I set up the CAM for 4 operations (settings shown are the ones I converged on after 2 failures)
- Adaptive clear the interior
- 1mm depth per pass
- 250mm/min feed
- 0.2mm stock left radially
- 2D contour the G
- 0.5mm roughing depth
- 120mm/min feed
- 2D contour the inner part of the outer ring
- 0.5mm roughing depth
- 120mm/min feed
- 2D contour the outer part of the outer ring
- 0.5mm roughing depth
- 120mm/min feed
I used a 1/8" 2 flute end mill at 2500rpm.
The first attempt, I broke an end mill entering the first 2D contour. I initially add a 1mm depth per pass on the roughing and it dug in and broke right away. I measured after the cut and it was actually 1.5mm deep! Z must have slipped, and admittedly I had "0" set a bit low, in that the first move across the part cut a channel about 0.2mm deep.
Second attempt I dropped the roughing depth to 0.5mm and set zero on the Z axis by setting the end mill on a razor blade on top of the part. This left it slightly shallow but I wanted to be careful. This worked much better, but I finally did break and end mill on the final contour. This was caused by 2 factors. One, once again the end mill dug in, secondly the part was not perfectly flat. Thirdly, perhaps somewhat contributing, is that I didn't have the center of the cut aligned so the outer 2D contour left the stock and re-entered. I think the combo of unloading, reloading, and having too deep a cut did me in. The cut was 1.mm on that side and 0.8m on the other. You think you'd see that but I swear the stock was flat! Combo of using a diced-up part as a shim and getting swarf between the milling vise, shim and stock.
Anyways the Z digging was an obvious common culprit that needed to be addressed. When I have a problem I can't figure out I generally do one of two things. The first? charge ahead and just start doing stuff and generally screw up a lot and get frusturated. The other, rarely, walk away do something else and then find a eureka moment. Fortunately this was one of the rare times - I had my commute bicycle partially torn apart to fix something, so I worked on that awhile and eureka. I was thinking about the GRBL settings and recalled that the X and Y have 126 steps/mm or something like that and the Z has a mere 25 steps per mm because I'm using an 8mm lead screw with direct drive. So it only takes a few missed steps to make a notable impact in the Z whereas in the X and Y I can lose a few and never know it. So, what if I enabled microstepping? A 1/4 step would give me 100 steps/mm and would be the same order of magnitude as the other axes. Sure, it masks a problem, maybe, but it 'scales' the problem to be the same order of magnitude as the other axes.
Third try - success! Code ran to completion, no broken end mills, and the dimensions were good to within 0.1mm or so - again, 10 steps, but margin of error, that's 4 thou which is pretty damn good if you think about it, for a mini mill with printed parts!
I did get some wicked backlash in the Y axis - the screws came loose on the affixing print. I need to add some threadlocker.