Re: [CAD_CAM_EDM_DRO] Lead screw twist... some musings

dave engvall wrote:

> >
> >
> > Accuracy is of no use in an inherently unstable system. As soon as the
> > phase of the response is more than 180 degrees out of phase with the
> > driving signal, massive oscillations will build up until either something
> > breaks or some overload limit is tripped.
> >
> > Jon
>
> Jon I need some interpetation here. Are we talking 2 counts out in quad. out or can we define this in terms I can understand.

No, the 180 degrees I mention is the phase relationship between the
driving signal to the motor and the response at the encoder. The encoder
response isn't really a linear signal, of course, but we can consider
its position as one for discussion. In any servo system, if the
response of the feedback is delayed by 180 degrees in phase
relative to the signal driving the system, then oscillation is bound
to develop without extreme means (like DSP) to correct it.
The only solutions are to lower the servo system bandwidth to
below the frequency at which this delay is 180 degrees (90 is
a lot safer), or reduce the delay (by increasing stiffness, reducing
mass, etc.)

This is a basic tenet of all control systems theory. I'm no expert
there, but when tuning a servo, this sort of thing is the demon you
are fighting to obtain a high gain, rapid responding system that remains
stable under all circumstances. Those three parameters are mutually
contradictory, and a balance has to be obtained between them.
Increase gain (for less following error at steady-state) or increase bandwidth
(for less following error during transients) and stabilty may be compromised.

Jon