Zero Backlash

Does anyone mind explaining to me why this machine causes zero backlash?

The fact that you have three different vectors of force continually acting on the print head tends to cancel out all unintended movement on a delta printer. It's not zero, but it does reduce to a very, very small amount of backlash.

Thanks for the reply,

if I understand you correctly, does this mean the forces from all 3 arms (acting in different directions) "pins" the print head between them constraining unwanted movement?

Yes.

Ah yes, the lovely Kossel... I feel a third 3D-printer being built later this year

When it's finished, it's time to sell one of my older ones.

I suspect the zero backlash has more to do with the fact that all axes are gravity loaded, which takes the play out of the joints. You could take a standard Cartesian printer with sloppy joints, tilt it at 45 degrees and get the same effect. (Of course, you'd have some pretty weird overhang issues)

Unless those ball or universal style joints are loaded with a spring or elastomer it is highly unlikely they are zero backlash. The Berrybot magnetic ball joints are truly zero backlash, and stay that way even as they wear. A more important issue is that a delta like this has a very low moving mass, and the forces are straight through the center of the pushrods so there is very low deflection. The weakness, visible in many videos is that the frame is tall and not very stiff and the top wobbles all over the place. That does move the nozzle out of position, and should be addressed. A lot of this slips by because it very difficult to measure true position of the moving nozzle in real time.

I've also heard that the precision with a Delta is very high in the center, but decreased towards the edges of the print bed.

I believe the precision can be very good everywhere. However the speed potential is greatly reduced out at the edges if you consider that a horizontal move takes a huge vertical move when the pushrods are nearly horizontal. So the axis with horizontal pushrod must move very fast, limiting the actual horizontal speed. If you were using servos and had significant servo lag then the fast moving axis might lag behind causing a following error.

I find it amusing that people demonstrate the machines using a tall funky shape vase that by it's shape will not show warping, and can't even be measured for comparison to the model without a coordinate measuring machine! Print me a simple cylinder, print a cube with rectangular holes. I'll tell you in a few minutes whether your machine is accurate. Flat surfaces show defects far better than funky curves. The commercial machines are used not only for look and feel but also to establish fit. We should be striving for the same thing if you want 3D printing to advance,

BerryBot3D Test cube with support structure

A new worlds record for the largest print at least for the BerryBot3D!

Delta resolution picture.

Any non-linear printer is going to have non-uniform accuracy over the build area. This is because any irregularities (play in joints, unequal length arms, unequal spacing of the towers, parallelness errors in the towers or arms) will manifest themselves to different degrees at different points in the workspace because the kinematics are highly non-linear. e.g, say one arm is a little longer than the others. At one point in the work area, that error may cause a 0.02mm error. At another point, it might cause a 0.1mm error. If you have a finely machined printer like the BerryBot, the errors may be small overall, but they will still vary.

The trickier thing about non-linear mechanisms like the delta printers is that the equations used to solve the kinematics rely on lots of things being exactly parallel and exactly equal length, etc. If they are not, there is no place in these simplified equations to include correction factors. If you were to use the full kinematic equations that explicitly included every dimension that could be out of whack, the equations would be hideously complicated and take forever to solve.

With a Cartesian printer, you can usually look at a calibration print (say, a cube) and if the sides aren't parallel, or what-not, it's pretty easy to figure out the required mechanical correction. With a non-linear printer, it's next to impossible to figure out what combination of dimensional errors are creating your problem. You just have to build the printer dead-accurate to begin with.

Edited 1 time(s). Last edit at 01/08/2014 10:16PM by LoboCNC.