How to interpret autocalibration results in delta printers

giostark

Hi all,
More or less noob questions...
Aside all the reading stuffs, I would like to know the direction on how to interpret the autocalibration results.
I leave the M666 to zero so that the autocal build all the parameters by it self all the times.
The length of the rods are fixed. 6 factor for the bed , max radius as i can set, etc etc.

 6/4/2020, 6:56:45 PM 	Calibrated 6 factors using 16 points, (mean, deviation) before (-0.009, 0.045) after (0.000, 0.043)
6/4/2020, 6:55:32 PM 	Calibrated 6 factors using 16 points, (mean, deviation) before (0.009, 0.039) after (-0.000, 0.034)
6/4/2020, 6:54:20 PM 	M98 P"0:/macros/AUTO CALIBRATIONS/3xG32"
Calibrated 6 factors using 16 points, (mean, deviation) before (-0.002, 0.048) after (-0.000, 0.045)
6/4/2020, 6:45:48 PM 	Calibrated 6 factors using 16 points, (mean, deviation) before (0.007, 0.034) after (-0.000, 0.031)
6/4/2020, 6:44:34 PM 	Calibrated 6 factors using 16 points, (mean, deviation) before (-0.001, 0.042) after (0.000, 0.040)
6/4/2020, 6:43:20 PM 	M98 P"0:/macros/AUTO CALIBRATIONS/3xG32"
Calibrated 6 factors using 16 points, (mean, deviation) before (0.010, 0.034) after (-0.000, 0.031)

If the result before the pass present less deviation and then more , there are mechanical issues , right? This mean that the printer rebuild new parameter at any pass.
If the pass are all similar , same deviation more or less, it mean that the autocal is coherent and should not be present much mechanical assembly errors, right?
Which is the ideal results? Make the after = to 0.000 should mean zero mechanical errors? Or maybe just that printer compensate very well?
Could be interesting put in the guide how to read those data for give the right meaning to them.
https://duet3d.dozuki.com/Wiki/Calibrating_a_delta_printer#main

Thanks for your time!

oliof

Varying results means either a probe with bad repeatability, or build defects that aren't consistent across probe runs, i.e. lack of rigidity/play somewhere.

If the deviation converges towards a number then that mechanical error is not affected by variation in measurement, but still is present. But it still is there. The firmware can compensate for those (but it's likely better to remove them if you have the means -- tools and skill. I don't and trust the software beyond a certain point).

giostark

@oliof said in How to interpret autocalibration results in delta printers:

towards

I see , ok I'll go investigate my build deeper...