Lead screw compensation

Greg Jackson wrote:

> Does anyone know if CNCpro or other inexpensive parallel port G-code
> interpreters offer compensation to make the rolled thread ball screws work
> as accurately as the ground thread screws? It has been suggested to me
> that I would need a laser interferometer ....

I might be missing something here, but after our machines are built,
the actual travels can be measured by various means, the differences
from theoretical can be determined, and finally any gcode that you are
prepared to execute can be run through a program to add & subtract
these variances depending on the current position (the errors being
kept in a 2D or 3D compensation matrix) to make a new NC program
which compensates for the machines errors.

I guess no-one has written that program yet! Excepting arcs
(which probably have to be piecemeal replaced by shorter segments),
such a program is quite straightforward. It would compensate nicely
for ballscrew & acme errors over both short and long distances.

Assuming such a program existed (eventually inevitable), I suppose
measuring 10 times an inch to a 0.000 10" on each axis, once at the
beginning of the machines life, would be entirely sufficient until
significant wear becomes a problem. For the hobbiest who does
not have a laser interferometer in his bottom drawer, measuring
every inch to a thou could form the basis of a rough compensation
matrix (and perhaps even allow 'hardware threaded rod' to be used
with sub .001 accuracy).

How to achieve this calibration? Might I suggest using the "mechanical
leverage of pulleys attached to an encoder" idea again (darn, that
concept is powerful). Such an arrangement could quickly fabricate
a compensation matrix of any required density.

Doug Fortune