Rolled Ballscrews

I've been thinking about rolled ballscrews and their potential lead error.
It seems this would be a problem for two reasons:

1) inconsistent preload

2) cyclic and cumulative positioning errors

I think the preload problem might be solved by using two nuts fairly
widely spaced with a low spring rate, high damping coefficient device
between them. If such a device were several inches long, the force
variation due to a whopping .010" lead error would be miniscule. But
it would have to be damped enough not to resonate at certain traverse
speed & load combinations.

For the cyclic and cumulative positioning errors, perhaps it's time
that we recognize that software is cheap and iron is expensive.
Programs like EMC could be modified to include a full screw correction
formula. I think this should be done at the motion planner level - hand
the servo loops precorrected position commands so that they can run on
raw rotary encoder feedback.

The problem with that approach is how to accurately map out the
errors. What methods are available for a hobbyist to measure
distances of one or more feet to an accuracy of .001"? For things in
the range of a few inches micrometer setting standards or guage blocks
might work, perhaps coupled with a dial indicator to allow testing for
short period cycling errors. But how linear is the dial indicator,
really? How good are your dial calipers?

Also the machine must have an absolute home sensor - something tricky
on a lathe where certain setups may make it impossible to always visit
the home location.

Chris

-------------------------------------------------
Christopher C. Stratton
Engineer, Instrument Maker, and Horn Player
ccs@... 617 628 1062
http://web.mit.edu/~stratton/www/brassbuild.html