Re: Why not lineal feedback, instead of rotary encoders?

Hi,

Torsen gave a very good reply. Let me just add this to his reply.

The encoder is there because the motor must have it in order to close
a stable feedback loop. That requires a tight "compliance" between
the motor and the feedback element. What that means is there should
be no (or vey little) backlash in either the motor reduction
gearing, or more significantly, in the rotary to linear transmission
(read leadscrew).

This can be difficult to achieve in a single feedback loop.

A more elegant solution uses an interior and an exterior loop. The
interior loop uses a rotary encoder on the motor; its function is to
stabilize the motor only. The exterior loop uses a linear encoder to
read the actual position of the axis, and based on its readings,
adjusts the commands to the motor (interior loop).

This way the motor (interior loop) is stable and the position of the
axis (exterior loop) cancels any backlash and leadscrew error.

It is more expensive, (two loops, two encoders) and you must have
software that accomodates position feedback to make the error
adjustments based on the exterior loop.

If you can live with 0.001" to 0.0005" error, live with a single
loop, (encoder on the motor or the leadscrew for a toothed belt
reduction). If you need better, you need to go with a dual loop
system.

Mariss


--- In CAD_CAM_EDM_DRO@yahoogroups.com, "Shane M Dwyer"
<dwyersm@p...> wrote:

> I have been looking around the groups, www, commercial vendor

sites,

> etc and have noticed that positional location feedback is always
> determined by an encoder mounted to the screw or motor. I can't
> figure out why the actual position of the object (table, workpiece,
> tool, or whatever)is not determined by a linear transducer, eg a
> scale not unlike one used with a dro.
> Wouldn't this be more precise. We are intending mostly to measure
> linear - planar, and not a rotating object's final position

afterall.

>
> And "final position" doesn't have to mean stationary, it means

final

> before the next reading, position update, polling, and so on.
>
> It is the object's location that we are currently interested in,

not

> the rotation of or position of the screw or screw motor spindle.
>
> Measuring the screw's location (ie. number of turns or part thereof)
> is not quite the same as the actual location of the work - due to
> sloppiness in anything, the screw, bearings, screwnuts, collars,
> retainers, gibs, gears, belt drives, transmissions, friction, etc .
>
> A positional feedback system derived from measurement upon the

actual

> article we are forcibly locating would then be somewhat immune to

the

> abovementioned sloppiness, wouldnt it?
>
> So why isnt it done?