Re: Rolled vs. Ground ballscrews

Be aware, a servo has to oscillate between encoder counts, it never
sits still. If you add up the distance between encoder counts, the
backlash in the screw and the backlash in the axis drive system then
you get a servo that will be continuously taking up that backlash in
either direction as it 'hunts' for those encoder count edges.
Basically, you are making the distance between encoder counts larger
(from the accumulated backlash), not a good thing.

I've seen in 'professional' retrofits using OpenCNC (high dollar
controller) where they use 2 encoders per axis to get around this.
One encoder on the motor and one on the axis itself.

When it's all said and done, is it really necesary? Get your
ballscrews and set them as tight as you can while still having smooth
movement over the range of travel. Use a motor mounted encoder and
run the thing. I think you will be surprised. If you still think
you need it better later, then you can always swap the motor encoder
for an axis driven encoder. This stuff is complicated enough, don't
make it harder. Just my opinion.


--- In CAD_CAM_EDM_DRO@yahoogroups.com, "Jim Colley" <jcolley@g...>
wrote:

> Dan,
>
> So, having used double ballnuts and preloading them, you got the

backlash down to about

> five tenths. Now, I know that you can never remove all of the

backlash, but I'm wondering

> if you can get it a little lower at sacrifice of a little freedom

of motion. What I mean is,

> assuming you preloaded the ballnuts by using two of them with a

method of opposing

> force separating them (or pushing them together even), that you

would run across some

> areas of the ballscrew where the lead of the screw deviated more

than other areas, causing

> the nut pair to bind a bit.
>
> Initially, I thought that the lead error of a rolled ballscrew was

a cumulative error from end

> to end and that .003in/ft would be a relatively constant deviation

across the length of the

> screw. However, when I look at your numbers, it seems that the lead

error is somewhat

> erratic (assuming accuracy of you measurements.)
>
> But, leading back to my orginal question, as long as I can keep my

backlash around the

> same as what you have, I will be happy. If I were to use a direct

measurement system for

> axis position, such as linear encoders or rotary encoders driven by

a timing belt from the

> slide, I would still have to calibrate the axis position indication

via some highly accurate

> means. Now, assuming that is done correctly, I think it is possible

to hold the accuracy of

> the machine within .001in. Now, before anyone flame sprays me about

thermal variations,

> flexure, etc., I'm just looking at what to take in to account for

error budget from the drive

> system. Any thoughts?
>
> Jim
>
> --- In CAD_CAM_EDM_DRO@yahoogroups.com, "Dan Mauch" <dmauch@s...>

wrote:

> > I installed some Thompson rolleed ball screws (.003 tolerance)

on a mill

> > drill. I used two ballnuts and preloaded them so there is

about .0005

> > backlash on the X axis. I then proceeded to check the accuracy

by using a

> > long drift dial indicator and resetting it. The .0005 backlash

was taken out

> > before the X axis was moved in the + direction