I want to share my results and compare with others to work toward getting the highest accuracy possible. –
V1.03 is giving the best results yet!
My sled weighs about 30 lbs.
Using Logan’s linkage – Entered 270mm for initial RR, so calibration cuts are located near the center of the sheet, vertically speaking
After calibration ran the test pattern – Horizontal distance = 1905mm top, 1905 mid, & 1904 bottom – Yay!
Vertical distances = 898mm left, 895.5 mid, & 897.5 right
Small squares – I think I’m supposed to measure them from 1 side of a cut to the same side opposite, not the size of the square itself? They are all around 100mm (+ or - 1mm off)
Measured diagonals from outside corners of bottom left square to outside corner of upper right square, and vice versa – Matched exactly – Yay! No parallelogram.
Trying to use the calibration benchmark score proposed by “blurfl”, I get 1.7 for the long measurments – The short ones, since there are 2 dimensions to each of the 5 squares, my total absolute error is 3, and divide by 10 gives me .3 – So, 1.7 - .3 – Pretty good I think
Still would like to tighten up on the vertical distances –
– Philip
That’s great news! Does your frame have a top bar? @Bar gets the credit for inventing ‘the score’, yours score is very good.
I had similar results (3.44 : 0.73) using a top-mounted linkage on a top bar frame 11 feet between motors with a 26 pound sled. My score not as good as yours, though! I didn’t change the RR, it was 100 if I remember, but the calibration corrected that to 128.1mm. Edit - chainSag 35.75.
I hope others will post their results as well.
can you guys run the calibration again and see if you get different numbers?
at one point we were seeing that the final numbers varied depending on the
initial values, so running it a couple times got closer to perfect.
128.1 for the top mount kit is a bit short, which could be a calibration error,
or could be an error in measuring the motor spacing (which will throw off the
rest of the calibration)
That looks more like what we would expect to see, with very little variation in parameters between calibration runs.
@dlang, I’m guessing the source of the variation in the measurements we were seeing before was because the Gcode was cutting the lower marks about 2 inches below what the algorithm expected. This lead to a variable chain sag for each subsequent calibration process, as the simulated and actual cut locations were different. But hopefully we’ll get a few more people to repeat this test to confirm.
[quote=“rjon17469, post:2, topic:1987”]
What were your final rotationRadius and chainSagCorrection values after calibration?
Is your frame stock? Does it have a top beam? /quote]
RR = 272.7mm (with Logan’s linkage)
Chain sag = 15.271mm
We have a top bar with an L-shaped cross section – A piece of Poplar about 1-1/2" x 3" with a piece 3/4" x 4" screwed to the edge, for stiffness in 2 planes.
Questions –
Am I right that, for the accuracy score, the total errors on the 5 small squares should be divided by 10?
Am I right that the small squares widths should be measured from the right side of the left vertical cut to the right side of the right vertical cut? (or vice versa)
To run the calibration a 2nd or 3rd time, how far back in the procedure should I go, before starting again?
Thanks – Philip
Am I right that, for the accuracy score, the total errors on the 5 small squares should be divided by 10?
yes
Am I right that the small squares widths should be measured from the right side of the left vertical cut to the right side of the right vertical cut? (or vice versa)
yes
To run the calibration a 2nd or 3rd time, how far back in the procedure should I go, before starting again?
ideally back to the beginning, but the motor distance probably doesn’t change.
We are looking to see if there is a difference between different calibrations
depending on what the rotation radius was set to before you started doing the
cut for the calibration.
I don’t think you need to recalibrate the motors or the chains, just run Step10. Everything before that point is based on counting encoder pulses, which won’t change between runs unless the frame is altered.
OK, I think we’ve got it dialed in! 3rd calibration run got us to a benchmark score of 1.0 - .227 – RR = 272.9 and Chain sag = 22.34
Seemed to help when we increased the sled weight (before this series with V1.03) from 20lbs. to 30 – Also helped eliminate chatter on the cut edges – I think we’ll try going up to 35 lbs., or maybe even 40.
The proof for us was to cut out a large shallow arc piece – a template for a piece of arched casing molding to go over an arch top door frame – We make a template, then use that to create the same size pieces in wood – So we compared the arch template (1/4" MDF x 3-1/2" wide x 42" long with an outside radius of 40") with an actual size drawing, created with CAD and printed on a Designjet plotter – Because the arch casing has to match the curve of the door jamb, with a 3/16" or 1/4" reveal, it has to be pretty exact – The first few we tried were about 3/16" to 1/8" off (lining up the lower corners and measuring the error at the top center of the arch – But the error on our latest attempt was only about 1/32" – Entirely acceptable!
I’m not sure why, but the whole template was a bit oversize – We could easily fix that by setting the bit diameter a little undersize.
It’s not about the accuracy so much – I like the extra weight for the dampening effect I think it has on the vibration caused by the cutting force – The force of the bit cutting through the material must feed back to the sled and contribute to the chatter we were seeing on the edges of our work pieces.
Bar – I will put up some more pictures – But basically we’re still using the first frame, but stiffened up with L-shaped cross pieces near the motors and just below the top edge of the plywood backing sheet.
