Delta calibration variation

General Discussion
1

Dear klippers

I have a delta printer where I’m using klipper.
However, I find it hard to have consistent dimensions without the head bumping in the part when the height of the print grows. (it behaves ok as long as the print remains small enough.)

I find it remarkable the parameters (distance, endstops, bed_radius, …) of the printer change heavily when executing subsequent calibration procedures.

I’ve started with the default distances (coming from the original firmware), which all have equal arm_lenghts, end stop positions etc.

  1. I start with the automated calibration (DELTA_CALIBRATE) with this probe attached to the nozzle. Since there is no x,y,z-offset, no tilting effects should be experienced.

This results in:
[stepper_a] angle = 208.970492 arm_length = 348.701581 position_endstop = 508.242816
[stepper_b] angle = 329.933911 arm_length = 351.813490 position_endstop = 507.859019
[stepper_c] angle = 90.000000 arm_length = 346.976579 position_endstop = 509.226186
[printer] delta_radius = 165.946893

  1. I start doing several enhanced calibration and measure the test prints, I’ll not post the entire result but only the stepper_a, to not clutter the topic.

stepper_a: position_endstop: 508.860334 angle: 209.262733 arm: 345.047713
stepper_a: position_endstop: 508.466941 angle: 209.835267 arm: 347.127509
stepper_a: position_endstop: 507.501071 angle: 210.036512 arm: 352.269988

As you can see there is a really big variation in e.g. the arm length. in the last print, the first layer is just complete rubbish.

My questions:

  • Am I the only one experiencing such effects, any advice how to get this more consistent?
  • is the entire enhanced calibration concept not missing a check of dimensions in the z-axis to make this a more reliable procedure?
2

My advice is to not use the automated probe, and instead use the manual probing method described in the documentation.

FWIW, I’ve seen way too many reports of issues with a delta that were ultimately due to the probe that I, personally, wont spend time investigating a calibration issue when using a probe.

is the entire enhanced calibration concept not missing a check of dimensions in the z-axis to make this a more reliable procedure?

Distances on the z axis of a linear delta is entirely controlled by the tower’s rotation_distance parameter. Thus, it is not “calibrated” as one can more accurately calculate it by inspecting the hardware - Rotation distance - Klipper documentation

Cheers,
-Kevin

3

thanks for the response
But it doesn’t really matter, if I do calibration manually or using a zero-offset sensor. I always seem to have quite some variation.

I also have to do manual editing of the bed_mesh after calibration before I get the first level correct over the entire bed. I do this by printing a bunch of levelling squares over the bed until it becomes a good consistent first layer.
Since after that tuning effort, the first layer is repetitive I assume it’s not related to the hardware of the printer causing such possible variation.