Z-Probe Repeatability Test

This is not enabled in the firmware as per marlinfw.org/docs/gcode/M048.html. I can't remember seeing it as an option disabled as I've worked on firmware. I will have a look at my versions. I have checked all the firmware that I have edited so far, and this option is not part of the code you will need to edit this into the code using a text editor, my choice is notepad++.

Ok, enabled M48 in the firmware and ran few tests:

Cold bed, Ultrabase glass surface:
Mean: -0.183341 Min: -0.187 Max: -0.180 Range: 0.008
Standard Deviation: 0.002573

Mean: -0.185099 Min: -0.190 Max: -0.182 Range: 0.008
Standard Deviation: 0.002928

Mean: -0.188112 Min: -0.192 Max: -0.185 Range: 0.008
Standard Deviation: 0.002561

-------------------------

When bed was heated (Ultrabase glass surface):

As soon as printer reported the bed and nozzle is heated up:
Mean: -0.592611 Min: -0.616 Max: -0.559 Range: 0.058
Standard Deviation: 0.020834

1 minute later:
Mean: -0.624499 Min: -0.632 Max: -0.614 Range: 0.018
Standard Deviation: 0.004869

5: minutes later when teperatures were stabilized
Mean: -0.636551 Min: -0.642 Max: -0.627 Range: 0.015
Standard Deviation: 0.003890

--------------------------

When removed glass (test with original Buildtak-like surface, bed heated):
Mean: -4.609233 Min: -4.624 Max: -4.599 Range: 0.025
Standard Deviation: 0.006812

Mean: -4.609986 Min: -4.616 Max: -4.601 Range: 0.015
Standard Deviation: 0.004797

Mean: -4.612748 Min: -4.634 Max: -4.599 Range: 0.035
Standard Deviation: 0.009273

---------------------------

Ultrabase glass surface back on, bed heated, surface lightning turned off:
Mean: -0.665928 Min: -0.672 Max: -0.662 Range: 0.010
Standard Deviation: 0.003186

Mean: -0.664171 Min: -0.667 Max: -0.659 Range: 0.008
Standard Deviation: 0.002511

Mean: -0.658898 Min: -0.664 Max: -0.652 Range: 0.013
Standard Deviation: 0.003630

---------------------------

Lights back on:
Mean: -0.655885 Min: -0.667 Max: -0.642 Range: 0.025
Standard Deviation: 0.008774

Mean: -0.651365 Min: -0.662 Max: -0.642 Range: 0.020
Standard Deviation: 0.007316

Mean: -0.654127 Min: -0.667 Max: -0.644 Range: 0.023
Standard Deviation: 0.007448

---------------------------

Double the Z-probe speed (HOMING_FEEDRATE_Z), lights off:
Mean: -0.459033 Min: -0.463 Max: -0.448 Range: 0.015
Standard Deviation: 0.004901

Mean: -0.463553 Min: -0.468 Max: -0.456 Range: 0.013
Standard Deviation: 0.004414

Mean: -0.466064 Min: -0.473 Max: -0.458 Range: 0.015
Standard Deviation: 0.005552

-----------------------------------------------------------

From the numbers I got, it is clear that:
- it is good idea to heat bed and nozzle before some time before you start your print (Prepare -> Preheat ABS/PLA)
- it is good idea to do the bed leveling (setting Z-offset) when your bed and nozzle is heated
- the IR sensor I use for glass surface isn't that immune to ambient lightning as the blue-paper says. Try to avoid direct light/sunlight on the bed surface during bed leveling process if you use this kind of sensor.
- it looks that speed of Z axis when doing leveling has some impact on the precision. Would be interesting to find a sweet-spot... a trade of between speed and accuracy. It might be possible to go faster and do more probe points for bilinear mesh rather than go very slow and have less points... A question for long winter nights ;-)

You should try running M303 PID Autotune for your nozzle, the danger of allowing the hot end to heat without extrusion is carbon build up in the nozzle which in turn causes blockages. Better to autotune the hot end to the temp you want, it will come up to temp faster and be more stable. If you are running the original firmware though for some reason the autotune function seems to be buggy. Marlin 1.1.4 is stable and autotune is good to run.