CNC machining game rules for prototyping: what changes between one part and a million

• 1
  Geometry is a machining decision, not a modeling decision
• 2
  Tolerances are a budget, and prototypes should spend less of it
• 3
  Setups drive cost more than machine time in prototyping
• 4
  Surface finish is a process choice with a real cost curve
• 5
  Build the prototype so the production part can be different
• 6
  Prototype vs production: where the rules change

A CAD model has no toolholding. The machinist adds that later, and the geometry you drew decides how much of it can be reached. Internal corners modeled as a sharp 90° will come out with the radius of whatever cutter reaches them. On a Ø6 mm end mill that means a 3 mm corner radius. Draw the radius you can live with instead of letting the shop pick it for you.

Deep pockets are the other common trap. A pocket deeper than about 3× its width forces a long, thin cutter. That cutter deflects, chatters, and needs slower feed rates. If a pocket is 10 mm wide and 45 mm deep, expect a slower cycle and a rougher floor than the rest of the part. Sometimes the fix is to split the part into two pieces and bolt them together.

Undercuts, internal threads at the bottom of blind holes, and features on five faces all push the part toward more setups or a 5-axis machine. That is not automatically wrong. It just changes the price and the lead time. Send the model early and we will tell you which features cost the most to reach.

One habit saves money: model the part as the machinist will hold it. Add a small flat where the vise will grip, or a boss you can remove later. It is easier to delete a tab than to fixture a part with nothing to clamp.

Every tolerance on a drawing costs money. The tighter the band, the slower the passes, the more inspection, and the higher the chance a good part gets scrapped. Our general machining tolerance is ±0.005 mm (±0.0002 in) when a feature needs it, but applying that number to every dimension on a prototype is wasteful. Most prototype dimensions only need to be functional.

Ask one question per dimension: does anything touch this? A bearing bore, a shaft diameter, a mating hole pattern, and a seal groove all matter. An outer profile, a cosmetic edge, or a clearance hole usually does not. Mark the critical few and let the rest run at general tolerance. The quote gets simpler and the part gets cheaper.

Datums need the same discipline. If a drawing calls three datums that never get used in a tolerance stack, the inspector still has to set up for them. Pick the datum that matches how the part sits in the assembly, then reference everything back to it. On a prototype, a clean datum scheme is worth more than a long list of tight callouts.

Materials behave differently at the same tolerance. Aluminum 6061 and 7075 hold a ±0.005 mm band more easily than a thin-wall titanium or magnesium part, where springback and heat move the number after the cut. If the feature is thin, tell us and we will suggest a realistic band instead of failing the first article.

For a one-off part, the clock spent finding and clamping the workpiece often costs more than the cutting. Each setup means a new zero, a new probe, and a new chance to shift a feature. Reducing setups is the single biggest lever you have on prototype price.

Design for one or two setups when you can. A part that can be made from one side, flipped once, and finished is cheap. A part that needs four orientations on a 3-axis mill is not. This is where 5-axis work earns its place: a single simultaneous setup reaches five faces, keeps one datum, and removes the re-fixturing error between operations.

We run 16 simultaneous 5-axis centers, 12 four-axis mills, 27 three-axis machines, and 16 mill-turn centers across 127 high-precision machines. The point is not the count. It is that we can pick the setup strategy that fits the part instead of forcing it onto whatever is free. A mill-turn center, for example, can cut a turned shaft and its cross-holes without ever releasing the part.

Size matters too. Our maximum processing size is 4,000 mm, with travels like 750 × 1,150 × 550 mm and a Ø400 mm rotary table for parts that need continuous rotation. If your prototype is large and thin, say so early. Large thin parts need different workholding than the same shape at 100 mm.

Ra numbers are not decoration. As-machined finish lands around Ra 1.6–3.2 μm and comes with the cut. A high-quality finish at Ra 0.8–1.6 μm needs a finer stepover or a finishing pass. A fine finish at Ra 0.2–0.8 μm needs more time still, and sometimes a different tool. Each step up roughly adds cycle time.

Pick finish by function. A sliding surface, a seal face, or an optical mount deserves the tighter band. A bracket, a housing interior, or a prototype that will be painted does not. If you are not sure, ask for the as-machined surface and check it against the assembly before paying for polish.

Coating and anodizing change dimensions. Hardcoat anodizing builds thickness on the surface, and a tight bore can close up after treatment. Electroless nickel and plating do the same. Tell the shop which dimensions must survive finishing, and mask or compensate before the part goes into the tank.

Cosmetic expectations need to be set before the first cut. Bead blasting hides tool marks and gives a uniform matte look. Brushing leaves directional lines. Polishing removes material and softens edges. A prototype that will be shown to customers is usually worth a blasting pass; a prototype that only gets measured is not.

A prototype is a question, not a product. The best prototypes answer the question fast and leave room to change the answer. That means avoiding features that only exist because the prototype process made them easy.

Watch for prototype-only geometry. A pocket added to save weight on a machined part may not survive the move to die casting or injection molding, where draft and wall uniformity matter. A machined rib that is 1.2 mm thick might be fine here but too thin for a cast wall. If you already know the production process, design toward it now.

Keep the material honest. If the production part will be 6061-T6 and the prototype is 6061, the prototype will behave close to the real thing. If the prototype is a plastic stand-in for a metal part, treat stiffness and thermal results as a rough guide only. We machine aluminum 6061, 7075, 2024, stainless 303 and 17-4PH, titanium TC4, PEEK, POM, and carbon fibre, so a like-for-like prototype is usually possible.

Finally, capture what you learned. Note the dimensions that were hard to hit, the corners that had to be opened, and the finish that did not matter. That record is the cheapest DFM input you will ever have when the design goes to production tooling.