new tertiary    2011...
The tertiary reflector was initially installed with fixed supports. A drive system is being developed with 5 motors to move the tertiary. Below is included some of the testing results for this control system.

15jan01 Tracking error on sine wave, calibrate torque output, and measure friction.
16jan01 Ramp the dac output -10 to 10 counts (static,moving friction).
30jan01 Calibrate sliderule potentiometer versus encoder.
06jul01 Fiber,cable interconnections.
09jul01 Tertiary simulator output.
02aug01 Sine wave with motors installed on the tertiary.
02/08aug01 rfi from tertiary electronics and motors.
mar02 Focusing with the tertiary.

15jan01 Tracking error on sine wave, calibrate torque output, and measure friction. (top)

The sine wave: 80000*sin(2pi*.055) was tracked for 4 cycles on the vertical actuators under load. The pi values were: loop bandwidth:1hz with damping=.707, and feed forward gain=1. The maxVel:1024 counts, accLim:2048cnts/.5secs, kiThr=500 encCnts. The sine was 80000 encCnts or 5 inches with a freq of .055hz. The data was recorded every 5 milliseconds. The_Figures:
  1. Figure 1 plots the tracking error in encCnts versus time. The a scaled version of the sine wave is over plotted for reference. 10 encCnts= .625 mil. so the tracking isn't  to shabby.
  2. Figure 2 is the force measured via the monitor port for both vertical drives. A scaled version of the sine wave is over plotted for reference. For the calibration the dead weight used was 500 lbs with a 3/2 factor (3 loops, 2 drives) so each motor saw 750 lbs. The zero offset for the monitor ports was measured to be 2060,2040 cnts for the left/right drives (these values drifted by +/- 20 counts over time). The maximum force went down to 1700 counts. The calibration technique was:

  3.      ma = forceUp - (w*factor + friction)
       -ma=forceDwn - (w*factor - friction)
    So (forceUp+forceDwn)/2=w*factor and (forceUp-forceD)/2=friction-ma and ma was assumed 0. The conversion factors were 3.0604 lbs/count and 3.5673 lbs/count for verLeft,verRight. The computed friction was 356 lbs. This includes any friction in the test frame pulleys. There is a large torque step to reverse direction. The small curvature at each end is the ma acceleration term of the mass do to the sine wave.
    processing: ter/010115/doit.pro

16jan01 Ramp the dac output -10 to 10 counts. (top)

Ramp dacs (open loop) from -10 to +10 and then back down to -10 counts. Sit 5 seconds at each dac value. The velocity range was -.023 in/sec to .023 in/sec with 540 lbs in the test rack (840/motor). Vertical drives only.  The figures:
  1. figure 1:  plot motor torque versus time. Over plot dac values and velocity. Shows that dac values 0-3 give no voltage change at amp. Constant velocity at low speeds takes increasing torque!  Can see the transition region from not moving friction to moving friction.
  2. Figure 2: plots motor torque versus time.  Move along the top at min torque till 0"/sec then drop down as we changed direction. Stay at max torque for .01"/sec ->.02"/sec -> 0"/sec. Then move up to min torque as the velocity reverses. Shows the hysteresis in the motion.
  3. processing: ter/010116/doit.pro

30jan01 Calibrate sliderule potentiometer versus encoder. (top)

An incremental encoder is used for positioning on the tertiary with 2000 counts/turn (20000 cnts/inch vertical and 16000 cnts/inch horizontal,tilt). A slide rule potentiometer is used on each screw to check this encoder. The potentiometer lengths are: 18" ver, 12" hor, 24" tilt. They are spec'ed to 1% accuracy. To calibrate the pot, each axis was run from minimum to maximum position and back two complete times in the test jig. The vertical motors connected to a 500lb load via 3 pulleys, the horizontal drives shared a 250 lb load, and the tilt motor had 250 lbs.The move time was 120 seconds for 1 direction.
      The data was recorded at 5 millisecond intervals. The 96000 samples where smoothed and decimated to give 22 points (the first and last point were included in this). A cubic spline was then done on Encoder(pot)  for these points. The figures show the residuals of encoder - encoder(pot) for the motions. For each axis the residuals are plotted as encoder counts and in inches.
  1. Figure 1 shows the residual for the vertical drives. The top two traces are moving up while the bottom two traces are moving down.
  2. Figure 2 is the residuals for the horizontal drives. There is little hysterisis in the horizontal drives.
  3. Figure 3 is the residuals for the tilt drive. Note that the vertical scale is 2.5 times larger than the previous 2 plots. This drive is the only one where the error is not within 10 mil.
  4. processing: ter/010130/calpot.pro

06jul01 Fiber,cable interconnections. (top)

  • Terminology.
  • Simulator output. (top)

    A simulator has been written in idl to model the different parameter setting for the tertiary: (PI loop parameters, maxAcceleration time, maxVelocity, etc..). The idl code is in idl/ter/sim. The file setup.datdef holds the default setup info. The parameters are: The figures show the results of the simulator. The horizontal and tilt drives have the same maximum velocity (3.5 in/sec) while the vertical drives can go at a maximum velocity of 2 inches/sec. There are separate plots for vertical and tilt/hor.
  • Fig 1. 5 inch move at maximum velocity.
  • For the upper plot is the vertical drive while the lower plot is the hor/tilt drive. The PIbw,damping,kiThr, maxDACnts, kf were fixed. The accTm was varied from .1 to 1 seconds (we use .5 seconds for the real tertiary). The vertical axis is the position error but an offset for each accTm has been added for plotting purposes. The horizontal value on the right of each line is posErr=0.  The acceleration limit causes the overshoots. By making the accTm smaller (accel limit larger) you can reduce the overshoot. You can also see that the vertical drives take longer to do the move since they have a smaller max velocity.
  • Fig 2. 1 inch move at max velocity and then start a sine wave.
  • Fig 3. 5 inch tertiary move varying kiThr value.
  • Fig 3,4. Plot the maximum freq. and amplitude of a sine wave by accTm.
  • Figure 3 is for the vertical drive while figure 4 is the horizontal tilt. If the position is Asin(wt) then the velocity is w*A*cos(wt) and the acceleration is -w^2*A*sin(wt).  The maximum velocity is 2 in/sec and 3.5 in/sec for the ver, hor/tilt motors. The acceleration limit is maxVel/acctm. The top plot on each page shows the maximum frequency by maximum requested velocity for different accTm's. The rightmost velocity (2 for the ver) is for 2048 dacCnts. If we limit the system to 1024 dacCnts then we use the 3rd point from the left.
    The lower plot shows the maximum sine wave amplitude versus requested velocity for various accTm's. These amplitudes are at the maximum allowable frequency.
    processing: ter/sim/dosim.pro

    02aug01 Sine wave with motors installed on the tertiary. (top)

    The motors were mounted on the tertiary and then the sine wave: 10000*sin(2pi*.05t) was tracked for 2.5 cycles on all actuators simultaneously. The central encoder position was 501000. The PI loop parameters   were: loop bandwidth of 1hz, damping=.707, and feed forward gain=1. The maxVel was1024 Dac counts, the accLim was 2048 Dac counts/.5secs, and the kiThr was 500 encCnts. The sine wave amplitude of 10000 counts is .39 inches in the vertical and .5 inches for the horizontal and the tilt.The data was recorded every 5 milliseconds. The_Figures:
    1. Figure 1 plots the tracking error (requested - measured) position in encCnts versus time. Top to bottom the drives are left vertical, right vertical, left horizontal, right horizontal, and tilt. The top plot has a scaled version of the sine wave overplotted. The +/- 50 count scale is about +/-  .002 inches.  The left horizontal and tilt drives are oscillating.
    2. Figure 2 is a blowup in time of the left horizontal and tilt drives. The oscillation has an amplitude of 30  counts (.0015 inches) and a period of .25 seconds.
    The 1.5 mil oscillation is small. It is interesting that it is 15 times bigger than the other drives. The other thing to note is that the oscillation in the two drives is in phase. If it was two motors  going bad, you wouldn't expect them to do it the same way.
    processing: ter/010802/doit.pro

    02,08aug01 rfi from tertiary electronics and motors. (top)

        The tertiary drive system has two electronics boxes.  Box1 is a huffman box containing the motor power supplies and amplifiers. Box2 is an rfi shielded rack containing the digital electronics and computer. There are cables connecting these two boxes and the motors. The lband narrow system was used to measure the rfi from the system. 1337 to 1437 Mhz was covered in 4 sub bands. The resolution was 25Mhz with 1024 channels. 1 second integrations were done.
  •     02aug01 plot of digital box on,off. Data was taken for 60 seconds with the motors off, digital box2 on. Then 60 seconds with everything off. The plot shows the strength of the signal as a fraction of Tsys. It is averaged over the 60 seconds. There is an 8 Mhz comb (marked with the green *).  It gets to about 8% of tsys. This is about 20db higher than the noise for a 60 second 25Khz integration. The black line is pol A and the the red line is pol B.
  • 08aug01 motors on, motors off, all off cycle (image 2.9 mb).  Pol A was used for the image. It  has 60 seconds motors on, 60 seconds motors off, then 60 seconds  all off. This cycle was repeated twice (the cycle starts from the bottom of each plot). Horizontal lines were drawn at the transitions. The image was normalized by the average "all off " bandpass. It was also flattened in the time direction by averaging a set of frequency channels and then dividing this time series into all frequency channels (this is why the motor on birdies have some negative going regions).The motors have a large spike every 25 Mhz (the dsp chip in the amplifier runs at 25 Mhz). You can also see a comb at 2 Mhz (the wavey line are some new cameras mounted on the service platform.)
  • 08aug01 motors on,motors off average plot. This plot averages the motors on (black line), and motors off (red line) (pol A). They were normalized to "all off" so the units are Tsys (28 K for lbn). The birdies at 1366-1368 is a radar. The birdie at 1420 is the galaxy.

    The motors on cover most of the band with the 2 mhz comb (at about 2% tsys). The 25 Mhz comb is about 10% of Tsys. The 8 Mhz comb with the motors off ranges from 2 to 8% tsys.
    processing: x101/010802/rfiplot.pro,x101/010808/doit.pro

    <- page up