Coming back to this - I don't believe it's my arm length since I got my arms from Ultibots before they closed and I used the arm length written on the arms.
Since this post I've checked on my carriage tensions and belt tensions - one carriage was tensioned a bit wrong and I've got all my belts at about 30hz.
My cubes are a lot more…cubical now when you just check the raw side lengths. Unfortunately, they still have non-square sides. I've attached pics of the latest cube.
My printer is a delta running RRF 2.02(RTOS) (2018-12-24b1) with DWC 1.22.6 and WiFi Server 1.22 running on Duet2 WiFi. My auto-calibration uses six factors and 12 probe points; config.g and bed.g attached. I have ball-cup CF arms with AL ball joints on the effector; the carriages have magball ends instead of single-piece ball joints.
I've been trying to sort out dimensionality/accuracy issues on my delta. 20x20x20 cubes are printing at about 21x20x20mm. A 50x50x5 print was 51.47 x 49.86 mm (measured each side 3x and averaged). Also, one corner is visibly non-square. The short side of the shape is always oriented towards the Z column. Prints are sliced with KISSslicer and the Gcode looks fine in Gcode.ws. There's some odd gapping in the top layer that may be related.
I'm trying to figure out what to look at next; I found this issue after I did a number of printer upgrades (Delrin to polycarbonate wheels, new belts, new bed; carriages unchanged). Belts are tensioned to about 30hz (measured as well as possible with EasyTension).
13, 1, 7, 2, 15
I'd love to see support for the Prometheus-style multi-filament hot ends as well. Both Prometheus and Prusa have made them. They retract filament when it's not in use and then advance the alternate filament. I think the only big change would be to filament movement distances.