Re: [CAD_CAM_EDM_DRO] Re: Open (i think) design for a parallel robot (reprap may be interested)

Graham Stabler wrote:

> --- In CAD_CAM_EDM_DRO@yahoogroups.com, Mike Pogue <mpogue@...> wrote:
>>So, as long as I specify coordinates in 0.001" increments (which is
>>what I normally do), there's an exact representation in motor
>>coordinates (and no quantization error).
>
> You are rounding which is fine if you are cutting out a square say but
> what about a circle or a diagonal or a spline? You are effectively
> adding the quantization error before you move the motors.
>
> Unless you have a noiseless analogue computer or you stick with square
> components there will always be some quantization error even before
> the motors :)

Maybe I don't understand what you mean by "before the motors"?

With the algorithms I am using, there is no added software quantization
error (other than the error that the steppers themselves introduce). I
am guaranteed to always be as close to the circle in stepper motor
coordinates as it is possible to be with steppers. It's not a PID
algorithm, so I don't have overshoot to deal with.

Note that the quantization introduced by the steppers is way smaller
than the backlash in the mechanical stuff. I'm running the steppers at
200 steps/rev, 10TPI leadscrew, with 16X microstepping. So, the
resolution is about 1/(200*10*16)" = 0.00003125" per motor step.

Circles and diagonal lines would certainly have the largest quantization
error. But, the furthest the steppers can get from a "true" diagonal
line or circle is about 1/2 that number, or about 0.000016".

Mike