Adjusting backlash board, checking oscillations.

Time line

• early aug14: the turret was not always arriving on position. I asked jorge and willie to check the adjustment of  the backlash compensation board for the master/slave motors.
• 14aug14: afternoon" jorge, willie adjusted the backlash compensation board.
• they set the dc offset to 0 for the individual motors, and the backlash board itself
• they set the backlash gain to about -.7 volts.
  
• i then took some data to look at how things were working.
• I noticed an oscillation in motor speeds and torques.
  
• evening: turret was not getting on position. Off by up to .2 degrees
• willie came in and adjusted the balance (dc offset) for the backlash board. Turret now comes to requested position.
• 15aug14:
• morning: i take some data to see how the turret is operating now that the turret arrives at final position.
  
• afternoon: i update the dc offset of the little star dac that drives the backlash board/motors. it was off by about 30 counts.
• Jorge and willie readjust things with the dac board set to have zero volts when no motion is requested.
• Jorge and willie then try various values of compensation gain to get smoothest motion.
• The day before that had about -.7 volts.
  
• They ended up leaving it at +4 volts
• I took data at different backlash settings to see if it changed the oscillations we saw in the motor speeds and torques (didn't seem to).
  

14aug14: first attempt at adjusting the  compensation board.

• The dc offset in the compensation board was set to zero (as it is supposed to be).
• After this adjustment, the turret had trouble arriving at the requested position (stopped up to .2 degrees from requested).
  

• Data is sampled once a second
• The requested max velocity of the turret was:
• 3deg/sec: 14.83 -> 14.91
• 6deg/sec: 14.925 -> 14.965 hours
• Page 1: amplifier speed feedback
  
• Each motor amplifier returns a voltage proportional to the motor speed.
  
• Top: position of the turret versus hour of day.
• 2nd: turret velocity computed from the encoder.
• The green lines show the maximum velocity
• The turret is not arriving at the maximum requested velocity.
• This sets a maximum dac output. 
• 3rd: Amplifier speed feedback:
• black is master, red is slave motor.
• You can see an oscillation in the 3deg/sec. amplitude about 10% of signal and period of about 20 seconds.
  
• bottom:Amplifier speed difference (master - slave)
• This shows the oscillation a little clearer (since the master and slave are out of phase by 180 degrees).
• The 6deg/sec oscillation is not visible (because we are sampling too slowly).
• Page 2: amplifier current feedback (torque)
• Each amplifier returns a voltage proportional to the current (torque)
• We only  get the absolute value of the torque (no directionality)
• Top: turret position
  
• 2nd: amplifier current feedback
• You can see the oscillation. slave looks to be 90 degrees from the master.
• bottom: master - slave torque
• Page 3: converting amplifier speed counts to deg/second
• a fit was made AmpSpd[ampCnts]= a0 + a1*encVel to convert amplifier speed to deg/sec (from the encoder)
• both amps gave about the same fit :
  
• ampVel=-3.89 + 59.3996 * encVelDeg/sec
• ampVel=-5.1356 + 59.1853 * encVelDeg/sec
• so 60 amplifier counts is 1deg/sec

15aug14: speed,torques after willy adjusts the backlash board balance.

    Willy came in 14aug14 evening because the turret was not getting to the requested position (about .2 deg away). By changing the backlash board dc offset, he got the turret to arrive at it's position. Data was taken on 15aug14 am to see if this adjustment
the plots show the turret moving between 40 and 320 degrees (.ps) (.pdf)

• Page 1: amplifier speed feedback
  
• Page 2 :   amplifier current/torque feedback
  
• Results are similar to what was seen before willies adjustment
• Page 3: blowup of amplifier speed and torque
• i blew up the 3deg/sec and 6deg/sec sections for the speed and torque
• You can see the oscillation in the speed and the torque
• The oscillation period for the 6deg/sec is probably faster than our ..5Hz (limit of our 1 Hz sampling)
• Page 4: fit ampSpd to encoder velocity.
• Got close to the same values from previous day,.

Summary:

• When the turret does not get to position , it is probably a problem with a dc offset.
• This turned out to be a dac dc offset.
• What probably happened was that the dacs had a dc offset, and willy compensated by making the dc  offset in the backlash board opposite to the dac offset.
  
•  later in the day we adjusted the dac and set the backlash board back to 0 volts offset.
• The speed and torque is oscillating
• at 3deg/sec the period is 21 secs
• at 6 deg/sec it is faster than 2 seconds (our sampling rate)
• the torque oscillation got min to max
  

home_~phil