CNC PCB milling essentials for hardware engineers

• 1
  How CNC PCB milling removes copper
• 2
  From Gerber files to a cutting path
• 3
  Substrates that behave and substrates that fight back
• 4
  Where CNC PCB milling stops making sense
• 5
  CNC PCB milling vs chemical etching

A CNC router holds a single-flute or two-flute end mill and drives it along a tool path generated from your Gerber and Excellon files. The cutter spins between 20,000 and 60,000 rpm and plunges a few hundredths of a millimeter into the copper foil, shearing it away from the substrate underneath. What remains is the trace pattern. No mask, no etchant, no photoresist.

Depth control is the whole game. Copper on standard FR4 is 35 μm (1 oz) or 70 μm (2 oz). A 30° V-bit cutting a 0.2 mm isolation channel may only reach 50–80 μm below the surface, so the machine has to know where the surface actually is. If the bed or the laminate is not level, one side of the board gets a shallow cut and the other side cuts into the glass weave.

The cutting action itself is abrasive, not smooth. FR4 is epoxy bonded to woven glass fiber, so the tool alternates between soft resin and hard glass. Carbide tooling holds up; high-speed steel does not. Expect to change cutters every few boards, sometimes every board once you get below 0.2 mm trace and gap.

Vacuum hold-down matters more than most people expect. A 100 × 100 mm board that lifts 0.05 mm during a cut will ruin fine-pitch work. Tape alone is not enough for boards under 0.3 mm thick. We use vacuum chucks and dedicated fixtures, and we surface the spoilboard before each job so the reference plane is true.

The CAM step converts copper layers, drill data, and board outline into vector paths. Isolation routing traces around every copper feature. Rubout clears large copper planes with overlapping passes. Contour routing cuts the board free, and the drill file becomes a set of peck cycles.

Trace and gap size drives everything else. A 0.2 mm tool needs roughly 0.25 mm of clearance to pass between two traces without rubbing the sidewalls. If your design has 0.15 mm gaps, either reduce the tool diameter or accept that the isolation pass will not fully clear the gap. Many shops will tell you it worked. It usually did not.

Tool diameter sets the copper width you can hold. A 0.8 mm end mill cuts a 0.8 mm channel, which is wider than most people want between fine traces. Most boards in the 0.5–2.0 mm pitch range run best with a 0.2–0.4 mm carbide cutter and a 30° or 45° V-bit for the isolation pass.

The Excellon drill file is separate from the routing. Hole sizes must match the physical drill bits you have. A 0.9 mm finished hole becomes a 0.9 mm drill plus plating allowance, but milled boards usually have no plating, so the drill size equals the finished hole. Confirm this with your vendor before setting up the file.

FR4 is the default. It machines cleanly at moderate feeds, holds ±0.005 mm on hole position, and costs little. The tradeoff is edge quality: the glass weave frays slightly, so a light sanding pass or a 0.2 mm finishing cut helps. For RF work above 3 GHz, FR4 loss becomes the problem, not the cut.

Rogers laminates (RO4003C, RO4350B) cut well but generate fine dust that clogs vacuum filters. Their ceramic-filled PTFE is more abrasive than FR4, so tool life drops maybe 30–50%. The upside is stable dielectric constant and low loss, which matters for antennas and high-speed differential pairs.

Aluminum-backed boards and thin aluminum sheets machine differently. The copper layer cuts like copper, but the aluminum backing grabs the tool if the feed is too high. Plan on lighter chiploads and a dedicated aluminum cutter. This is common in LED and power electronics work.

Flexible substrates like polyimide are the hardest to hold. They deflect under cutting force, so the tool skips or tears. Vacuum hold-down with a sacrificial backing plate helps. Very thin flex, under 0.1 mm, is usually better sent to a flexible circuit shop than milled in-house.

The process excels at single-sided and simple double-sided boards with 0.2 mm and larger features. It struggles below 0.15 mm trace and gap, where tool deflection and surface flatness dominate the result. If your design has 0.1 mm lines, choose a fab house with a photo-lithography process.

Plated through-holes are not part of the milling workflow. You can drill a hole and press a rivet or a wire, and you can solder a via from both sides. You cannot get a plated barrel without plating chemistry. For boards with dense via arrays, that limitation alone decides the method.

Solder mask and silkscreen are optional and crude. Milling leaves bare copper, which oxidizes in days unless you tin plate, apply a conformal coat, or use an OSP finish. Some shops screen a mask after milling. Most do not. For a functional prototype, bare copper with a quick tinning pass is often enough.

Board size matters too. A 4,000 mm × 400 mm travel envelope covers almost any prototype panel, but a large thin board will chatter at the edges. If flatness across a 300 mm span is critical, plan on a fixture that supports the middle, not just the perimeter.