CNC routing

A CNC router is a spinning cutter that a computer moves through a workpiece along a path you specified. That is the whole idea, and everything else is consequence: because the path is a file, the part is repeatable; because the machine does not get tired, the tolerance is consistent; and because it follows the file exactly, a mistake in the file is executed at full speed with complete confidence.

The workflow has three stages and they are genuinely separate skills. You draw the part in CAD. You decide how it will be cut in CAM — which cutter, which order, how deep each pass, where the machine holds on — and that produces a file of machine instructions. Then you set the material up on the machine, tell it where the workpiece starts, and run it.

Beginners expect the drawing to be the hard part. It is not. CAM is where the craft is, because it is the stage that decides whether the cut succeeds, how long it takes, and whether the cutter survives — and it is the stage where the machine cannot tell you that you are wrong until it is already cutting.

The three stages

CAD, then CAM, then setupCAM is where the skill and most of the mistakes live.

What it is good at

Repeatable, accurate, awkward shapesA single one-off is often quicker by hand.

The real constraint

RigidityA flexible machine cannot be made accurate by going slower. It is the specification that matters most.

The hidden cost

Dust extractionBudget it with the machine. It is a health requirement, not an accessory.

The workflow is the thing to learn first, and it is best learned on a part that does not matter:

1. 1

   Step 1: Draw something simple in CAD

   A flat part with a couple of holes and an outline. Free options are genuinely capable, and the goal at this stage is to produce a shape rather than to design anything.

   You will know it worked when: A drawing with real dimensions.

2. 2

   Step 2: Set up the toolpaths in CAM

   Choose the cutter, decide the operations and their order, set the depth per pass conservatively, and add tabs so the part does not break free at the end of the cut. This is the stage that decides whether it works.

   You will know it worked when: A toolpath, with a simulation you have watched all the way through.

3. 3

   Step 3: Hold the work down properly

   Clamps, screws into the spoilboard, tape and glue, or a vacuum table. Workholding failure is the commonest cause of a ruined part and one of the commonest causes of injury, because a workpiece that comes loose leaves at speed.

   You will know it worked when: Material that cannot move, with clamps clear of every toolpath.

4. 4

   Step 4: Zero the machine

   Tell it where the workpiece starts in all three axes. Getting Z wrong is how people cut through the spoilboard into the bed, and it is a rite of passage worth skipping.

   You will know it worked when: A known origin you have double-checked.

5. 5

   Step 5: Air-cut it first

   Run the whole program with the cutter raised well above the work. Ten minutes here catches a wrong origin, a toolpath that leaves the material, and a collision with a clamp — all of which are expensive to discover during the real cut.

   You will know it worked when: Confidence that the program does what you think.

6. 6

   Step 6: Cut it, and stay with the machine

   Hand on the stop, extraction running, hearing and eye protection on. Do not leave a cutting machine unattended, and do not reach in for offcuts while it is moving.

   You will know it worked when: A finished part, and a machine you were watching.

7. 7

   Step 7: Clean up, wet or with a vacuum

   Not with a brush or an airline. The clean-up is a health step rather than a tidiness step, for the reasons in the safety section.

   You will know it worked when: A workshop that is not full of airborne dust.

Almost every specification quoted in a listing is secondary to one property: rigidity. A machine that flexes under cutting load produces a cut that wanders, and no amount of slowing down fixes it — it merely makes the wander smaller. Everything below is really a statement about rigidity and the working area it is achieved over.

The specifications worth reading, in order:

• Frame rigidity and gantry design — The one that decides everything. Steel or heavy extrusion, a low gantry, and as little unsupported span as the working area allows.
• Spindle against trim router — A proper spindle is quieter, more powerful, speed-controlled and far more pleasant to work near. A trim router is cheap and is the commonest first upgrade to regret not making.
• Working area, honestly measured — The usable area is smaller than the advertised travel once workholding and the spoilboard are accounted for. Bigger is not free — a larger gantry is a less rigid gantry.
• Leadscrews or ballscrews against belts — Belts are cheaper and stretch under load; screws hold position better. This shows up as accuracy under heavy cuts rather than in an idle machine.
• Controller and software openness — A machine locked to one vendor’s software is a machine you cannot fix or improve. Widely-used open controllers have a community behind them.
• Whether an enclosure is possible — For dust and noise. Retrofitting one to a machine never designed for it is much harder than choosing a machine that anticipated it.

Two people can cut the same drawing from the same material on the same machine and get very different results, and the difference is almost entirely in how the cut was planned. This is the part that repays study and the part no beginner expects to matter.

The decisions a toolpath actually encodes:

• Order of operations — Pockets and holes before the outside profile, always. Cut the outline first and the part is loose while you are still machining features into it.
• Roughing and finishing as separate passes — A heavy pass that leaves a small allowance, then a light pass that takes it off. The finishing pass cuts at a consistent load, which is what produces a consistent surface — one pass trying to do both does neither well.
• Depth per pass against total depth — Three shallow passes load a light machine far less than one deep one, and cost only time. This is the variable to reduce first when anything goes wrong.
• Climb against conventional — Which way the cutter rotates relative to its travel. Climb generally leaves a better finish and pulls the work, which matters more the less rigid the machine and the weaker the workholding.
• Ramping and plunging — Most cutters cut badly straight down. Ramping or helical entry lets the tool enter at an angle, which is easier on the cutter and on the machine.
• Tabs, and where to put them — On straight edges rather than corners, and away from surfaces that matter, because you have to remove them by hand afterwards.
• Where the clamps are — A toolpath that is correct and passes through a clamp is a broken cutter. Simulating the whole program is what catches it.

The habit that separates people who improve from people who keep breaking cutters is watching the simulation all the way through before every cut. It costs a minute, it is the only place a mistake is free, and it is the step most often skipped once somebody is confident.

• Resurface the spoilboard — Skim it flat occasionally so the bed is parallel to the machine rather than to the floor. Everything that depends on depth accuracy depends on this.
• Keep dust off the rails and screws — Grit in a linear bearing is what ends a machine’s accuracy. Brush and vacuum rather than blowing it further in.
• Lubricate to the maker’s schedule — Leadscrews and rails, with what the manufacturer specifies rather than what is in the shed.
• Check belt tension and squareness — Belts stretch. A machine that has gone out of square cuts parallelograms and it happens slowly enough not to notice.
• Retire cutters before they are dead — A blunt cutter burns, deflects and loads the machine. They are consumable, and running one too long costs more in ruined parts than a new one.
• Test and empty the extraction — A full extractor or a blocked filter quietly stops protecting you while appearing to run.

Cutters are consumable, they are the cheapest thing that decides whether a cut succeeds, and beginners economise on them hardest. A cutter is chosen by what it is cutting, how thick, and what finish you want — and using the wrong one is the second most common cause of a ruined part after workholding failure.

• Upcut — Pulls chips up out of the cut, which clears the slot and keeps the cutter cool. Leaves a ragged top edge on plywood and veneer.
• Downcut — Presses chips down, leaving a clean top edge. Packs the slot, so it needs shallower passes and generates more heat.
• Compression — Both, on one cutter — upcut at the tip and downcut above it — for a clean edge on both faces of sheet material. More expensive and worth it for plywood.
• Single flute — For plastics and aluminium, where clearing the chip matters more than the number of cutting edges. Plastic melts and re-welds if the chip stays in the cut.
• V-bit — For engraving and chamfering rather than for cutting through. The basis of carved lettering and most decorative work.
• Ball nose — For three-dimensional surfacing, where the rounded tip blends between passes. Slow, and what carved relief work is done with.

On materials: softwood and plywood are forgiving and where everybody starts. Hardwood needs more power and sharper cutters. MDF cuts beautifully and produces the worst dust in the workshop — see the safety section, because that is not a minor caveat. Acrylic and other plastics melt if the chip is not cleared, which is a chip-load problem rather than a speed problem. Aluminium is genuinely possible on a rigid machine with the right cutter and lubrication, and genuinely not on a light one. And composites, carbon fibre in particular, produce a dust that is a serious respiratory and electrical hazard and should not be cut on a hobby machine without controls designed for it.

Ordinary workshop care covers most of this hobby. Three hazards are different in kind, because two of them harm you slowly and invisibly and the third does not give you time to react.

The rest, briefly:

• Fire — A dull cutter dwelling in one place scorches and can ignite fine dust. Do not leave the machine, and keep an extinguisher suitable for the workshop nearby.
• Electrical — Spindles, VFDs and controllers involve mains voltages and stored charge. Wiring a machine is not a beginner electrical project, and a fault in extraction ducting can carry static.
• Composites and unknown plastics — Carbon fibre dust is conductive and hazardous to breathe; some plastics release harmful products when overheated. Know the material before cutting it.
• Manual handling — Sheet material is heavy and awkward, and most machine-room injuries in woodworking generally are backs and hands rather than cutters.

• From 2D to 2.5D — Pockets, profiles and V-carving — most useful work lives here, and it is where the machine earns its place.
• Into three dimensions — Ball-nose surfacing and relief carving. Slow, satisfying, and a genuine step up in CAM understanding.
• Into metal — Aluminium on a rigid machine, then a mill conversion. Effectively a different hobby with different tolerances and different risks.
• Into furniture — Sheet-goods joinery cut on the machine and assembled by hand. The route most woodworkers take, and where a full sheet machine starts making sense.
• Into building machines — A substantial number of people end up more interested in the machine than in what it cuts, which is an entirely respectable outcome and a real branch of the hobby.
• Into selling work — Common, and it changes the safety and insurance position rather than only the workload. Worth understanding before rather than after.