random leg movement into hardstop

[GoogleCodeExporter]

The leg moved randomly into hard-stops. Ryans theory is, that this is a 
calibration problem with the optical encoder.
Check the problem, fix it, develop a procedure to mount the sensors right and 
calibrate them.

Original issue reported on code.google.com by tra...@gmail.com on 21 Aug 2013 at 10:26

[GoogleCodeExporter]

Issues with the optical encoder would not cause the stated symptoms. My theory 
is that, because we have no sign convention on the motor, the medulla's halt 
state acted as positive feedback on the motor velocity, not negative feedback.

This theory may be checked easily by looking at log data from one of the 
incidents.

Original comment by jrvanwhy on 23 Aug 2013 at 6:51

• Changed state: Accepted

[GoogleCodeExporter]

Original comment by jrvanwhy on 23 Aug 2013 at 6:52

• Added labels: Priority-Critical
• Removed labels: Priority-Medium

[GoogleCodeExporter]

I have looked at some log data and confirmed that the medullas went into the 
halt state during the incident.

My solution is to create a sign convention for the motors. When looking at the 
face of the motors (the side the shaft comes from):
        1) Sending a positive direction command to the amplifiers should cause the motor to rotate counterclockwise
        2) Rotating the motor counterclockwise should cause the incremental encoders to count upwards

Additionally, for the EtherLabs system (or our current system, if we decide 
against using EtherLabs), I will implement checks that monitor the rotor 
velocity during a halt to confirm that the motor is, in fact, slowing down.

To clarify my previous statement about the optical encoders: The optical 
encoder alignment did not cause the acceleration, although it may have caused 
the robot to enter halt in the first place.

Original comment by jrvanwhy on 27 Aug 2013 at 5:06

• Changed state: Started

[GoogleCodeExporter]

More info, and busted hypothesis:

Looking at the data again, it appears as if the motor begins to accelerate in 
the incorrect direction prior to halt mode's engagement; the acceleration 
corresponds with the controller's current command, but has the wrong direction 
(the controller was limiting itself to 10 amps in the data I have). It only 
begins to accelerate rapidly when halt mode activates.

The direction the motor accelerates in matches the direction it would move if 
the direction pin from the Medulla to the amplifier were to become 
disconnected. Therefore, we now suspect that the direction pin may have become 
disconnected during the tests, leading to the runaway motor condition.

Original comment by jrvanwhy on 29 Aug 2013 at 1:49