Almost there, waiting on a z-axis motor but in the meantime, I tried the calibration benchmark and left the bit down cutting all throughout the benchmark to expose if my straight lines are actually straight and I’m seeing some sink in the middle. I’m 10mm lower in the middle than on the sides.
My sled weighs exactly 20lbs. It used to weigh 30lbs and I noticed the sink on my first calibration(score 2.8 - 0.13) so I lost some weight thinking that was it but the cut is almost the same second time around(score 2 - 0.38).
What could this be?
I have configured my machine manually using the “Manual Calibration” method, however I noticed at the end of the regular “Calibrate” there was a section to cut some lines and tell Ground Control the actual measurements of those cuts so GC could have some feedback but I don’t see this after manual calibration. Could this be what I’m missing?
I went through the entire calibration process without the automatic z-axis and I it did improve accuracy.
By the way I didn’t find the “Simulation” part of GC until now! Had I seen it earlier I might have overcome errors quicker but that’s on me! I’m definitely replacing my 2x4 with a metal uni-strut after seeing GC bend the 2x4 during the “pull chain tight to measure”. I believe this is where the sinking in the top middle is coming from.
I noticed sag in the y values especially near the top of the machine while working on optical calibration. I measured the motor distance with a tape measure (instead of using the pull-tight-to-measure routine) and it improved.
So life is good until I noticed my Maslow is still not calibrated. Comparing measurements of a piece I cut, to the measurements of it’s drawing let me know that I’ve got some error. If I attempt to cut 8" square, my vertical is 8 1/16" and my horizontal is 8".
To limit variables I’m looking at my sled and I’m thinking @dlang Pantograph kit is looking pretty good.
Felt I needed to update this thread to try to be consistent on what actually works and what might not.
I don’t mind it except I was attempting a chair for my office desk through sketchup using a slicer plugin. This requires X and Y accuracy if you’re trying to get it all on one 4x8 sheet with 6" buffer for Maslow then some of the cuts end up with different orientations. Also the sliced pieces will need to go back together.
Also afraid I might see my error in other ways after I’m done cutting.
I personally find pleasure in organizing parts by hand to make the most of the material… but then again I might be weird. It’s like a puzzle that doesn’t have a specific solution, but can be optimized!
(It turns out that human brains are exceptionally good at this type of problem and computers are surprisingly bad at it. I spoke with a senior engineer at Cisco Systems about it once, he mentioned that since a computer needs to actually iterate through all possible situations it becomes computationally prohibitive to actually find the “best” option because the actual options are effectively infinite. There are programs that will nest parts and they do a pretty good job but their accuracy is dulled a little because asking a computer to calculate an infinite number of solutions is mean (and slow))
If you need a linkage right away I’m shipping same or next day right now: linkage link
I can say that it’s within the range that many people have experienced, including myself. The ultimate goal (as I understand it) is 0.5 mm (1/64-inch) accuracy across the entire sheet of plywood. I don’t know if anyone is achieving that, particularly in the bottom corners. But to put your question into perspective, if a cut is 1/16-inch too long, it could in be that the start point was 1/32-inch too high and the stop point 1/32-inch too low. If that’s the case, that’s pretty darn close to the 1/64-inch goal. Of course, it could be that the start point was 1/16-inch too high and the stop point was dead on accurate, or vice versa.
If the real question is whether or not you can improve upon it, that really depends upon how accurate you built your machine and how accurately the model used by the controller reflects the real world. See this post relating to sources of errors:
There’s lots of room for error and I suspect they all need to be eliminated to achieve the ultimate accuracy goal. I don’t know if that is universally feasible with a hand-assembled system unless we can find a way to compensate for “less-than-perfect” construction.