Re: Arm length calculation help on own-design delta needed

Quote
sheepdog43

Quote
hercek

Quote
sheepdog43
Also, if you plan to auto level or auto calibrate, just reaching the edge of the build plate is not enough. The further out you go, the more flat one arm will be, leaving you little to no leverage to depress the probe. Even if you do have a little extra, I recommended not probing more than 75-80% from the center of the build platform to reduce wear and tear as it is VERY hard on your diagonals doing this.

I'm not sure this is true. At the location most far away of a tower, the arm of that tower will have small angle to bed (in ybanrab situation about 24° which is not that small anyway). But the arms to the other too towers will have quite big angle to the bed (about 63°), those arms should take the most of the probing force (which is in the Z axe direction). Maybe this situation may be bad when the pairs of arms going to the same tower are too near to each other, but hopefully all people realize that the distance between arm pair should be as big as possible (within the design constrains of the effector platform).
An original rostock here does not have problems probing about 80% away from centre.

All three sets of arms are needed to hold an effector in place horizontally. If one set of arms is horizontal, they are no longer holding the effector flat and the effector can rock. It doesn't take much to throw off calibration. Yes, two sets are holding it down, but we have three sets for a reason. Take off a set of arms and you will see where the problem is.

In such a position, the horizontal pair does not need to prevent rotation in the horizontal axe perpendicular to it (other two arm pairs will do it just fine). The horizontal pair still needs to prevent rotation along vertical axe (which would lead to smaller platform tilting too). Maybe you are right and these positions are significantly worse even when arm pair distance is big enough but I have doubts without proper momentum propagation analyses.

Quote
sheepdog43

Quote
hercek
I do not understand why bigger printer would need bigger accuracy of parts barring the problems of frame/rod elasticity (probably what you mean by deflection).

When you calibrated your printer, remember how the print was bowl shaped? It's not linear, the further out you go, the worse the bowl shape gets.

So lets say your Kossel diagonal is accurate to within 0.04mm giving you great precision. However, in reality, you probably aren't perfectly flat. There can still be 0.01mm bowl shape in your print, you just can't spot it and it's irrelevant.
That 0.01mm bowl shape at 170mm wide balloons up to 10mm by the time you reach 500mm wide. You need to dial down to something like .0015 diagonal accuracy just to get a .01mm accuracy across the bed at 500mm.

Here we claim essentially the same thing. Non linearity causes the errors to be bigger at the heat bed edges. But we are looking at it from different angles:
* you say this is bad because small calibration errors lead to big effector errors at bed edges;
* I say this is good because the small calibration errors can be detected at the edges (and thanks to that we can correct them).

Btw, I did not have bowl problem at the beginning. I had a serious tower position problem which leads to a different error shape. Anyway diagonal rod and tower position errors can be easily fixed with a zprobe (measuring just below the head) and the math in this notebook [github.com]
The comments in it are a bit obsolete (e.g. it is known now that it converges nicely if at least 8 calibration points are used). But it allows to calibrate as precisely as micro-step length and printer play (slop) allow with both marlin and repetier. I'll update it sometimes. To get good results, one needs to probe points when one arm pair is almost horizontal (where the errors are biggest thanks to the non linearity).