Kossel Mini - Diagonal Rod Length Headaches (Magnetic Joints)

Hello everyone, I've been lurking and using the great resource that is this forum ever since I built my i3 last year. Well, delta calibration has kicked my butt enough that I actually had to make an account and post

Anyway, I know I have to be missing something obvious but I've gone over this stuff time after time and I still can't figure it out. I think it might have something to do with the variant of the Kossel Mini I built but I'm not sure.Here is a link to the design.
Let me also state that I took my time with this build and put a lot of effort into making sure everything was as sound and square as it can be. I've also designed a magnetically removable Z-Probe. It allows me to not have any x or y probe offset and is extremely precise.

So, I'm using diagonal rods with magnetic joints and steel 10mm balls. The rods are exactly 230mm from the tip of each ball. I subtract 1.1mm from each end because that is the measured depth of the countersunk magnets. This gives me the actual joint to joint measurement. So, they should be 227.8mm (230mm-2.2mm).

-I'm running RichCattell's Marlin FW and having a heck of a time getting it dialed in. When I let it auto calibrate rod length it takes it all the way down to 217.28mm. A whole 10mm less than the actual length! (But G29 reports very level results minus a 0.15mm bump in the middle of the bed)
-If I auto calibrate with a fixed rod length of 227.8mm the process always gets way out of whack. I'm talking individual tower radius corrections over 5.0mm. If it does actually complete, it's not correct and it's obvious after running a G29.
-If I adjust diagonal rod length based on printed parts scale I get all the way down to 211.08mm! By this I mean if I print a 30mm cube and it comes out 27.8mm (92.3% smaller than anticipated) then I decreased rod length by 92.3%. This yields dimensionally accurate parts but circles are out of round and things just don't look right.

Here are my current M666 settings. I let auto calibration set diagonal rod length and used an accuracy of 0.05mm. With these values G29 returns almost perfect results and parts are just slightly undersized but only by about 2%.

X (Endstop Adj): -3.77
Y (Endstop Adj): -3.53
Z (Endstop Adj): -2.97
P (Z-Probe Offset): X0.00 Y0.00 Z-21.20
A (Tower A Position Correction): 0.65
B (Tower B Position Correction): -0.37
C (Tower C Position Correction): -0.28
I (Tower A Radius Correction): -0.00
J (Tower B Radius Correction): 0.78
K (Tower C Radius Correction): -1.03
R (Delta Radius): 93.2632
D (Diagonal Rod Length): 217.2805
H (Z-Height): 247.53

I should be happy that it's printing really well with these values, as seen below, but I'm so confused as to what is up with my diagonal rod length. It just leaves an uneasy feeling about this otherwise perfect build.

Any help, guidance, or feedback would be greatly appreciated.

Thanks,
Rob

Quote
robschwieb
Hello everyone, I've been lurking and using the great resource that is this forum ever since I built my i3 last year. Well, delta calibration has kicked my butt enough that I actually had to make an account and post

Anyway, I know I have to be missing something obvious but I've gone over this stuff time after time and I still can't figure it out. I think it might have something to do with the variant of the Kossel Mini I built but I'm not sure.Here is a link to the design.
Let me also state that I took my time with this build and put a lot of effort into making sure everything was as sound and square as it can be. I've also designed a magnetically removable Z-Probe. It allows me to not have any x or y probe offset and is extremely precise.

So, I'm using diagonal rods with magnetic joints and steel 10mm balls. The rods are exactly 230mm from the tip of each ball. I subtract 1.1mm from each end because that is the measured depth of the countersunk magnets. This gives me the actual joint to joint measurement. So, they should be 227.8mm (230mm-2.2mm).

-I'm running RichCattell's Marlin FW and having a heck of a time getting it dialed in. When I let it auto calibrate rod length it takes it all the way down to 217.28mm. A whole 10mm less than the actual length! (But G29 reports very level results minus a 0.15mm bump in the middle of the bed)
-If I auto calibrate with a fixed rod length of 227.8mm the process always gets way out of whack. I'm talking individual tower radius corrections over 5.0mm. If it does actually complete, it's not correct and it's obvious after running a G29.
-If I adjust diagonal rod length based on printed parts scale I get all the way down to 211.08mm! By this I mean if I print a 30mm cube and it comes out 27.8mm (92.3% smaller than anticipated) then I decreased rod length by 92.3%. This yields dimensionally accurate parts but circles are out of round and things just don't look right.

Here are my current M666 settings. I let auto calibration set diagonal rod length and used an accuracy of 0.05mm. With these values G29 returns almost perfect results and parts are just slightly undersized but only by about 2%.

X (Endstop Adj): -3.77
Y (Endstop Adj): -3.53
Z (Endstop Adj): -2.97
P (Z-Probe Offset): X0.00 Y0.00 Z-21.20
A (Tower A Position Correction): 0.65
B (Tower B Position Correction): -0.37
C (Tower C Position Correction): -0.28
I (Tower A Radius Correction): -0.00
J (Tower B Radius Correction): 0.78
K (Tower C Radius Correction): -1.03
R (Delta Radius): 93.2632
D (Diagonal Rod Length): 217.2805
H (Z-Height): 247.53

I should be happy that it's printing really well with these values, as seen below, but I'm so confused as to what is up with my diagonal rod length. It just leaves an uneasy feeling about this otherwise perfect build.

yQI0XQO.jpg

Any help, guidance, or feedback would be greatly appreciated.

Thanks,
Rob

The rod length should be the distance between the centres of the balls ignore the cup bit just measure the centre to centre of the balls once it is all assembled (I suspect you are 10mm to long in your config)

Quote
The rod length should be the distance between the centres of the balls ignore the cup bit just measure the centre to centre of the balls once it is all assembled (I suspect you are 10mm to long in your config)

I've seen this stated elsewhere but no real facts to back it up? Why would it be from the center of the ball, I could see if the counter sunk magnet was 5mm deep and covered half of the ball, but that's not the case. Standard joints are in the center of eye, which makes sense, it's the center of rotation. However, the center of the ball is not the center of movement with such shallow counter sunk magnets.

Here's an image to describe what I mean:

It is certainly the center of the ball, not where you have it drawn. Take the max length you can measure from the outermost spherical points, and subtract the length of one ball diameter.
For "facts to back it up" just watch your own joint in motion. Does the ball rise and fall in the socket? If it doesn't, then it's rotating about the center of the ball, you know, the whole point of a ball joint.

Edited 1 time(s). Last edit at 01/28/2016 04:23PM by Koko76.

Also the proof is in your initial statement where you say auto Cal compensates your rods by 10mm the same as your ball diameter.

The center of the ball is the center of the joint.

Thanks for the clarification guys. It finally clicked when I imagined the joint at a 90 degree rotation.

can i get a copy of your firmware? i cant get my to move correctly, i also have magnet rods