Re: [CAD_CAM_EDM_DRO] Gecko servos

Tim Goldstein wrote:

> I did get the driver to fault
> if I did a reverse of direction at 180 ipm. This sort of surprised me
> as I
> expected EMC to apply the acceleration to both the stop and start, but
> this
> did not appear to be true.

One thing to know is that EMC is much more aware of position than
velocity.
It has a desired position, and computes a desired velocity to reduce
position error on the next sample. The velocity that this will take is
not
the smoothed one. What this means is if you are running near the speed
limit, and a substantial following error is present, and you decelerate
rapidly, the error will switch sing very quickly, because the motor's
actual position was already heading in the right direction due to the
buildup of following error.

(I haven't expressed this very clearly, so please try to visualize a
smoothly
computed trajectory, and the motor's vain attempt to follow it, and then

the computer trying to smoothly reverse everything.)

What happens here, is that the motor goes suddenly from being behind
the curve to being in front of it. This causes a big discontinuity, and
you
can see why the driver gets caught in the middle, and is forced to
exceed
its capabilities.

This is why you have to limit the max velocity and acceleration a
little, to
give some headroom at the top end of the curve. Probably reducing
either the
max speed OR the max acceleration will prevent the faults.

>
> I can't wait to try this setup out on my machine. The steps work out
> to be
> .0000625" apart which is a bunch better than the .00025" I have with
> the
> steppers. If I can get the machine to do rapids at 90 ipm and reduce
> the
> stair stepping on long tapers I will feel the time and money to switch
> will
> be well invested.

Yes, it should reduce stair-stepping quite a bit. What you will find,
though,
si that you really can't make the machine move in .00006" steps.
Probably, it
will not even move in .00012" steps. It would be interesting to put a
sensitive
dial test indicator on it, and make incremental jog moves of .0001" and
see
how many of these it takes to make the machine jump. (With a servo
system,
you have a LOT of gain between commanded position and actual position,
so it tends to jump every time. With the open-loop steppers, the gain
is
1, I guess. With a closed loop stepper, it will be very interesting to
see the
performance.)

Jon