Re: tach output

Jon,

What you say may have been true at one time but things have gotten a
little more sophisticated and refined now. There is an anti-dithering
circuit in the current versions that takes care of this very real
objection. It functions by knocking back the gain and damping to 5%
of set value anytime the motor is within +/- 2 counts of the command
position. The integral term dominates within this range and remains
fully effective.

The net result is you would have to be in a very quiet room to even
hear the motor dither, let alone to thrash back and forth enough to
cause motor heating or cause wear of the mechanical parts. You really
would need to see it.

Mariss

--- In CAD_CAM_EDM_DRO@y..., Jon Elson <elson@p...> wrote:
> David Micklethwaite wrote:
>
> > "Jon Elson" <elson@p...> wrote:
> >
> > >
> > > Well, I have a strong preference for real servos. The
difference with the
> > > Gecko drives is that the CPU has no position feedback. It
sends step
> > > pulses to the Gecko, and hopes it got there.
> >
> > Is it possible to use a linear encoder of some sort to provide
the feedback
> > to the Gecko driver in place of a rotary encoder on the servo
motor shaft? I
> > realise that this would still not provide position feedback to
the CPU but
> > it might provide some compensation for backlash in the lead screw.
>
> Well, sort of. the problem is the Gecko will sit there and bang
back and forth
>
> across whatever backlash continuously, all the time that the geckos
are
> enabled. This will be not only nerve-wracking, but will wreck the
screw
> and nut quickly. It may also overheat the gecko or the servo motor.
> If the backlash is more than .001", I think you would find the
system
> unusable. The gecko would accelerate across the backlash in open
> loop runaway, as it is, indeed open loop, because the motor is
moving, but
> no feedback of that is coming back from the encoder (and there's no
> separate tach). Then it hist the other side of the nut and bumps
the
> encoder hard, overshooting the one encoder count it was trying to
> adjust. Then it starts accelerating rapidly the other way.
>
> Clearly, not an example of stability.
>
> Jon