Furniture Design Made by CNC: How the Cutting Actually Works

• 1
  What Happens Inside the Machine
• 2
  Where 5-Axis Motion Changes the Answer
• 3
  Wood, Metal and the Materials Between
• 4
  Tolerance, Finish and What Drives Cost
• 5
  When CNC Is the Wrong Answer
• 6
  Process Choice by Part Geometry

Furniture design made by CNC starts as a CAD solid, usually a STEP or native file. CAM software slices that solid into passes, assigns each pass a tool, a feed rate and a stepover, then writes G-code the controller can run. The cutter follows those coordinates and removes stock until the remaining material is the part. Nothing about the shape is limited by a mold draft angle or a press brake radius.

Cutting is a force problem before it is a shape problem. A 12 mm carbide end mill at 18,000 rpm and 3,000 mm/min in 6061 aluminium pulls a few hundred newtons sideways. On a 2,500 mm table leg that load bends the blank, and the finished wall can drift 0.1–0.3 mm from the nominal surface.

That is why long, slender furniture parts are usually cut in two or three setups with support blocks, or from thicker stock that gets trimmed in a second pass. A 40 × 40 mm solid leg with a 6 mm wall through the middle is a different job from a 40 × 40 mm leg machined from a 60 mm block. Both are possible. Only one holds tolerance over 2 m.

Climb milling, sharp tools and light radial engagement keep that force predictable. When a shop skips those settings to save cycle time, the symptom shows up later as chatter marks on the visible face.

A three-axis machine moves the tool in X, Y and Z. The part stays put. That covers flat panels, straight rails, simple brackets and most joinery plates. It is also the cheapest way to hold ±0.005 mm on a face that only needs one approach angle.

A five-axis machine tilts the tool or the table, so the cutter reaches an undercut without re-fixturing. A chair leg that sweeps in two planes, a backrest with a compound curve, a seat shell that wraps down at the edges: those are five-axis shapes. The gain is not only reach. Shorter tools can be used on deep pockets, which cuts vibration and improves the floor finish.

Simultaneous five-axis work costs more per hour than indexed three-axis work, so the honest rule is simple. If the geometry can be reached in three setups, do that. If it cannot, five-axis usually beats building three fixtures and chasing datums across them.

GreatLight runs 16 simultaneous five-axis machining centers, 12 four-axis mills and 27 three-axis machines. That mix matters because it lets us route each feature to the machine that fits it rather than pushing everything onto the most expensive spindle.

Solid hardwood cuts cleanly at high spindle speed and moderate feed. Hard maple, white oak and walnut are common. The catch is movement. A 900 mm oak panel can swell 4–6 mm across the grain between 30% and 70% relative humidity, so any tight tolerance on a wood furniture part has to be called out relative to a controlled environment, not in absolute terms.

Aluminium is the usual choice when the design shows metal. Grades 6061 and 6082 machine well, anodize evenly and hold thin walls better than steel. Grade 7075 gives higher strength for load-bearing joints but anodizes to a darker, less uniform tone. Stainless 304 and 316 are used for legs, frames and hardware where corrosion or a brushed finish matters.

Sheet materials behave differently again. Plywood, MDF and veneered board are dimensionally stable in-plane but weak through the core, so screws and dowels need enough depth below the surface. A 18 mm panel with a 12 mm deep pocket leaves only 6 mm of material under the cutter, which will bow under clamping pressure.

Composite and plastic parts are worth a second look for prototypes. POM and PC machine fast and take a thread, which is often enough to validate a joint before committing to aluminium tooling.

Tolerances on furniture parts are rarely as tight as the machine can hold. A visible gap between a leg and a seat wants ±0.2 mm. A bearing pocket in a swivel base wants ±0.02 mm. A locating shoulder in a knock-down joint sits somewhere in between. Writing ±0.005 mm across a whole drawing does not buy a better chair; it buys inspection time on features that never touch another part.

Surface finish follows the same logic. As-machined faces sit at Ra 1.6–3.2 μm, which reads as a fine stepover pattern under raking light. Bead blasting or tumbling at Ra 0.8–1.6 μm hides tool marks and gives an even sheen. Polished and anodized faces reach Ra 0.2–0.8 μm and are the ones that show every fingerprint.

Cost is driven by three things in this order: the number of setups, the volume of material removed, and the finish specification. A part that removes 70% of its stock as chips pays for that stock twice. A hollowed shell with a 4 mm wall can cost less than a solid block of the same outer shape, because the machine spends less time in the cut.

Batch size changes the calculation too. At one piece, programming and fixturing dominate. At 500 pieces, cycle time dominates and a light redesign of a fillet or a pocket depth can move the unit cost more than a supplier change.

CNC is a subtractive process, so it is inefficient whenever the finished part is mostly air. A hollow cabinet shell, a large drawer box or a panel with wide openings wastes most of the block. Sheet metal fabrication, vacuum casting or 3D printing will beat it on cost and often on weight.

Volume is the second limit. A molded or die-cast part spreads its tooling cost across thousands of units. Below a few hundred pieces the tooling rarely pays back, and machining stays competitive. Above that, the math usually flips, especially for simple shapes with generous draft.

Size is the third. GreatLight machines up to 4,000 mm on the largest travel, with a Ø400 mm rotary table for round work. Beyond that, a part has to be split and joined, and the joint becomes a visible design element rather than a hidden detail. That is a design decision, not a machining one.

There is also a finish limit worth naming. Laser engraving needs a minimum character height of 1.5 mm to stay legible after anodizing. Fine text and thin logos on a curved surface will not survive the process cleanly.