Unsolved Mystery. Weird Z0 behavior around build perimeter.

[Cedric_K]

Sorry to re-open an old topic, I heard about this Rostock Max which can't be calibrated between tower for long, and that was exactly the reason I didn't bought this machine (not rich enough to buy a printer I was not sure to make work properly).

I wanted a big delta printer, and I couldn't find a better deal than Rostock MAX, so I decided to design and to build a looking-like printer by myself. Dimensions are quite the same than Rostock Max, it also use official Onyx bed as buildplate. But no belts, no roller on extrusions, each tower is made of 2 smooth rods and 1 ballscrew. The frame is VERY rigid, 2020 aluminium extrusions everywere.

Something was really crazy at first calibration tests : I got exactly the same behaviour as described on some Rostock MAX, I was getting the nozzle raising between towers (it's RADDS powered, so 32bit arduino due based, and running repetier-firmware). It was just impossible to fine-calibrate, I had adhesion issues one these areas of the bed, and it took months to understand about that and to solve it.

I tried everything : added a Zprobe, used rotation matrix, used bed_correction_distortion, etc... nothing was working.

I was also using 288mm Magballs arms from Ultirobots, and configured "Delta_Arm_Lengh" to 288mm, and that was the mistake. The day I took a dial gauge instead of the hotend, in order to calibrate every parameters (2 endstops, 2 angular correction, delta_radius and arm_lengh), thanks to echer3d least_square calibration routine (also implemented in reprapfirmware), I found the correct value I had to enter for "arm_lengh" was around 296mm. Some iterations later, I was launching my first big print, so happy to see my nozzle keeping exactly the same exact height in each point of the bed. Just a good setting of the z-height, and It was perfect. No more z_probing, No more software correction, its now totally useless.

Hope it could help. Maybe Initial arm_lengh from SeeMeCNC into repetier are a bit wrong, but works on some printers because they compensate bad geometries, or even, there are several bad parameters, succeeding in canceling together enough to get prints, but not enough about this behaviour between tower.

I'm sure about something : each parameter MUST BE modified individually on each printer. When I change something on my delta, running a new least-square calibration changes some critical parameters as "arm_lengh" and "delta_radius" (just a bit), I have to update these value if I want to keep a perfectly-calibrated machine. If not done, there are always some random points (not so random, always after I reach a virtual triangle were towers are the corners), with some deviation. Sometimes, the deviation is so small , that you could always print. Sometimes not.

ENDSTOP settings are not enough to keep a delta printer well calibrated.
RADIUS setting is not enough to print flat,
YOU CAN't change only one parameter, this is always a balance between each of them.


don't worry about dimentionnal accuracy when modifing "arm_lengh" , there is always one unique good value, if you print perfectly flat, your value will be close to perfection.

Or maybe what I say has nothing to do with Rostock MAX behaviour, i not sure about anything, I dont get this printer. just wanted to help a bit the community, with what I heard here, and what I discovered on another delta with same dimentions.