Re: [CAD_CAM_EDM_DRO] Re: How bout this idea??

Doug Fortune wrote:

> "Paul R. Hvidston" wrote:
>
> > Having only a linear encoder on the table for feedback would lead to
> > infinite loop gain during back-lash and like stated before, the result is
> > oscillation/instability, position error and poor performance.
>
> That statement is true, IF the backlash is a problem....
> but IF the backlash is not a problem (ie has been sufficiently
> nullified by various means), then I maintain linear encoders
> on the table for feedback are are OPTIMAL.
>
> Having the linear encoders on the table (directly measuring
> the table's movement) is therefore the goal (ie not something
> to avoid as might be read into your above statement).

Absolutely true. But, on a light and sloppy machine with Acme
screws, it is more likely to be a real problem. On a very tight,
new, high-class machine with anti-backlash nuts in good
condition, it is very good. But, one rule of motion control
systems, is that you really want 10 x the desired system resolution
in the encoders. If you want .001" system resolution, then you
need a linear encoder with .0001" resolution. These linear encoders
are quite expensive! I have a 1000 cycle/rev encoder on my
5 TPI ballscrews, and that gives me .00005" resolution for a
LOT less expense. (That is 50 uIn, not a typo, there. 1000 Cyc/rev
= 4000 counts/rev in quadrature, x 5 TPI = 20,000 counts/in.)
This "rule" is based on the position quantization inherent in any
encoder causing velocity fluctuations in the servo loop.)

Jon