Daren SchwenkeI was able to get the inverted tripod kinematics dialed in enough to home the machine.
It looks like my beads on the push-rod cables idea for end-stops is going to work out just fine.
Homing and first coordinated moves!
A project log for Arcus-3D-C1 - Cable 3D printer
3D printable, Open Source Hardware, tripod kinematics, cable driven 3D printer
Discussions
Become a Hackaday.io Member
Create an account to leave a comment. Already have an account? Log In.
That's beautiful ! Hope it won't take long before we see it print something !
So what's the printable volume ? Have you tried to go way up ? Are the wires still following the right path ?
Are you sure? yes | no
The volume as I'm using it inscribes the circle generated by the triangles. It can go a bit larger and still track correctly by a bit, but I wanted my bed below the plywood for simplicity. If you go much beyond where the steppers are, the lines can then jump off the pulleys. As you move from the bottom of the build volume to the top, you trade Z accuracy for XY accuracy.
Right now it's a 300mm dia cylinder 300mm high without loosing too much accuracy. You could just about double that and change nothing other than the rods. To go bigger than double, you would need a larger end effector spacing to accommodate the cable wrapping.
Are you sure? yes | no
That's more than enough for the fixed objective of a cheap 3D printer, but it's so beautiful there could be less cheap more accurate versions :
What I'm thinking about is replacing the rods with threaded ones and the bearings with nuts, and let the motors slide when rotating, that will prevent the error from the rotating pitch.
If my idea is not clear (but interesting) I could make an illustration.
Are you sure? yes | no
I get where you are going with this and it's a novel idea, but the error from the cable wrapping per axis is at most 0.012mm over the entire build area currently. I doubt threaded rods will have that kind of diametric accuracy.
The distance from the pulley to the spooling rod is about 500mm, and the width of the wrapping at worst is 3.5mm. A little Pythagorean math gives us the actual length: sqrt(500x500+3.5x3.5) = 500.0122. So when fully wrapped up, we have introduced a positive error of 0.012mm. Since that error is directly proportional to the current length of the axis and linear, it would be pretty simple to compensate for that in the math if I come to care about it. :)
Where there could be significant gains in accuracy made would be actually modeling the correct shape for the virtual pulleys on the top of the end effector. I guessed.
Are you sure? yes | no
super cool , cant wait to see it print
Are you sure? yes | no