Auto_Bed_Leveling: Z-Probe Repeatability code

[Arno_Millenaar]

Hi Roxy,

Thanks for writing this nice feature for testing the Auto_Level sensor. I noticed your post at the Marlin GitHub and hence found this forum. Next i patched the Marlin_Main.cpp and tested my setup.

The results are very interesting indeed. Without retraction the deviation is at what I consider at the theoretical limit of the Z-Probe design I have as I get an astonishing 0.000691 (yes that’s right, 3 leading zero’s) i.e. better than 1 Mill accuracy. (1/1000mm).

However when adding ‘Retract’ this drops to 0.004797. This indicates that the ‘shaft-rotation’ in my design still induces a small deviation or I need to make a more accurate ‘head’ for the sensor probe. However if you then think about the BED flexing up and down when being heated even when this is done under PID control, then I believe that any further improvements will be futile as Bed Bending caused by temperature fluctuation and moving cooling fan’s around while printing will most likely be a bigger factor.

FYI: I use a M5 Ultra-Mini Proximity Sensor in combination with a bronze probe with a small metal nut on top, retracted by a mini servo. However the Servo is not part of the measurement and only used to lift the probe. Still I noticed that due to the lifting the probe, it slightly rotates each time its gets lifted and retracted, hence the difference in results when adding ‘retract’ to the M48 test command.

This Z-probe design I made can be applied to many extruder setup’s so if anybody is interested you can find a description of the setup at the following link: http://creatrtips.freeforums.org/vie....php?f=8&t=160

Here is a link for the mini proximity sensor I use: http://jimou.en.alibaba.com/product/...ty_sensor.html Type 10-30V NPN N.O. with a switching distance of 0.8mm

I also noticed (especially when using mechanical switches) that slowing down the second and final probe movement increases the accuracy significantly. I therefor patched the ‘homing_feedrate[Z_AXIS] ‘ lines within marlin_main.cpp to be divided by 8 where the default is only 2. This makes a notable difference. Even for a non-mechanical sensor such as an induction proximity sensor.

Results from M48 testing:
Without Retract:
10:25:02.150 : M48 Z-Probe Repeatability test. Version 1.85
10:27:22.410 : Mean: 12.281874
10:27:22.410 : Standard Deviation: 0.000691
10:27:22.410 : echo:endstops hit: Z:12.28

With Retract:
10:33:23.254 : M48 Z-Probe Repeatability test. Version 1.85
10:36:09.816 : Mean: 12.280522
10:36:09.816 : Standard Deviation: 0.004797
10:36:09.816 : echo:endstops hit: Z:12.29

Thanks again for sharing this nice feature with the 3D community,

Rgds, Arno

[Roxy]

Here is a link for the mini proximity sensor I use: http://jimou.en.alibaba.com/product/...ty_sensor.html Type 10-30V NPN N.O. with a switching distance of 0.8mm

I also noticed (especially when using mechanical switches) that slowing down the second and final probe movement increases the accuracy significantly. I therefor patched the ‘homing_feedrate[Z_AXIS] ‘ lines within marlin_main.cpp to be divided by 8 where the default is only 2. This makes a notable difference. Even for a non-mechanical sensor such as an induction proximity sensor.

Thanks for the hint!!! I'm going to set my homing_feedrate[Z_AXIS] to divide by 8 also!!! And once its benefit is confirmed, I'll change the directions in the first post of this thread.