Servo Busts Into Oscillation Only When Homing

[polaraligned]

.


PLEASE SEE THIRD POST FOR CURRENT ISSUES ON THIS



I replaced the servo on my W axis with a smaller one with a planetary gear box on it. It is a brand new Baldor servo with 1000 line encoder. Old servo worked fine other than it was not strong enough to overcome the stiction of the axis with the available gearing, so I figured that I would add gearing as I don't need speed on that axis. The gearbox dictated that I use a smaller motor for clearance, but I still will get a lot more torque at the shaft it has to drive.

When doing an MPG move one click at a time, it sometimes starts to move in proper direction and then goes backwards. Video will give you a good idea of what is happening. I am turning the MPG in the same direction and not reversing. Servo is disconnected mechanically from machine to show what is happening.



When it goes backwards the DRO will show the proper commanded position, but the encoder counts will actually go in the wrong direction reflecting the actual motor position. When moved again with another click of the MPG, it will move the extra distance to make up for the amount it went backwards on the previous click. So when the shaft moves in the wrong direction, the encoder reads the proper position but the DRO is off by .002".

Worst part is that when I go to home this axis, the motor busts out in violent oscillation as soon as the home switch is hit.

It runs fine when doing a back and forth move in the PID tuning menu.

I ran Autotune several times and every time this servo puts the load meter hard red and faults the machine when it gets to doing the Ka and Accel values.

I spent a lot of time trying to tune this servo, but it's weird behavior continues.

Any thoughts?

[eng199]

It seems like an encoder problem or bad connection on an encoder channel. It does seem odd to get such good results on PID live tuning with an encoder problem.

Move the motor to 0. Put a mark on the motor shaft. Move 10. Move back to 0. Is the mark at the starting location? Repeat for some other move distances and check the mark for position loss.

Your rates are also incorrect. For example, max rate is less than fast jog rate. I wouldn't expect this to cause issues, but it is best to fix the obvious things first.

[polaraligned]

.
I appreciate your response Eng199. I have narrowed down the problem shown in the video above and I will address that at the bottom of this post.



MORE PRESSING PROBLEM: My motor busts out into oscillation when homing. Please see video below.



This was the original battle with this replacement servo. When the planetary gear box is connected and the drive belt installed, I get a position error every time the W axis hits the home switch. When nothing is connected to the motor, it busts out into oscillation ONLY when home switch is tripped during a M92 command. When the motor is running and the switch is pressed, it just sees it as a limit tripped and stops.

What is it during the homing cycle that would cause this? I am lost as this axis worked and homed fine with the old motor.

---------------------------------

Follow up on the issue of the motor jumping backward. As I stated above, when servo jumped back the encoder and the DRO disagreed by .002". Thinking about it, my backlash was set to .002" on that axis. So I set the backlash to zero and the issue disappeared. I checked other axis and when I raised the backlash they did the same thing though a bit harder to detect.

The MPG is causing the issue as it only occurs during MPG moves. It seems that even though I am turning in the same direction, the Centroid pendant is for some reason thinking that sometimes the direction is being changed while I am clicking in the same direction.

Possible debounce issue? I know the Centroid pendant has a microprocessor and communicates serially with the MPU11, so I assume that there is no debounce settings available for this as debouncing is probably done by the microprocessor. Would that be correct?

[cncsnw]

Regarding the oscillation issue when homing:
I see you have already reduced Ki and Kd. I would try (in this order):
Reduce Ka significantly (to maybe 16 or 8 or 0).
Reduce Kp (to maybe 0.5)
Reduce Ki further (to maybe 0.001)

Regarding the MPG issue:
You could watch the MPG encoder counts by setting up a "manual" axis (labeled 'M') and assigning it encoder input #15.
Ignore the warning about a duplicate encoder assignment, it will work fine.
Watch encoder counts on that axis (on the PID screen) while you slowly turn the wheel forward and backward.
My guess is that the detents in the handwheel are too "mushy", and allow the encoder to bounce back a count while remaining in the detent.

[polaraligned]

Thank you very much Marc. Dropping the Kp and Ki did the trick to stop the oscillation. The Ka didn't seem as helpful, and it is set at 8 as of now. When axis hits the home switch now, it does a very short back and forth. I will spend a little more time working on the tuning to hopefully iron that out.

I am going to assume that even though the motor was working perfectly during commanded moves, when the home is tripped the drive reverses the current in the motor as fast as it can therefore causing the oscillations that otherwise did not show up during any other moves.

As for the MPG, I did set up a manual axis for it as you suggested (thank you) and it looks like DRO is following the MPG with no issues, yet the encoder count is off when I get this "jumping back" motion from the motor. The MPG is a high quality Japanese unit that was a bit pricey when it was purchased, so I am not surprised that it is working correctly. I need more time to troubleshoot this problem and get better data on what exactly is happening...Very busy with work and the garage is really hot