Easy backlash calibration (experimental)

thingiverse

I have a couple of Dollo3D printers that I've been building and developing for a couple of years. I've used both of the quite a lot for printing mostly functional parts that aren't too tight on tolerances. But recently I started noticing some inconsistencies, especially on circles and started exploring backlash compensation. Of course to combat backlash, I need to know how much backlash there is so I took a look on thingiverse's backlash related offerings. Tried few but as usual, a design of my own had already started to form on the back of my mind. Bear in mind this is still a bit of an experiment, I've tested this on only one printer which might have other factors skewing the results. Results: on the one printer I've tested this on the result's were 0.1 on X and 0.25 on Y, which seems to yield good enough results for that particular machine. So my backlash measurement involves two models, pro-backlash and anto-backlash. Bot are basic cube, with either added or removed pieces on each side. - pro-backlash is a model that will have backlash when changing print head direction on cube corner. The cubic indents are there the ensure that when the head changes direction from X to Y (and vise versa) the previous direction on new axis has been opposite, i.e. the direction change will need to deal with backlash - anti-backlash is the opposite of pro-backlash, the cubic extrusion make sure that the previous direction on new axis has been the same as the new direction, i.e. backlash has already been accounted for So how to use this? For most accurate result I suggest printing the pro_backlash and anti_backlash models separately WITHOUT backlash enabled. To save time, print them in same print which should work also. Measure both model's X and Y sides (they are marked on the models) and compare the results: pro_backlash should have less length than the anti_backlash, the difference is the backlash compensation needed. For best results measure on both edges of each axis, to account for travel moves and seam that might affect the backlash. Extra verification step: print the models with the backlash valued applied. If my reasoning is correct, there shouldn't be any differences in length of axes between the models. Do let me know if this works or not. My initial solution was a Python script to generate print movements, but it turned out too cumbersome and probably wasn't as accurate so I'm hoping this would serve as a quick and reliable method of backlash measurement. OpenSCAD source: https://github.com/spegelius/3DModels/blob/master/Calibration/backlash.scad