The CNC machine briefly explains how a CAD file becomes metal

• 1
  What a CNC machine briefly explains about motion
• 2
  How axis count changes what you can cut
• 3
  Why the setup decides the tolerance, not the spindle
• 4
  Surface finish: what the numbers mean in practice
• 5
  Where CNC stops being the right answer
• 6
  Which machine setup fits which part

Strip the enclosure away and a CNC machine is three systems working against each other: a spindle that turns the tool, a set of slides that position it, and a controller that decides where it goes next. The controller reads a program of coordinates and feed rates, then drives each slide to the number written in the block. Nothing about the cut is decided by hand once the cycle starts.

That is the whole difference from a manual mill. On a manual machine, the operator reads a dial and feels the cut. On a CNC machine, the geometry is already fixed in the program, so the same part can be produced on a Monday morning and again three weeks later with the same result. Repeatability comes from the program, not from the person standing at the door.

The program itself is short. A typical finishing pass on an aluminum bracket is a few hundred lines of G-code: rapid to a safe height, plunge to depth, follow the contour, retract. CAM software generates most of it from the solid model. The programmer's real job is choosing tools, stepover and feeds, not typing coordinates by hand.

So when someone says a CNC machine briefly explains a part, they usually mean the machine only executes what the model and the setup already decided. Change the model, and the part changes. Change the setup, and the part may change too, even if the model is identical.

• 1
  Cutting motion is programmed, not freehandEvery move is a coordinate and a feed rate in the block.
• 2
  Repeatability lives in the programTwo runs weeks apart hold the same dimensions.
• 3
  CAM writes the codeThe engineer picks tools, stepover and speeds.

An axis is a direction the machine can move the tool or the work. Three linear axes, X, Y and Z, cover most prismatic parts: plates, housings, brackets with features on one face or a few faces reached by refixturing. A three-axis machine with travels of 750 × 1,150 × 550 mm will handle a large share of industrial work.

Add a fourth axis and the work rotates. The common form is a rotary table, often Ø400 mm, that indexes or turns continuously while the tool cuts. That lets you reach features around the perimeter of a part without unclamping it: cam slots, cross-drilling patterns, engraving that wraps a cylinder.

Five-axis means two rotary motions on top of the three linear ones. The tool can tilt and the work can rotate at the same time. On a simultaneous five-axis center, the controller keeps both rotations moving during the cut, so a ball nose tool can stay normal to a curved surface. That is how impellers, turbine blades and deep pockets with drafted walls get machined in one setup.

The trade is not only cost. Five-axis toolpaths are harder to verify, stock needs more clearance, and rigid setups matter more because the tool reaches out at an angle. For a simple plate with holes, five-axis adds nothing. For a part with features on five sides, it removes three setups and the position errors that come with them.

• 1
  Three axesPrismatic parts, one main direction of access.
• 2
  Four axesRotation around the work; perimeter features in one clamp.
• 3
  Five axesTilt plus rotation; contoured surfaces in a single setup.

A machine that holds ±0.005 mm on a test coupon will not hold it on a part that moves in the vise. Clamping force distorts thin walls. A long tool deflects under load. A part that is only supported at two points will ring during a finishing pass and leave chatter marks on the surface.

Good setups follow a simple order. Locate on a machined face or a bored hole rather than a raw casting surface. Support the part under the cutting zone, not just at the ends. Keep the tool as short as the geometry allows. If a feature needs two operations, reference the second setup to a feature cut in the first, not to the raw stock.

For thin-wall aluminum parts, we often rough with the part still attached to a tabbed blank, then finish after a stress-relief pause. The blank keeps the wall from springing. Cutting it free at the end costs one extra operation and saves the dimensional drift that shows up hours later.

This is also why a quote asks for the drawing and not just a photo. Datum callouts, tight tolerances on one face only, and a note about which surfaces are cosmetic all change how the part is held. A setup that ignores those notes will pass inspection on paper and fail at assembly.

• 1
  Locate on machined featuresRaw cast or saw-cut surfaces vary too much.
• 2
  Support under the cutUnsupported spans chatter and drift.
• 3
  Reference setups to each otherSecond-op datums should come from the first op.

Ra is the arithmetic average of surface roughness, measured in micrometres. As-machined surfaces typically land at Ra 1.6–3.2 μm. A high-quality finish on a well-supported part reaches Ra 0.8–1.6 μm. Fine finishing with a small stepover and a sharp tool can reach Ra 0.2–0.8 μm on aluminum and some steels.

The number is not free. Getting from 3.2 to 0.8 usually means a slower finishing pass, a smaller stepover and sometimes a second tool. Cycle time may double. On a cosmetic face, that is worth it. On a hidden mounting face, it is money spent for nothing.

Finish also depends on the material. Aluminum 6061 and 7075 cut cleanly and take a fine finish well. Stainless 316 and 17-4PH work-harden, so light finishing passes with the right feed can glaze the surface instead of cutting it. Titanium TC4 (Ti-6Al-4V) needs sharp tools, low cutting speed and plenty of coolant, or the finish suffers and tool life drops fast.

Machining marks are directional. If a part will be anodized later, the marks show through clear anodizing and may look uneven across faces cut with different tools. Bead blasting before anodizing hides most of that. Where a brushed or polished look is required, the grain direction has to be specified on the drawing.

• 1
  Ra 1.6–3.2 μmStandard as-machined finish on most parts.
• 2
  Ra 0.8–1.6 μmHigh-quality finish with a dedicated pass.
• 3
  Ra 0.2–0.8 μmFine finish; slower cycle, tighter tool control.

CNC subtracts material from a solid block. That is efficient when the part is small relative to the stock, and wasteful when it is not. A large housing machined from a 200 kg billet may produce 170 kg of chips. Die casting or sheet metal fabrication will beat it on cost once quantities pass a few hundred pieces.

Very hard materials are a second limit. Tool steel and Inconel can be machined, but tool wear is high and cycle times stretch. For a one-off prototype, that is acceptable. For a production run, the cost per part rarely justifies it when a casting or forging gets closer to net shape.

Geometry sets the third boundary. Sharp internal corners cannot be cut by a round tool. A 6 mm end mill leaves a 3 mm corner radius at best, so a square internal corner needs EDM or a redesign. Deep holes with a high depth-to-diameter ratio need peck drilling and a long, thin tool that deflects. Both are possible, both cost more.

On the other side, CNC handles the cases that casting cannot. One prototype, ten thousand parts, no minimum order quantity, design changes between runs. That flexibility is why it stays the default for prototypes and low-to-mid volume production.

• 1
  High material removalCasting or sheet metal wins past a few hundred parts.
• 2
  Hard alloysMachinable, but tool wear raises cost per part.
• 3
  Sharp internal cornersRound tools leave a radius; EDM or redesign needed.