo system. He modified the LVDT connections to the wind-baffle servo system to eliminate a ground loop, and connected the new altitude motor/brake assembly to the drive amplifier. By the end of the day, work on the altitude axis was nearly finished.
 
 

Friday, May 15.

During the morning, Paul finished working on the altitude wind-baffle control system and connected the chain between the motor and the wind-baffle structure. The altitude wind-baffle servo was ready for testing. We once again tested the telescope altitude servo system by moving the telescope within the wind-baffle. We saw some high frequency oscillations in the altitude servo, which we removed by reducing the gain in the PID loop. (Is this accurate?) We also saw some low frequency perturbations that we were not able to remove.

The weather at APO was clear with wind gusts on the order of 15-20 mph, so we decided to open the building to check the behavior of the telescope with wind perturbations. We moved the telescope in altitude and azimuth at several different velocities, ranging from sidereal rate to approximately 1/10 of the maximum slew rate. Throughout these tests, the wind-baffle tracked the telescope without problems.  The performance of the telescope servo system was almost the same as what we saw inside the building. The telescope held its position well, but the low frequency perturbations we saw inside the building were still present.  We also moved the telescope 20 degrees in altitude and saw that the encoder position agreed with the altitude clinometer to within 1 degree.

At about 5 PM, we closed the telescope enclosure and tried to understand why we had such oscillations in the control system. With the wind-baffle servo system turned off, we re-tuned the PD loop and the low pass filters in the amplifiers for both the altitude and azimuth axes. With the wind -baffle servo turned on, we observed that noise was being coupled to the telescope control system. This appears to be causing the low-frequency perturbations we are seeing in the telescope controls and is a problem we will have to address during our next trip.
 
 

Saturday, May 16.

On Saturday morning, we once again tested the altitude and azimuth axes. We worked some more on the missing count problem to better understand if we can use a gain of 100 with the MEI controller. We determined that at a maximum velocity of 3.5 degrees/sec, the maximum output frequency delivered by the encoders will be about 10 kHz. If the interpolators are set to a maximum input frequency of 11 kHz and gain of 50, the output pulses to the MEI controller will have a separation of 380 ns, which the MEI can read without problem. With an interpolator gain of 100, the minimum separation drops to 180ns, which is slightly below the 200 ns minimum  specified by the controller and may lead to dropped counts by the MEI.

We tested this on the rotator by setting the interpolator gain to 100,  the input frequency to 11 kHz, and the output pulse separation to 180 ns.  Moving the rotator back and forth several time, we saw an average error of a few minutes over 60 degrees of motion.  We are not sure, however, whether the discrepancy was due to missed counts or to the accuracy of the marks that we placed on the rotator.  We started to run out of time at APO, so we will do more testing with the interpolator settings at Fermilab or at the Heidenhain branch office near Chicago to understand what the optimum settings should be for the interpolators.

Prior to leaving APO, we set the encoder interpolators for all three axes to a gain of 25, input frequency of 22 kHz, and minimum output pulse edge separation to 380 ns. With these settings, the MEI controller should not lose counts under normal telescope motions. It is possible that counts will be lost, however, if the telescope is moved by hand at velocities exceeding 6 degrees/sec.

The control system is currently capable of closing on position in all three axes.  Expected position error at this time is on the order of 1000 encoder counts.  With the interpolator gain set to 25, this equates to 14 mas at the altitude and azimuth drive disks.
 

Questions regarding this trip report should be addressed to Claudio Rivetta at (630) 840-8035 or Charlie Briegel at (630) 840-4510.

Last modified 05/27/98
boroski@fnal.gov