This can be more technically termed a physical 'steady-state offset' in the way the extruder head, or any robotic part experiencing friction or loose parts for that matter, comes to rest after a movement.
|Early indication: Note the ridges out of line on the near-left face; structurally insignificant here, but quite visible.|
So I set about trying to fix this with some tests in PLA. I created a simple test-print of a 25x25mm square and 25mm diameter circle wall next to each other so that I could measure them with calipers and see how the real object compared to the design.
Since this problem can sometimes be caused by a pulley working itself loose from a motor shaft, I first tried tightening up the grub-screw holding the x-belt pulley in place. This made no observable difference so I turned my attention to the belt.
While the next test print was running, I observed how the belt fed over the pulley as the perimeters were drawn at low speed, in case there was any damage to the belt. What I noticed was that at a certain point along the belt as the extruder turned around the circle, the belt would suddenly slip across the pulley by almost half a tooth-length as it wasn't meshing properly. At the exact same time, the extruder jolted to one side by a tiny amount, causing the line drawn to not weld properly with adjacent layers.
The resulting malformation can be seen here:
|Also notice how squashed the original 'square' is in the X-axis.|
My next remedy for this was to take the x-belt out and swap it around, since the pulley teeth only ever saw half of the belt during its full travel, I guessed that using the un-worn half of the belt would allow it to grip better. I also noticed while taking the belt out that there was some very slight wear on the ABS X-axis pulley, so I relpaced that with a spare.
This helped for a while, resulting in the print improvement that you see above, although the problem started to creep back in recently, so I suspect that the belt is nearing the end of its service life and may have to be replaced before it snaps.
In the interest of reducing this problem in future, and also getting rid of the annoying round-four-corners path followed by the Mendel X-Axis, even with its main alternative design to decrease belt wear that I happen to use, I have been designing new end brackets for the X-axis, with the belt travelling in a similar simplified style to the way Josef Průša completely redesigned it, however this will still be compatible with the old Z-axis fittings of the original Mendel. This should shorten the belt needed, thereby removing some of the flexibility, while simplifying the construction of the axis slightly.
|The first prototype, bottom-right part had issues due to the extruder gradually losing grip during print.|
The parts for this redesign are here, and tomorrow I should be fitting and testing this system, so I can update how this goes then. If all goes well, then hopefully I can build a better extruder with the steel fastenings I recently got, and complete another set of parts to sell so that I can replace the belts before they snap. O_O