[CAD_CAM_EDM_DRO] Re: Newbie trying to help others make sense of it all

Karl Peterson wrote Sat, 11 Aug 2001:

> I am working towards a X-Y-Z-Rotary table CNC 3 in one
> Mill/Lathe using a CAD/CAM solution in my house tied via
> sneakernet to an EMC controller in the shop using steppers
> with encoder controls. I have the Box for the Linux/EMC to
> run in, am getting the BDI CDRom downloaded tonight. I have
> a slew of weak Steppers, four 5A controllers from
> CAMTRONICS, and the DRO board.
>
> I am stymied on the encoders. I have seen pictures about
> home built ones with Parts from US Digital, but they don't
> make sense to me. I cannot visualize how they attach to the
> table and provide position data and cannot see how to put
> them on my little hunk of iron.

Ray Henry wrote Sat, 11 Aug 2001:

>I've seen and heard about dozens of ways using wire, fishing line, toothed
>belts and such that run alongside each axis but for my money/time I'd go
>for ball screws with rotary encoders. I used the 5/8 stuff from McMaster
>and found about 0.004 backlash with single nuts. Zero lash with Roland's
>approach using tapered spring washers between two.

Hi:
This boils down to using linear encoders or rotary encoders.
True linear encoders attached to directly sense the position of the
table are the 'Rolls Royce' of position feedback methods. But
they are pretty expensive (USDigital is a 'bargain') and have
to be protected from swarf and shocks.

Rotary encoders (much more affordable) can be used two ways:
- directly attached to the screw or the motor. The downside is
that you are NOT directly measuring the table movement, but
are trying to interpret the table movement from the rotary
motion. This means these inaccuracies are introduced:
- if the threads are not true, then you will have displacement
that varies along the length of the thread, from a low value
of maybe 0.001" per inch, up to lots (for hardware store
ready rod). Theoretically you can compensate this in software
but no-one has done it for hobbiest software yet (that I know).
- backlash of nuts (and maybe motor couplings) could range up to
10 thou or more.

- attached to wires (attached to the table), or attached to pinions
which run along racks. The errors include cable stretch, and for
racks, the degree of trueness of the rack.

IMHO, if you want feedback under 0.000 50" then linear encoders are
the only way to go. The USDigital stuff can give you 1/1440" ==
0.000 6944" which is 1440 steps/inch (obtained by 360 steps/inch in
quadrature).

Doug Fortune
http://www.cncKITS.com