Rapid prototyping turns a CAD (computer-aided design) file into a physical part without building a mold or die first. The machine reads the file directly, either adding material layer by layer in a 3D printer or cutting the part from solid stock on a CNC (computer numerical control) machine. A first physical version can exist within hours or days of the design. Understanding how rapid prototyping works comes down to two decisions: which process builds the part, and what you do with what the part teaches you.
The method matters because every prototype exists to answer a question. A part built to check whether two pieces assemble has different needs from one a client will hold in a review, and different again from one that has to survive a load test.
How Rapid Prototyping Works, From CAD File to Part
Every route starts with the same thing: a clean 3D model exported for the machine that will make it. From there the processes split into two families. Additive processes build the part up: FDM (fused deposition modeling) lays down melted plastic filament, SLA (stereolithography) cures liquid resin with light, and SLS (selective laser sintering) fuses nylon powder with a laser. Subtractive processes start with a block of plastic, wood, or metal and cut away everything that is not the part.
Neither family needs a mold or die, which is what makes both fast enough to iterate with. The trade-offs sit elsewhere: in surface finish, strength, accuracy, and cost. Our pillar guide on getting started on prototyping covers how to scope a first round before any of those choices come up.
Which Method Fits the Question the Prototype Has to Answer?
Pick the process by what the part needs to prove. Each of the four common routes answers a different kind of question well.
FDM is the low-cost starting point for early form and fit checks, since a simple part in small numbers costs less this way than by any other route. Its weak spot is where one layer meets the next. Across the layers, plan on roughly half the plastic’s strength, and orient anything that carries load so the force runs along the layers instead.
SLA gives the smoothest surface of the printing processes and holds fine detail, so it suits appearance models, small features, and anything headed to a client review. It costs more than FDM, and some standard resins are brittle, so a clip or hinge that has to flex belongs in a tougher material.
SLS builds without supports, since the unfused powder surrounding the part carries it through the print. That opens up internal channels and undercuts, and a nylon part comes out almost as strong across its layers as along them. Packing many parts into one build also brings the cost per piece down on a batch.
CNC cutting makes the part from solid stock rather than building it, which suits prototypes that need a particular sheet or block material, flat faces, or a machined surface. For a one-off it usually costs more than printing, because the programming and setup all land on a single part. At Kemperle we run all three printing processes plus CNC routing for plastics, wood, and soft metals, so the choice follows the part rather than the equipment on hand.
What Tolerance Can You Expect From Each Method?
Published service standards give a useful baseline. Protolabs Network, a large manufacturing marketplace, holds desktop FDM parts to ±0.5% with a floor of ±0.5 mm, prototyping-grade SLA and SLS parts to ±0.3% with a floor of ±0.3 mm, and industrial SLA to ±0.2% with a floor of ±0.13 mm. Machined parts with no tolerance on the drawing default to ISO 2768, the general-tolerance standard: medium class for plastics and fine class for metals.
The percentage matters once parts get bigger. On a 200 mm part, ±0.5% is ±1 mm for desktop FDM, and ±0.3% is ±0.6 mm for prototyping SLA or SLS. A bracket that fits in the hand sits near the floor figure, while a housing the length of a forearm is governed by the percentage.
That default is worth knowing before you assume a machined part is automatically the accurate one. On a feature between 30 and 120 mm, ISO 2768 medium allows ±0.3 mm in plastic, the same band a prototyping SLA part carries. Machining pulls ahead when the drawing calls for tighter numbers, so we would put a tolerance on every dimension that matters and leave the rest to the default.
These figures describe what a service commits to, not what every machine achieves. A well-tuned printer can beat them on a small part, and a large, thin, or warp-prone part can fall outside them, which is one more reason to measure the features that decide the fit.
What Happens After the First Prototype?
Rapid prototyping is iterative by design. Test the part against the specific question it was built to answer: does it assemble, does a feature that looked fine on screen work in the hand, does a wall survive handling. Write the findings down, because a written list keeps the next revision targeted.
Then put every fix back into the CAD model as a real dimension or geometry change, never as a file-down or shim on the physical part, since only the model carries forward into production. Change as few things as you can per round. If three dimensions move at once and the fit improves, nobody can say which change did it, and the next problem gets harder to trace. Our guides on why a structured prototyping process matters and how iterative design and prototyping works cover how many rounds to plan and how to know when to stop.
For assemblies or parts where fit is critical, we would scan the revised prototype and compare it with the updated CAD model before closing the round. That comparison lays a color map over the part showing which areas are in or out of tolerance, and it exposes any gap between the change you meant to make and the part the machine delivered.
Planning a first round? Tell us what the prototype has to prove and which dimensions decide the fit, and attach the CAD file. Start from our quote request form or phone 718-557-9578. From there we match the round to one of our 3D printing services or to CNC routing, along with a short list of what to measure before round two.
