What Is the Program That Runs a CNC Machine?

• 1
  From CAD model to the program that runs a CNC machine
• 2
  What the controller actually reads
• 3
  Why 5-axis programs are a different problem
• 4
  Simulation, post checks, and the first cut
• 5
  Where the program stops and the shop takes over
• 6
  Program type by machine and part

A machinist never types coordinates by hand for a complex part. A designer exports a solid model, a CAM programmer selects tools and cutting strategies, and the software writes the toolpath. That toolpath still is not machine-ready. It is a list of movements in the CAM system's own reference frame, with no idea which spindle, which fixture, or which controller will execute it.

The program that runs a CNC machine is the output of the post-processor: a text file of G-code and M-code written for one specific machine and one specific controller. G-code defines motion and feed. M-code handles machine functions such as spindle start, coolant, tool change, and program stop. The controller reads the file top to bottom and moves the axes.

The distinction matters because a CAM file and a machine program are not interchangeable. Send the wrong post output to a machine and the first rapid move can drive a tool into a vise. The post is where machine limits, work offsets, and safe retract heights get baked in.

One more layer sits on top: the setup sheet. It tells the operator which fixture, which zero point, and which tool numbers the program assumes. Without it, a correct program still runs wrong.

• 1
  CADGeometry only. No tools, no feeds, no machine.
• 2
  CAMToolpaths in a neutral frame. Still machine-independent.
• 3
  Post-processorConverts toolpaths into controller-specific G-code.
• 4
  Setup sheetTies the program to a real fixture and zero point.

A block of G-code is one line. G00 is rapid positioning, G01 is linear feed, G02 and G03 are circular interpolation. F sets feed in mm/min or in/min. S sets spindle speed. A typical finishing pass on 6061-T6 aluminium might run 8,000 rpm at 1,200 mm/min with a 6 mm carbide end mill, while the same cutter in 17-4PH stainless drops to around 400 mm/min to keep tool load sane.

M-code covers everything that is not motion. M03 starts the spindle clockwise, M08 floods coolant, M06 calls a tool change, M30 ends the program and rewinds. On a mill-turn center these codes also switch between milling and turning modes, which is why the same part can need two programs on one machine.

Controllers are not all equal. Fanuc uses a fairly terse dialect. Siemens 840D supports more advanced cycles and shop-floor variables. Heidenhain on a five-axis machine is written in conversational format that looks nothing like plain G-code, though the CAM post handles that difference automatically.

This is also where subprograms and loops appear. A bolt circle with 24 identical holes does not need 24 hand-written blocks. A subprogram with a rotation offset repeats the cycle, and the file stays short enough for a human to check.

On a three-axis machine, the tool axis is fixed. The post only has to move X, Y, and Z. On a simultaneous five-axis center, two rotary axes move while the linear axes cut, so the controller must solve kinematics in real time to keep the tool tip on the intended path. Our 16 simultaneous 5-axis machining centers also use a Ø400 mm rotary table, and the post has to know that table's pivot distance and the exact spindle gauge length, or the tool will gouge on a contoured surface.

Two numbers dominate the outcome: tool center point management and the inverse-time feed. Tool center point keeps the programmed point at the tool tip when the rotary axes tilt, so the operator can adjust work offset without reprogramming. Inverse-time feed keeps chip load constant when the tool moves along a curved path in five axes, where linear feed would otherwise spike on the inside of a corner.

Not every part needs this. A flat plate with holes is a three-axis job and adding rotary motion only adds risk. Simultaneous five-axis earns its cost on organic curves, deep pockets with undercut walls, and features that would otherwise need three or four separate fixtures.

• 1
  Good fitImpellers, medical bone plates, thin-wall housings, angled ports.
• 2
  Poor fitSimple prismatic parts, high-volume simple turning, flat plates.
• 3
  Watch forShort tool gauge length, rotary table clearance, fixture collisions.

Nobody runs a new program straight onto a finished workpiece. The first pass is a dry run or a machine simulation. CAM software renders the stock removal and flags a collision between holder, fixture, and part. That check catches most errors, but it trusts the same post that wrote the file, so it cannot catch a wrong post setting.

The second check is the post itself. Verify the machine's travel limits, the safe Z retract, and the tool library. A program written for a 750 × 1,150 × 550 mm machine will not run on a compact 500 × 310 × 200 mm machine even if the part is small, because the rapids and the tool change position assume different geometry.

The third check is on the machine. Run the program in single block with rapid override down and the feed override at 10 percent. Watch the distance-to-go display on the first approach. If the number does not match the setup sheet, stop.

Our own flow adds a fourth step: raw material check before the run, in-process monitoring during it, and a final inspection before shipment, with reports on request. A program can be perfect and still produce a bad part if the stock or the fixture is wrong.

A program cannot compensate for a machine out of square, a worn spindle, or a fixture that flexes under load. Those show up as taper, chatter, or a size drift across a batch, and no amount of code editing fixes them. The operator reads the symptom and changes the setup or the tool, not the file.

Thermal drift is the common one. A machine that just started will hold a different size than the same machine after four hours of cutting. On a ±0.005 mm tolerance, that difference is enough to matter. Shops that hold tight tolerances warm up the spindle and re-check the first part after the run settles.

Tool wear is the other. A carbide end mill that cuts 6061 cleanly will start pushing material in 316L after a few hundred millimeters of cut. Adaptive toolpaths and constant chip load help, but the operator still decides when to change the insert.

This is why we treat the program as one input among several. Material, tooling, fixture, and inspection all carry equal weight. A well-posted file on a poorly prepared machine loses to a modest file on a controlled process every time.