How Can CNC Machines Create 3D Shapes?

• 1
  Key takeaways
• 2
  What CNC machines create 3D shapes from: model to toolpath
• 3
  Choosing the axis count for the shape
• 4
  Workholding, stock and the shapes that fight back
• 5
  Verifying the shape after cutting
• 6
  When another process is the better answer
• 7
  Step by step: from file to finished 3D shape
• 8
  Which setup fits your 3D shape

Everything starts with a 3D model, and the model format matters more than most people expect. Send a STEP or Parasolid file. Those carry true surfaces, so the CAM system can offset a tool along them. An STL mesh is only triangles, and a coarse mesh forces the software to guess at the curve between facets. The guess shows up as a faceted surface on the finished part.

Once the file is clean, CAM software slices the surface into passes. Think of contour lines on a topographic map. Each pass sits at a fixed stepover, the distance between neighboring toolpaths, and each one is a line the cutter can follow. A 0.2 mm stepover on a curved surface produces a much smoother result than 0.5 mm. It also takes roughly two and a half times longer to cut.

The controller then interpolates. Modern controllers read a stream of point coordinates and fit a smooth curve through them, so you get arcs instead of a staircase of straight moves. This is how CNC machines create 3D shapes without a physical template. The shape exists only as math until the cutter traces it.

One check before you cut anything: look at the toolpath in simulation. Watch for sudden direction changes at the edge of a surface, and for passes that leave uncut islands. Fixing that in software costs minutes. Fixing it on the machine costs a scrapped part and a new block of stock.

3-axis machining moves the table in X and Y and the spindle in Z. All cutting comes from one direction. It handles pockets, steps, bosses and shallow sculpted faces very well, and it is the cheapest way to remove material. What it cannot do is reach the back side of a shape without a second setup. Every extra setup adds a re-fixturing error, usually 0.02–0.05 mm on a well-made vise.

4-axis adds rotation around one axis, normally A. That lets the part turn while the cutter works, so you can machine a cylinder with flutes, a cam profile or a ring of pockets in a single run. The rotary table we use is Ø400 mm, which covers most shaft-type work. It is a strong fit for parts that are round and have features around the circumference.

5-axis adds a second rotary axis, so the tool can tilt. This is the setup that reaches undercuts, deep cavities and organic surfaces in one pass. Impellers, turbine blades, medical implants and complex housings are typical. Tilting also lets the cutter stay on the surface normal, which keeps a consistent contact point and improves both finish and tool life.

Pick the lowest axis count that reaches every feature. A part that only needs faces cut from three directions does not get faster on a 5-axis machine. Setup drops, but programming time and cycle time rise. We quote the axis count the geometry actually needs.

A 3D shape has few flat faces, which makes it hard to clamp. The usual answer is a soft jaw set machined to the part contour, or a fixture plate with tapped holes and toe clamps. For thin or awkward parts, machine a pocket into a block of aluminium, set the workpiece into it and hold it with low-viscosity fixture wax. The part is supported across its whole lower face, so it does not ring or spring.

Stock allowance matters. Leave 0.5–1.0 mm on sculpted surfaces for the finishing pass. More than that means heavy roughing cuts that push the part around. Less than 0.3 mm risks cutting into a surface that has already been shaped, especially on castings and forgings where the skin is not where the model says it is.

Some geometry is simply hard to machine. Deep narrow slots need long, slender tools that deflect. Sharp internal corners need a cutter radius smaller than the corner, and small cutters break. Undercuts that a tilted tool can reach on one side may be unreachable on the other. A short DFM note on the drawing can save a whole revision cycle.

Wall thickness is the other limit. Below about 1 mm in aluminium, cutter pressure and heat start moving the wall. Rough with light radial engagement and finish with a spring pass. If a wall is cosmetic only, say so. We can leave a thicker section and blend it.

A curved surface cannot be checked with calipers alone. For profile work we use a CMM to compare the measured surface against the model, then report the maximum deviation. For a part held to ±0.005 mm, that comparison is the only honest answer. Surface finish is measured separately, usually with a portable roughness tester.

Finish values depend on the process. As-machined surfaces typically land at Ra 1.6–3.2 μm. A careful finishing pass gets to Ra 0.8–1.6 μm. Polished or lapped surfaces reach Ra 0.2–0.8 μm and cost more time. Tell us the finish callout that matters; specifying Ra 0.2 μm across an entire part rarely pays off.

In-process checks catch trouble early. On a long roughing job we stop after the first finishing pass and check the wall thickness and one critical profile. If the cutter has worn, we change it before the rest of the surface is cut. That costs a few minutes instead of the whole part.

Every order leaves here after 100% inspection. Raw material certificates are checked on arrival, dimensions are monitored during the run, and the final part is measured against the drawing before packing. Inspection reports are available on request.

CNC cutting is subtractive. It removes material from a solid block, which is the right approach when you need tight tolerances, good surface finish and a known material. It is not always the fastest or cheapest route to a 3D shape.

If the geometry is a thin shell with no tight tolerances, 3D printing may be quicker and lighter. If you need 50 identical complex housings with moderate tolerances, die casting or vacuum casting starts to win on unit cost. Sheet metal fabrication handles flat and formed panels far faster than milling them from plate.

The useful question is not which process is better in general. It is which one holds the features you actually need. A part with two critical bores and a cosmetic curved shell can be cast and then machined only on the bores. That combination often beats either process alone.

We run milling, turning, sheet metal, die casting, vacuum casting and 3D printing under one roof, so the recommendation is not tied to one machine. If we think a different process fits better, we say so with the numbers behind it.