Re: Lead screws - made more accurate

--- In CAD_CAM_EDM_DRO@yahoogroups.com, Doug Fortune <pentam@c...>
wrote:

> Don Rogers wrote:
>
> > Now, we are talking about lead screw accuracy, how do we get

0.001" per

> > foot accuracy when the best lead screw we have is 0.006" per

foot? And how

> > did we come up with one that accurate when a machine that

accurate didn't

> > exist yet?
>
> The machine doesn't have to be accurate, if you can calibrate
> the machine, then apply corrections.
>
> I believe Les Watts described on this list a year ago, a method
> of 'mechanically' correcting a leadscrew (ballscrew or acme) by:
>
> - determining the error (in fractions of a leadscrew-turn)
> - machining the counter-correction into a linear cam
> - a follower follows the spindle along the axis (along the cam)
> and the cam torques the nut appropriately (which applies the

correction).

>
> Of course its far easier these days just to apply the correction
> in software (conceivably even using hardware store all-thread and
> getting reasonable accuracy).
>
> I believe EMC/Linux can do this, I'm just waiting for the idea to

catch

> on in the other interpreter packages.
>
> Doug Fortune
> http://www.cncKITS.com
>

I come from a Coordinate Measuring Machine (CMM) background where
software compensation ("comp") became the norm a couple decades ago.
Since I owned a very large CMM (128" x 72" x 54" travels monitored by
one-micron resolution scales and located in a lab controlled year-
round to +/- a half degree F floor to ceiling), the annual
calibrations were pricey, about $6000, so I watched them like a
hawk. I also helped with, and observed, the initial assembly, setup,
and calibration of the machine. Having seen what goes into the
process, a complex laser mapping to dial out linear scale error, way
waviness, way alignment, axis twist, etc. all performed on components
that were as "perfect" as possible (in both manufacturing and setup)
before comp, I cannot get real excited about simply mapping (or even
worrying about) linear errors of a leadscrew on a home-built
machine. I'm reminded of Butch-to-Sundance (or vice-versa) "Can't
swim? Can't swim?! Hell, the fall alone will probably kill us!" In
our context, most people are going to have alignment/setup
imperfections that make the leadscrew error inconsequential. Axis-to-
axis squareness alone will usually "kill" you... and comp is only for
machine component inaccuracies, not slop/clearance/backlash problems.

However, I suspect most people building their own CNC aren't
machining helicopter gearbox housings. So, while a discussion of the
pros and cons of different types of components is very helpful, it's
not appropriate to make blanket statements I've seen here ruling out
one type or another. Remember that some people are just building
a "few hundred Dollar" machine to play around with or experiment
with, and it may never run enough to see even the slightest wear.
They may be using their machine to route shapes that would otherwise
be cut on a bandsaw following a pencil line. In metalworking, they
may only expect their machine to be a small improvement over (and
more fun than) the "Dykem/scribe line/drill at or mill to the lines"
class of work. I can find justification for everything from chains
to all-thread to ballscrews, and casters-on-angle-iron to THK slides,
depending on use and expectations. We don't all need ground
ballscrew driven carriages with air bearings on polished granite ways
and closed-loop control monitored by three laser interferometers per
axis.