Laying out my longest engraved scale on the honeycomb:
The zero-th step aligns the scale with the axis travel: slide one end of the scale to put the dot on the edge, jog to the other end, slide to put the dot on the edge, iterate until the dot is the same brightness on both ends.
The scale lines are a tidy 0.2 mm wide, the red laser dot might be 0.4 (it’s rectangular-ish), and a jog increment of 0.2 mm works well. I can manually align (pronounced “slide”) the scale on the honeycomb to center the dot within a line, whereupon moving the head a known distance to the other end of the scale and counting-while-jogging a few steps until the dot drops into the proper line gives the offset from the correct distance.
Jogging 590 mm along the X axis produced 589.8 mm of actual travel (one jog step short of the line 590 mm from the start), an error of -340 ppm.
Jogging 495 mm along the Y axis travels 494.4 mm, an error of 1212 ppm. That’s considerably more than I expected and required a few iterations until I believed it.
Both axes use steppers with 20 tooth pulleys and 3 mm pitch belts, so the laser head moves 60 mm per motor revolution. The stepper drivers are configured for 5000 steps/rev, so the axes should have a step length of 12 µm = 60 mm / 5000 step. Both axes arrived with Step Length values set to weird numbers very close to 12 µm, but, after a quick check showed incorrect travel distances, I reset them to 12 µm before making real measurements.
LightBurn provides access to the Ruida controller’s “Vendor Settings” (after a warning to not mess things up) and allows you to change them:
The values shown above come from multiplying 12 µm by the ratio of the actual to the intended distance:
- 11.9959 = 12 × 589.8 / 590
- 11.9855 = 12 × 494.4 / 495
Repeating the tests with those slightly smaller step sizes produces motions that are spot on to within my ability to measure them.
Neither of those changes was large enough to affect the outcome of cutting the Tek Circuit Computer decks, which are much smaller than the full extent of the axes and thus see much smaller errors.