@Julien hmm interesting - i must be honest when i saw that i was like ah ha! i know what it is, i thought you had feed optimisation on (slows down feed rate on corners by a factor you set) but looking at the file you sent, we can say the man form Wales say ānope feed optimisation is offā! going to sort kids out then pop up to workshop later to try this if i have enough stock!
ok so half way through typing this i wonder if you could do the inverse of this and instead of saying slow down with feed optimisation, increase it Julien to see if it still slows down?
Iāve cut quite a few patterned test sheets of black .5" cast acrylic. A honeycomb pattern for instance, every hexagon having a different cam adjustment. I have yet to achieve a reliable and acceptable off the machine finish with my Shapeoko for my purposes. Each of the 6 sides will have a slightly different finish. I get the same segmented finish on curves, which by itself, is generally acceptable to me. But being I have to post process every piece, requires addional hand machining. Iāve played extensively with resolution in Fusion cam, but nothing outside of that. For the most part, full DOC finish passes have proven difficult because a sufficient WOC seems to load the machine a bit too much.
For profile cuts, I was modeling addional geometry and then clearing before finish profile cut. But Iāve given this up for a profile cut with finish passes and a lead out. I contract out most of my work, sometimes with my own cam, and the results are not comparable. The finish from an industrial machine with a .375"+ cutter is rather substantial upgrade. The same or better than that produced by a quality router table, shaper, or jointer. I am also no where near proficient at these things, so take my observations for what they are. Iāve read comments on here about being able to achieve the same finish with Shapeoko, as say with a handheld router. While I believe this is likely possible, the effort per project hasnāt yet proven equitable to me.
I am very interested in thread, and look forward to hearing what others have encountered and suggest.
Reduce it to 0.002 (GRBL default by the setting guide)
The linear interpolation can be considered as a sag calculation (the sagitta formula is s=r-\sqrt{r^2-l^2}, where r is the arc radius, l is the length of the half-chord, and s is the sagitta).
Example: if the radius is 25mm and the sag is 0.02mm, the flats are 2mm long (l=1.0).
r=25, sag=0.005, then the flats are 1mm (l=0.5mm).
r=25mm, sag=0.00125, the flats are 0.5mm long.
At r=12.5, the 0.01mm sag gives flats 1mm long (l=0.5mm)
Iāll be very interested to see if the machine starts to slow down, jerk & stutter with this value reduced as the CPU in the controller is a little lacking in grunt.
@enl_public: very interesting bit of info, thank you. I had my hopes up! butā¦I adjusted $12 to 0.002, reran the cut (video for @LiamN: no visible/audible stuttering going on)
and the corner is still segmented, with the same segment length as beforeā¦
Is there anyhing that would make GRBL not taking my $12 into account ? To be sure, I power cycled the controller after adjusting $12.
And now of course I have to do the opposite test, putting a very large value in $12 to check if thereās any effect (i.e. that it IS actually taken into account)
itās a succession of very small G1 linear moves and very short G3 arc moves. I guess this is why $12 is irrelevant? (i.e. the arc length is so small that it does not even get to break it down into multiple segments?)
I find the segmented radius attractive. A digital and/or mechanical artifact. When I can, I resign from media blasting pockets and the like too, I enjoy the fingerprint. I can make something organic with my hands.
@LiamN : I still think this is unlikely, otherwise people would not be able to get smooth curves in e.g. metal, and they do. Something is off with my CAM or machine or both. Weāll know soon enough when Neil or Jon confirm what they see in their cuts. I am probably in for a facepalm moment when this is over.
@Lowbrowroyalty: it does have an interesting look, but to be honest at this point Iām more interested in understanding/finding why this happens than in the resulting piece.
I recorded an air cut, and one can hear/see the vibration/slightly jerky movements when doing the curves:
EDIT: aaaaand I realized I still had my ātoleranceā at 0.0025mm, which is the reason for the many short moves & arc. I just did the test with another extreme, 0.5mm tolerance, the file has much fewer arcs and they are larger (of course), and the air cut sounds much smoother.
Iāll try a cut tomorrow with a very large tolerance like that to see the effect on the finish, for the sake of testing.
It is irrelevant because the G1 moves make straight segments. The lengths look to be about 0.2mm. The arcs are much, much smaller, and likely at the 0.02mm broke to only a couple linear, but at 0.002 would have broken into more, but the arcs are so small in angle as to make it pretty much not meaningful.
I see that boosting the tolerance helped. Good deal.
I donāt know why Fusion would put the straight motions in between the arcs. It is not hard to fit the arcs without them for most ways of defining the curve.
I will have to look at this tomorrow evening as i need to break a project to get acrylic lol see what i do for science!
Also check this out, i was looking at a Fusion 360 vid - more training again :), but this part i think might be worth doing on your model Julien, it may make no difference but as he says this then confirms to G1 and G2 - interesting, it may be applied when you did the normal spline but worth a try, Fusion 360 constraints explained with examples fantastic vid would recommend to watch from start, but bookmarked at right location
itād not be insane hard to make a small tool that, say, takes a circle diameter, depth of cut and number of segments, and turns that into some gcode
ā¦
should make it sort of easy to see if this is digital or mechanical
Have you compared the pitch of the belts to the the āpitchā of the lines in the plastic? Your belts could be stretched and not meshing with your sprockets smoothly but sort of snapping into the sprocket as it turns.
Cheers
Mike
Acrylic is unforgiving.
No matter what I try, I get those marks. I did a redesign of @Julienās spline design and just used tangent arcs and lines. Same thing.
@fenrus, Iām sure by the time Iām writing this response, you have already implemented this ?
@JappieMike the pitch of the belts is 2mm, they may be stretched by a few percent,so not a really good match to the ~0.5 to 1mm periodicity I see, also it would very likely show in the āX onlyā and āY onlyā moves too, and it doesnāt. That said, I had in my list a test where I would purposefully untighten my belts to see if it changes anything.
@neilferreri thanks for the test cuts, appreciated ! Iām a little relieved that itās not only me, but even more curious as to what is going on there. Does your G-code has the succession of very short G1 segments and G2/G3 moves too? Iāll redo a test cut with a pure circle later today (basically a single G2 command) and see if it changes anything.
No. As soon as I saw that with yours I thought, āwell, thatās the problemā. So I tested and I still got the same edges.
Iām horrible at documenting (not good for this sort of thing), but I tried a number of different settings with tolerance, both in the CAM settings and in the user options when post processing. I changed the $12 values.
It was actually pretty frustrating. I could only seem to make it worse.
I think I saw signs of what I could only guess was from deflection.
Start fundraising for that 600fps cameraā¦ Iāll chip in!
This is why I really want to see a physics-based simulation which actually takes into account flute geometry ā could this be some sort of interaction between how quickly the endmill is turning and how quickly the machine is moving the endmill against the surface?
