I designed and fabricated a weird 3D printed 4-axis CNC machine (in ~3 months, oof) which uses carbide inserts to carve tiny isolation routes in chunks of PCB substrate. It's very much a finicky proof-of-concept, and may very well host fatal hidden gremlins which doom the project to novelty status, but with a bit of care I can produce boards with 6/6 design rules (0.15 mm spaces and traces) at 20-30 mm/s, roughly an order of magnitude faster than a desktop mill with significantly less noise.<p>I gave a talk this weekend on the machine at the eighth Hackaday Supercon, which will be on YouTube at some point, but for now here is a link to the project page, including design files and a few dozen hasty project logs.
After years of 3D printing, a technology I love, I thought I'd try CNC, and make a PCB.<p>Oh boy did was it ever a hassle building the machine then trying to sell everything on eBay.<p>I am super impressed by people who actually put in the effort to make this stuff work.<p>The biggest issue is the vias. They have to be done by hand or with more expensive machinery. Then there's solder mask. Then there's cutting the board outline, it's nasty fiberglass.<p>Once it's all done, you have to actually solder the boards, which is even harder than it would be with a commercial board.<p>I have a hard time imagining wanting to a DIY board or even hand populating a commercially made board again.<p>It's so much fun to be able to use accelerometers and USB-C and modern power management chips in DIY projects.
IMO fully mechanically milled PCBs are kind of stupid; you are never going to actually manufacture them at scale this way, so what is the point in jumping through the DFM effort to prototype with such a constrained process? There used to be a cost and speed advantage vs using a service provider, but pcbway and jlcpcb have absolutely changed the game on that front. There still are advantage for in-house manufacturing capability, but I don't think pure milling is a good approach - whatever small scale process you use for this needs to produce an end product that is highly similar to the mass manufactured process and as such will require mechanical (milling/drilling), optical (etch/ablate), and printing (mask/silkscreen) processes. I think there are some small machines getting quite close to automating the one-off production of 'real' pcbs, though.