Re: Heck of a time tuning servo motors.

Too many poles in the RHP?

This may be a contributor to my problems. I do not have access to
the I term as theses are gecko 320's. I do have the P and the D. I
really should break out the scope and see what the error term is
looking like when it goes into this state.

I have messed with setting the system up to be severley over damped,
but there were still problems. I think I'm just going to have to sit
down and get real systematic about my tuning and setup approach. so
far I've been in the "its almost working" rush to get it moving. I
think taking a step back and being methodical is called for at this
juncture.

My plan is this :
1. Adjust my gibs and backlash loose enough for very free play in
both the X-Y axis (while eyeballing the 5/8" ballscrew parts on
mcmaster).

2. Completely start over with the PID tuning procedure described in
the manual with an eye for the fact that it may be a bit more touchy
operation than I previously anticipated.

3. Start ordering parts for 2 stages of reduction in my mill spindle.

Thanks for the great info on tuning the servo amps. I have had some
control theory, but I must admit that it was a while back and more
geared towards filter design and what not. Lots of Matlab and not
too much rubber meeting the road...

-Erik Reikes

--- In CAD_CAM_EDM_DRO@y..., "Greg Jackson" <greg@b...> wrote:
> The problem is most likely in the tune. When you get high current,
but
> nothing is going anywhere, I suspect you are getting a high
oscillating
> current and the servo is doing a high frequency hunting.
>
> The problem is difficult to describe without quickly getting into
second
> order differential equations, root locus plots, and other issues
related to
> dynamics. Without getting into long textbook discussions, I might
give you
> a few suggestions:
>
> PID works fine with linear systems. Leadscrews, and their
associated
> backlash, create a very non-linear system. Conventional PID cannot
handle
> backlash without serious detuning.
>
> If you have access to the gains, the first one to lose is the
integrator,
> that's the I term. I would always start out with I as zero.
Proportional
> will act like a spring, with higher numbers being a stiffer spring.
> Derivative will act like an oil filled damper and calm it down a
bit. Using
> only P and D terms is like tuning a car suspension with springs and
shocks.
>
> Also, friction before the backlash will help, like the derivative
term in
> the controller. Friction after the backlash will hurt and will
emphasize
> the backlash non-linearity. Tightening the gibs is friction after
the
> backlash.
>
> G. Jackson
>