I found an interesting script for Eagle that can be used to help calibrate your CNC PCB milling settings:
I exported some GCode to go run on the mill. --I am running Eagle/PCBGCode on a Mac, PCB-GCode-Wizard (when needed) on an XP VM, and then LinuxCNC on Ubuntu.
I spent a few hours moving between the office and shop this afternoon tweaking Eagle/PCBGcode.ulp and my mill. I learned a few things so I will consider the Saturday afternoon a success.
It appears that a 0.2mm 60 degree vbit will cut an isolation trace about 12mil or 0.012" wide. This is pretty consistent with the best of the tests that I ran a week or two ago. With a SHARP bit I was able to leave behind a 1mil or 0.001" electrical trace. (Yeah, that surprised me too.) I have pattern of traces from 1-25mil that I used for the test scenarios.
Gotchya #1: It requires an ubber sharp bit. Asthe bit wears down or dulls that is a largely unachievable result. The reality is that I would not dare working smaller than 10-15mil electrical traces. 20mil will probably be my norm.
Gotchya #2: With the 0.2mm 60 degree vbit my minimum isolation is ~12mil. This going to limit my SMD parts as long as I use that mill bit.
I have the following bits on order:
45 degree - 0.2mm
30 degree - 0.2mm
20 degree - 0.1mm (probably too fragile to use)
End mill - 0.2mm
0.3mm = 11.8mil
0.2mm = 7.87mil
0.1mm = 3.9mil
Reality #1: My guess is that ~10mil of isolation might be my realistic limit. I should have the next round of mill bits here (with luck) by the end of this coming week. I will need to do some research on SMD package sizes to see how that matches my isolation realities.
Reality#2: For homebrew work this is probably will probably more than cover my needs for a while. Currently I am not working in the crazy small SMD components very often so it is probably not a major limitation. If I need something crazy small it would probably make more sense to send the board out to a fab house or to get an adapter board for prototyping work.
Next steps: Document my updated settings so that I can reproduce the results again in the future. Once the new bits are tested my next major testing task will be pin registration so that I can flip the board and mill both sides. Being able to prototype 2 sided boards would let me utilize auto-routers much more effectively and to have more control over board layout and size.
Here is a great link:
That site has some create info that closely aligns with my experiences during my initial testing cycles.
73 de NG0R