Delta printer Auto Calibration

Comments

• Repetier

  July 2015

  I have no stable converging solution that converges to the right values. But if someone makes a pull request with a working solution I will add it. Problem is we have too many variables and we need at least as many measurement point with new linear combinations of all varables.

  
  

  Using

  
  

  http://www.thingiverse.com/thing:745523
  

  
  

  you should get the most important values set quite good as this also measures angles and length while with pure z probing you can not measure them and would modify them just to hover better above the plate regardless or adding some other geometry errors. That is an other problem with autocalibration. Auto can only fix z errors not xy errors.
• MikeB

  July 2015

  I tried this with a Kossel Delta and it arrived at a good result :

  https://github.com/RichCattell/Marlin

  With my big delta I still have a problem of not plain movement.

  I'll try to follow your suggestion.

  Thanks

  
  
• MikeB

  July 2015

  It could be very usefull to calibrate a Delta printer, to have a command (G30 D? for Delta) that will do a Z probe on the various :

  0,0

  0, bed_radius

  -SIN_60 * bed_radius, COS_60 * bed_radius

  -SIN_60 * bed_radius, -COS_60 * bed_radius

  0.0, -bed_radius

  SIN_60 * bed_radius, -COS_60 * bed_radius

  SIN_60 * bed_radius, COS_60 * bed_radius

  And that show an output with the various heights. This can help a lot to calibrate the delta radius and X, Y , Z end stop offsets

  
  

  
  

  
  

  
  

  
  

  
  

  
  

  
  

  
  
• Repetier

  July 2015

  I use scripts for this. That is more flexible and can be better adjusted. Not all deltas can go to bed_radius especially if z-probe has a offset.
• RobertoReis

  November 2015 edited November 2015

  I believe that a tower as a reference to be chosen by the user and accuracy in time to give the command,  it can be a starting point. For example. My endstop Z is fixed because I spend all supply wires to plate lamp sensors and hotend the tower Z and  some wires cross over the endstop (on the side - more stable). He does not like being moved. Choosing the Z tower as a default, it could make the initial measurement in Z tower, take as a reference, adjust the X and Y Towers (endstops) by Z tower and then set the table center. After these four stored values, there would be another measurement between the towers and calculate the individual radius and arms corrections. It would be a relatively simple solution. When its found the value, remake tests to fine-tune and then finalize.

  
  

  I believe that these parameters should be calculated by the host rather than the Arduino as is done with the Marlin. This allows a more elaborate calculation and with better results, as the computer's processing power is much higher than the Arduino, and have a greater amount of memory to work with variables, because from what I saw the Arduino is already near its limit. The firmware could have 3 extra variables that would be just the right angles of the towers as Xp, Yp, Zp. Passed via serial, the host would do the "dirty work"