Draw-Wire Calibration/Positioning System

Technical Details
,
141

I’m liking the idea of using the current motors and chain for support and
movement, and Kevlar cord running from the sled to the bottom corners for
position encoding. The tension on the measuring cords would also serve to
reduce sag in the chain, and thus improve accuracy in the bottom corners.

sag in the chains wouldn’t matter as we are positioning based on the measuring
cords

I’m visioning it something like this: one end of cord on another ring or
linkage at the bottom of the sled, going to a very small pin or pulley beyond
bottom corner (mirroring motor placement?), then up to encoder shaft/spool
mounted on top bar (axis parallel to top bar), two turns around encoder shaft,
then down to a weighted pulley, then up to anchor point on top bar.

have some sort of cap on the router so that the measuring cords can attach to
the top of the router directly above the bit.

the encoders should be above the workpiece so that they are less likely to catch
sawdust and chips.

We can save money by using relative encoders rather than absolute,

the difference in cost between relative and absolute encoders is pretty small.

This would also get the measuring cords out of the way while loading and unloading sheet stock…

so would having them above the workpiece, and there they would be out of the way
of chips as well.

One problem I see with the thought of driving the sled with a triangle above,
and locating it with a triangle below, is that the opposite corner vectors
will seldom be collinear. I can imagine a scenario when the sled is somewhat
off-center, and the direction of motion desired is collinear with one of the
sensor vectors, but not collinear with either of the chains, so the intuitive
use of the bottom right sensor encoder to control the top left motor by simply
flipping the sign in the current code would fail.

the control is not that trivial, you don’t control the left motor based on the
right encoder, you use the encoders to figure out where you are and then power
the motors to move you towards where you want to be.

David Lang

142

Yes, that’s what I had envisioned. Rather than a takeup spool/encoder like the expensive assemblies mentioned, the cord goes up to an encoder at the top, then the excess is stored and tensioned by a weighted loop hanging down from the top bar.
Using a weight also keeps the tension on the bottom cords constant, which is better than a spring, which would provide the least tension when the most is needed.

143

Hi, hard to tell if this has been suggested, so in case not, as I understand it, we are thinking about the draw wire for position measuring (continuing to use triangulation - but not subject to chain sag - and needed because chain sag is proving to difficult to calibrate for). What if the draw wire fixed point was attached to the point the chain attaches to the ring/bearing, and then travels to the same point where the chain comes of the motor (motors remain in standard position - need the chain to pass over a fixed roller to have a fixed chain point also). After passing over the rollers in top left/right, could then travel to the draw wire encoder - where the internal spring keeps the tension?

I also noted the comments on duty cycles and temperature issues for draw wire encoders here although suspect the temperature ones are related to the wire used in this case.

144

KISS: A single encoder with high (enough) resolution, combined with the current scheme for tracking position, could be used to reveal the current system’s inaccuracies.

Our current positioning work should be augmented, not tossed out in favor of another method. As I understand it, we’re talking about identifying/compensating for error of approx -0.5-15mm, otherwise the current system is ‘accurate’.

A single draw wire placed on the Cross-Beam Center position, and attached to (any) point on one of the (3) mounting systems should allow us to compute the sled position, sled rotation in degrees, and ultimately: bit position, without need for other complicated apparatuses. (assuming the variables for each mounting system are input accurately)

145

Helping to minimize the draw wire length to cost calculation, one can attach a length to the encoder wire equal to the cross beam height above the workpiece. leaving us with maximum draw wire length of ~1.75m.

146

Furthermore, an accurate drawstring would allow a user to more easily measure the machine during assembly by attaching to each motor mount and measuring inward.

As it stands, I think we could consider a well tuned maslowCNC as accurate enough for ‘hobby making’ creations. The current results don’t lie. As is, a well built maslowCNC will suit many just fine.

However, Its not unreasonable to me to push the cost envelope out another 50-100% to gain superior (sub-mm) accuracy, especially without radical re-engineering of the machine.

https://www.phidgets.com/?tier=3&catid=4&pcid=2&prodid=632

Please shoot holes in my theory, but wouldn’t we be able to calculate and compensate for all positioning error with a single encoder?

We’d have ‘2 legs of the triangle’ from sled to crossbar to motor, and our chain distances. Isn’t that enough to solve it?

147

This is not a valid assumption. If this was a valid assumption, the original
quad kinematics would have worked pretty well. The entire reason to need to move
to triangular kinematics is that the actual physical values of the machine could
not be measured accurtely enough.

David Lang

148

but the draw wire can measure to within 1/128"

149

note the difference between repeatability and absolute accuracy in the specs. A
lot of the ones that I’ve been pointed at have very good repeatability, but the
accuracy isn’t that good.

The existing system with the encoders on the motors is very repeatable (to
about 1/1600" if you have the tensioning right). Our problem is absolute
accuracy.

David Lang

150

Hello, I have experience with these and they were a failure. Could never calibrate them to read constantly. Changed to Sick units with wire and no longer had a problem.