3D CNC carved woodwork: what a shop can and cannot hold

Wood carving on a CNC is a subtractive process with a ball nose or tapered tool. A 3D CAM model is sliced into passes, and the tool follows the surface at a set stepover. Deep undercuts, twisted reliefs, and organic surfaces are where 5-axis earns its place, because the head tilts to keep the tool engaged instead of dragging the shank.

Simultaneous 5-axis work is not automatically better. For a panel with shallow relief, a 3-axis pass with a long tapered tool is faster and cheaper, and the finish is the same. We move a part to a 5-axis center when the surface normal swings more than roughly 45° from the tool axis, or when the part is too tall to reach in a single setup.

Our 5-axis machining centers handle work up to 4,000 mm, with common travels of 750 × 1,150 × 550 mm and 600 × 600 × 600 mm. A Ø400 mm rotary table covers round and curved carvings that need indexing between faces. Those numbers matter when you are deciding whether a carved element ships as one piece or gets split for assembly.

The honest limit is depth-to-diameter. A 3 mm ball nose cutting 40 mm deep will chatter and leave witness marks, no matter the machine. If your design has narrow deep pockets, expect us to reduce depth per pass, add a roughing tool, and quote more hours. Redesigning that pocket to a wider radius usually costs less than machining it as drawn.

• 1
  Good fitOrganic relief, sculpted furniture parts, curved moldings, carved panels, art and architectural elements
• 2
  Marginal fitVery deep narrow slots, thin unsupported walls, features smaller than the tool radius
• 3
  Poor fitFlat parts with simple profiles, which a 3-axis router does faster

Species choice decides more of the final result than the machine does. Grain direction controls whether a cutter shears fibers cleanly or tears them. On a carved surface, the tool crosses grain in every direction, so some tear-out is normal on open-grain species and needs hand sanding or a finishing pass with a smaller stepover.

Softwoods such as pine and basswood cut quickly and take fine detail with little tool wear, which makes them useful for form studies before committing to a hardwood. They also dent easily, so a softwood carving is rarely the right choice for a part that gets handled daily.

Dense hardwoods hold crisp edges and thin sections far better. Ebony, logwood, and stabilized knots machine cleanly but blunt tooling fast; we adjust feed, speed, and tool coating accordingly. The key is matching tool, toolpath, speed, and feed to the specific species rather than running one recipe across the board.

Engineered wood behaves differently again. MDF and plywood are dimensionally stable and cheap for jigs, molds, and pattern work, but they are abrasive on carbide and produce fine dust that needs extraction. If a carved part will be painted, MDF removes the grain-matching problem entirely.

Moisture is the quiet failure mode. Wood moves as it equalizes with shop air. A carving cut from stock at 12% moisture can shrink or cup after it reaches a dry building. For parts that must stay flat, we ask for kiln-dried stock and, where the design allows, we cut a rough pass first and finish after a rest period.

On a 3D carved surface, visible scallops come from stepover. A 6 mm ball nose at 0.5 mm stepover leaves a different surface than the same tool at 0.15 mm. The first needs sanding; the second often does not, or needs only a light pass. This is the single biggest lever on both lead time and finishing cost, and it is worth deciding early.

We work to ±0.005 mm (±0.0002 in) on machined dimensions where the drawing calls for it, but that figure applies to the machine and the setup, not to a wood surface after it has moved. Wood is not a stable metrology material. If a carved feature has a true position or fit requirement, we machine that interface in a material that holds it, or we build a fixture that constrains the wood during the cut.

Roughing and finishing run as separate operations. A larger tool removes bulk at a coarse stepover, then a smaller tool follows the same surface at a fine stepover. Skipping the roughing pass saves nothing; it loads the finishing tool and burns it.

We measure surface finish on metals at Ra 0.2–0.8 μm for fine work and Ra 0.8–1.6 μm for standard high-quality surfaces. On wood we judge by hand and by comparison sample, because fiber pull-out dominates the reading. Ask for a sample coupon if the texture is critical and we will cut one from your species.

We work from STEP (.stp), IGES (.igs), SolidWorks (.sldprt), and Parasolid (.x_t) models. For organic carved surfaces where the model is already a mesh, a high-quality STL is acceptable, but the mesh resolution sets the surface quality ceiling. A coarse STL gives a faceted carving no toolpath can fix.

Send the model plus three things: overall size, the surface areas that matter, and the faces that must locate in a fixture. If the part is carved on one side only, we can hold it on a flat back and the setup is simple. If it is carved on several faces, we need a way to reposition it accurately, and that usually means a machined cradle or index pins.

Quotation and a free DFM analysis come back within 12 hours. Production can start within 24 hours, and parts ship in 3–5 days. Prototypes and one-offs are fine; there is no minimum order quantity, and a run can scale to 10,000+ parts. Uploads stay confidential, and an NDA is available on request.

One thing we check before anything else: is wood the right material? If the part needs a tolerance that wood cannot hold, carries load, or sits outdoors without a coating, metal or a composite usually costs less over the life of the product. We will say so rather than machine something that fails later.

Carving is efficient for one part to a few hundred, especially when the shape is complex and the design may still change. It becomes expensive per unit when the same carved shape repeats thousands of times, because every part spends machine hours on a surface that could be copied instead.

At that point the carving becomes a pattern. We machine the master in wood or MDF, finish it, and use it for vacuum casting or die casting depending on the material and volume. That keeps the hand-worked surface you approved while the per-part cost drops.

Multi-piece assemblies are another route. A large carved element split into three sections machines faster, fits inside smaller travels, and is easier to replace if one section is damaged. The joint lines need to fall where the design can hide them, so bring that decision into the model rather than the shop floor.