Not sure how to fix this calibration? error

Hi all,

I received a kossel mini that was assembled and used a few years ago. The previous owner used the FW that came with the motherboard because (as I found out) the provided Marlin 1.6.x wouldn't compile.

I have upgraded to Marlin 2.x, gone through a manual calibration (the z probe has issues), and it seemed to be OK, but there is a problem I'd like your help with. By manipulating the end-stop adjustments and DELTA_RADIUS, I now have the Z measurement of the towers and the center all within 1 mm. But the Z measurement of the left side between the A and C towers is way off, and the front side is slightly off. Please see attached diagram for details. What should I be checking for or adjusting to address this? (The build plate is flat glass)

Thanks,
Daniel

That's a strange one.

If you jog the head to X0 Y0 Z10 then to X-85 slowly, does the head move in parallel to the bed?
Are you using any of the automatic bed leveling features?

IMO If you can, it'd be worth fixing the probe to do automatic calibration, if nothing else it's faster.

Thanks for responding and your questions.

The probe has what I assume is a design flaw, although maybe it is related to this. The hot end is on a hinge. When the the nozzle pushes against the bed, the hot end pivots on the hinge, tripping a microswitch. The trip is off a few mm at extreme +y from x0 y0, so it isn't usable at the moment. If I can fix the -X problem I posted about, it may also fix that. Perhaps one of the reasons the previous owner stopped using it is that the thing I'm asking about popped up.

I have not set up bed leveling. That requires way too much hand probing. Since I'm using a glass plate, I shouldn't need it for initial prints.

When I go from X0 Y0 Z10 to X-85, the head gets closer and closer to the bed rather than moving parallel. If I go to X85 it travels parallel. I'm thinking there is a mechanical problem rather than software. Unfortunately, my expertise is software, so I was able to upgrade the configuration.h file pretty easily, but I'm out of my element if something mechanical is happening.

Deltas can be a challenge to get calibrated correctly. Marlin has example configurations for Kossel, you could compare yours to this one (Mini right?) if you haven't already. [github.com]

Have you confirmed that the rods are all the same length?
Are you using the auto-configuration function with manual probing or are you doing something else?
When the head is at X-85 Y0 Z10, are the rods tight or is there some play?
Are there any obstructions that could prevent the carriages from moving freely?

I'll answer all your questions below. Your question about rods being tight has pointed in a very valuable direction; details below. I need to work out how to address that, so will post another response after I work through that.

I used as a starting point the same file as the link you provided, then copied over the relevant values from the Marlin 1.6 config.h that came with the Kossel mini (and wouldn't build on the 1.6 that came with the mini).

Yes, I have checked the rod lengths. I had to re-epoxy one of them because the joint came out of the rod. At that time, I took them all out to use as a jig for re-epoxying.

I was not able to get the autoconfiguration with manual to run properly. I think the issue is that I don't understand how to use it. When I first started it, the effector dropped to X0 Y0 Z0, then nothing. I was expecting a prompt on the LCD. When I didn't see one, I did fully manual using this tutorial to manually set the height limit switch adjustments and the delta_radius: T3P3 Kossel Mini Kit R2 - 19 - Commissioning and Calibration.

When the head is at X-85 Y0 Z10, one of the rods was tight. That was interesting, so I tried the other cardinal directions and found some tightness other times, but not all of the other times. Further investigation revealed that X0 Y0 wasn't in the center of the delta, but pushed in the -Y direction. This is probably mechanical, so I will be checking the struts and corners to get things parallel and perpendicular. This may take a few days - I need to get some metric ball hex wrenches to access those screws. I'll post again when I have worked through this.

There are no obstructions that I can see. The carriages seem to move freely at higher and lower speeds.

Thank you for the great suggestions. You may have identified the issue.

To close the loop...

I have been working on this since then. It now works well enough that I was able to 3D print one of the pieces that holds the hot end in place and am using that instead of the original, which cracked. Here is what I have learned so anyone who sees this can follow in my footsteps.

1. I never did a calibration of the a, b, c stepper motors, only the extruder motor. Doing that manual calibration helped a lot.
2. I found a number of loose screws on the carriages, which I tightened.
3. The rods became tight at X70 and X-70. With all of the below fixes I still can't get to a radius of 80. This may be the best my build can do.
4. I stopped using the Marlin 2.x and went with Marlin 1.1.9 AC, found here: https://github.com/LVD-AC/Marlin-AC/releases/tag/G-force9-AC. There are some special fixes in this branch for the G33 calibrations that work pretty well.
5. With all of the above fixed, I was able to calibrate the z sensor so it was usable.

I had a lot of help from people in the FLSUN 3D Printers Facebook group https://www.facebook.com/groups/FLSUN3DP/ who actively use this exact model.

Thank you for your help, too.

I'm glad you got it sorted, enjoy!

Basicly you need to calibrate 3 things:

The delta height (or the centre of the build plate)
The end-stops to compensate for a titled build plate
The tower angles wich result in a saddle shape deformation (towers are ok, but in between not)

G33 does the 3 in one go.