search

misan / dcservo

Position control of DC motors
315 stars 114 forks source link

Backlash correction? #57

Open paukstelis opened 6 years ago

paukstelis commented 6 years ago

I'm using geared motors and there is a reasonable amount of backlash. What is your thought on where best to implement this? It seems that we could fairly easily track direction from the pwm output and just look for cases when direction changes and modify the target position with +/- the number of encoder steps needed. This seems too easy, so I'm sure I'm missing some critical factor here!

misan commented 6 years ago

That sounds about right. It may be useful to have the compensation value stored in the EEPROM so you can change it without recompiling the code. However I am not sure how well that will work (but I would like to know if you follow that route).

On Fri, Nov 2, 2018 at 8:35 PM paukstelis notifications@github.com wrote:

I'm using geared motors and there is a reasonable amount of backlash. What is your thought on where best to implement this? It seems that we could fairly easily track direction from the pwm output and just look for cases when direction changes and modify the target position with +/- the number of encoder steps needed. This seems too easy, so I'm sure I'm missing some critical factor here!

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/misan/dcservo/issues/57, or mute the thread https://github.com/notifications/unsubscribe-auth/AAccyL-RpSUs0gKg8ugOxVrB0H4tXJYUks5urJ5vgaJpZM4YMKsJ .

paukstelis commented 6 years ago

So it seemed simple, but becomes more complex in practice. I was basically doing what I described: tracking direction and applying modification to target when direction changed. The problem became that any overshoot (which I can't totally avoid because of large masses) would trigger a correction, which leads to stuttering. I'll have to think if there might be a better way to do this.

paukstelis commented 6 years ago

Oh, I think the fix is simple. I was monitoring motor direction via the sign of the PID pwm output. I can just watch when DIR changes. That should do the trick.

misan commented 6 years ago

Proper adjustment of the PID gains could give you a response without overshot

paukstelis commented 6 years ago

I've never been able to get it spot on. The larger the motors (and the higher the operating voltage) make it more difficult to tune, from what I have seen.

I tested watching changes from the DIR signal and that seems to work well if you assume isotropic backlash. The only complicating factor is that on one axis I have a somewhat restricted drag chain that soaks up some of that backlash depending on where it is. I'm not sure I want to get in modeling this to compensate.