With the difficulty of getting super-precise measurements for a successful calibration, I thought we could start sharing some ideas on how to improve this critical process.
1 Like
One method I think would work is, first, to complete calibration just like it’s done today (manual measurements, followed by the square test cuts). This is the base calibration.
Now, on to the refinement…
Next, Maslow would be told to cut two long, slender rods. The first one would be cut in the horizontal direction, and would be supposedly the exact same length as the distance between the motor mounts, with a couple of markers (e.g. small arrowheads cut into the rod) at exactly at the 1/3 mark, center and 2/3 mark on the rod. The user would then pick up this rod and physically place it between the motor mounts to verify that the cut and the actual distance are identical. The markers would be used for further verification against manual marks on the frame that the user makes with a pencil. If all is well, everything should align perfectly with no gaps and no shifts.
If there are shifts or gaps…then the calibration constants that were entered into Maslow need to be modified (I still have to think about how each type of gap or shift translates into what type of correction needed for the constants).
The process above would also be repeated with a vertically cut rod, that should fit perfectly between one of the motor mounts and the router bit, when the sled is parked at an exact calibration position.
1 Like
One method I think would work is, first, to complete calibration just like it’s done today (manual measurements, followed by the square test cuts). This is the base calibration.
The problem is that the current calibration is only cutting in one area, and it
assumes there is only one variable to tweak.
The simulation has shown that different errors can produce similar results. So
the whole approach is questionable.
…if this is true, then the test pattern selected is not the best choice. The test cuts should be designed to have a clear mapping between the Maslow variables and measurable errors in the test cut.
Easier said than done, I know.
Take a look at this post on cnczone.com:
http://www.cnczone.com/forums/diy-cnc-router-table-machines/166437-cnc.html
…and this test pattern with explanation, posted on inventables.com
1 Like
…if this is true, then the test pattern selected is not the best choice. The test cuts should be designed to have a clear mapping between the Maslow variables and measurable errors in the test cut.
play around with the errors in the simulator, defining such a pattern is
extremely hard.
Take a look at this post on cnczone.com:
unfortunantly, this is designed to catch errors on Cartesian designs, our design
ends up with very different types of errors.
1 Like
What about using a circle of numbered dots?
And a dot in the centre.
Then the distance between ots and the centere dot should all be the same
the dots that can form a square should all have the same distance
the dots that can form an equilateral triangle should also have the same distance.
eventually a second smaller circle of dots inside that circle, half the radius, to see if there are distortions?
not sure if this is needed…
1 Like
Calibration patterns need three major features
-
the pattern changes if the measurements that are entered are incorrect
-
it’s easy for the builder to measure the results
-
based on what’s measured, we can calculate how to change the entered
measurements to fix the errors.
I don’t think the circles approach satisfies these.
1 Like
O that was a quick reply, meanwile i was playing with the idea in inkscape and i’ll share my brainwaves on this.
In a way there is only one measurement that counts
The width between every dot to the nearest dot should be exactly the same distance… When there is distortion in any direction then that should be clear immediately.
Very easy to measure. More easy then square since the equilateral triangle allows to measure the roundness.
collect measurements of all triangles and do the math from there?
I also see an option to make use of 6 bigger triangles in the same configuration. Longer distances / less measuring errors… double the detail…
EDIT: uploaded new drawing showing the larger triangles
3 Likes