What Does the Language Look Like That CNC Machines Use?

• 1
  Five things worth knowing before you read further
• 2
  G-code: the cnc machines language at its core
• 3
  CAM software and post-processors: how code gets written
• 4
  What changes when you add rotary axes
• 5
  Where the code shows up on your part
• 6
  What the cnc machines language cannot do
• 7
  Which program format fits which job

Every CNC machine, whether it is a 3-axis mill in a job shop or a 16-pallet 5-axis cell, runs on the same idea. A line of text tells the controller one thing to do. G01 X50.0 Y25.0 F800 means move in a straight line to that point at 800 mm/min. G02 adds an arc. M06 changes the tool. M03 starts the spindle clockwise. That is the whole grammar, repeated thousands of times per part.

The standard behind it is ISO 6983, often called G-code or RS-274. It dates back to paper tape readers, which is why the syntax is so terse. A letter addresses a register: G for preparatory function, M for miscellaneous function, X Y Z for linear axes, A B C for rotary axes, F for feed, S for spindle speed, T for tool number. Nothing is spelled out because every character once cost a row of punched holes.

Controllers are only loosely compatible. Fanuc, Siemens, Heidenhain, Mitsubishi and Haas all read the core G01/G02/G03 set, but canned cycles, macro calls and high-speed look-ahead differ. A post-processor that works for one brand will not post clean code for another. This is why a shop with mixed machines keeps multiple post configurations, and why sending a program between shops is rarely plug-and-play.

For a buyer, the practical point is this: the code is the last translation step between your CAD file and the metal. Tolerance, surface finish and cycle time are all decided in that translation. A ±0.005 mm callout does not survive a program that feeds a Ø6 mm end mill at 2,500 mm/min through 4140 steel.

• 1
  G01 / G02 / G03Linear, clockwise arc and counterclockwise arc moves. The workhorses of every program.
• 2
  F, S and T wordsFeed rate, spindle speed and tool selection. Set per operation, not per part.
• 3
  M-codesMachine actions: spindle on/off, coolant, tool change, program end.
• 4
  Work offsets G54–G59Define where the part sits relative to machine zero. Wrong offset, scrapped part.

Nobody hand-writes 40,000 lines for a contoured mold cavity. A CAM programmer imports the solid model, sets stock, picks tools, and generates toolpaths. The CAM engine works in its own internal representation: cutter location data, CL data for short. That is geometry only. The post-processor is what converts CL data into the dialect your specific controller understands.

The post is where most real-world problems live. A generic post will produce code that runs but wastes time, or worse, drives a tool into a fixture. A tuned post knows the machine's travel limits, its rapid retract plane, its spindle warm-up routine and its tool-change positions. At GreatLight we run 127 high-precision CNC machines across 3 wholly-owned plants, and each machine family has its own post. That is not bureaucracy. It is the difference between a 4,000 mm gantry part running clean and a crash.

Modern CAM also handles rest machining, adaptive clearing and toolpath smoothing. Adaptive or trochoidal paths keep radial engagement low, which lets you run higher feed rates without burning the tool. On 17-4PH stainless this can cut cycle time substantially compared with a conventional offset path, and it holds a more consistent Ra. The code looks strange to a human reader. It performs better.

One caution for engineers reviewing a quote: cycle-time estimates are only as good as the post and the tool library behind them. If a shop quotes a 3-hour cycle on a part that CAM says needs 5 hours, either they have a better process or they have not simulated the program. Ask which one.

• 1
  CL data vs G-codeCAM outputs geometry; the post converts it to machine-specific words.
• 2
  Post-processor tuningRapids, retracts, safe planes and tool-change positions set per machine.
• 3
  Adaptive clearingLow radial engagement, higher feed, longer tool life on hard alloys.

A 3-axis program moves X, Y and Z. A 5-axis program adds two rotary words, usually A and C or B and C, and can move them at the same time as the linear axes. That simultaneity is the whole point. It lets a ball nose cutter stay normal to a curved surface, which gives a better effective cutting speed and a more even scallop height.

The syntax itself does not grow much. The complexity moves into the CAM strategy and the post. A 5-axis post must solve the machine's kinematics, check that the tool holder does not hit the table or the trunnion, and manage singularity points where two rotary axes line up and the controller cannot decide which way to turn. Get that wrong and the machine makes an unpredictable move. In a 16-machine 5-axis department, we simulate every new program before it touches metal.

There is a second family of formats worth knowing. STEP-NC (ISO 14649) tries to describe features and operations rather than raw motion, so a controller could in theory decide its own toolpaths. It has been in development for decades and remains rare in production. APT, an older higher-level language, still appears in aerospace post chains. Neither replaces G-code at the spindle today.

The engineering meaning is simple. More axes in the code means fewer setups, tighter true position on angled features, and better surface finish on complex geometry. It also means programming time goes up and the cost of a mistake goes up with it. A part with a single datum and 20 angled holes is a 5-axis job. A flat bracket is not.

• 1
  Simultaneous 5-axisAll five axes move together. Required for contoured surfaces and undercuts.
• 2
  3+2 positioningRotary axes index, then cut in 3 axes. Simpler code, still fewer setups.
• 3
  Singularity and collisionThe two failure modes a 5-axis post must catch before the run.

Feed and speed are the two words with the most visible consequences. Push feed too high and the tool deflects, leaving chatter and a Ra that misses the callout. Push it too low and the tool rubs instead of cutting, which work-hardens stainless and burns the edge. The window is narrower on titanium and Inconel than on 6061 aluminium, which is why a shop's material experience matters more than its machine list.

Tool selection is encoded in the T words and the CAM tool library. A Ø12 mm end mill with a 4-flute geometry behaves differently from a 3-flute aluminum-specific cutter at the same feed. Chip evacuation changes, and so does the risk of recutting chips in a deep pocket. On a 4,000 mm part with deep pockets, the tool sequence in the program decides whether the job runs in one shift or three.

Offsets matter for tolerance. Work offsets set the part origin. Tool length offsets set the Z reference for each tool. Cutter compensation lets the operator adjust for tool wear without reprogramming. If a shop runs 100% inspection before shipment, as we do, the offset log is part of what the inspector checks against. A ±0.005 mm band leaves little room for a stale offset.

Surface finish callouts map to specific program decisions. Ra 1.6–3.2 μm is a normal as-machined result. Ra 0.8–1.6 μm usually needs a finishing pass with a smaller stepover. Ra 0.2–0.8 μm typically means a separate finishing strategy or a secondary operation. Telling your machinist the finish target in Ra, not in adjectives, is the fastest way to get a program that hits it.

• 1
  ChatterUsually feed or rigidity, not the code syntax. Check tool overhang first.
• 2
  Work hardeningToo low a feed on stainless or titanium. Raise feed, keep the tool cutting.
• 3
  Offset driftA worn tool with an unadjusted offset will walk out of a ±0.005 mm band.

Code cannot fix a bad design. A 0.5 mm internal corner radius that needs a Ø0.5 mm cutter will run, but the tool will deflect and the finish will be poor. A deep, narrow slot will force a long-reach tool and the same problem. These are geometry issues that a better program cannot solve. A DFM review catches them before the program is written.

Code cannot create precision the machine does not have. A worn ballscrew or a spindle with runout will put error into every line, no matter how clean the G-code looks. That is why machine maintenance schedules and calibration records belong in a supplier audit. We hold ±0.005 mm on production parts, and that number depends on machine condition as much as on programming.

Code cannot inspect the part. A program can drive a probe and log the result, but it cannot decide whether a feature is functionally acceptable. That is a metrology question. We inspect 100% of parts before shipment and run raw material checks, in-process monitoring and final inspection, with reports available on request.

The honest summary: G-code is a precise but narrow instruction set. It controls motion and machine state. Everything above that layer, from material choice to fixture design to inspection strategy, is where the outcome is actually decided.

• 1
  Tool reachDeep pockets and small internal radii need long-reach tools, which deflect more.
• 2
  Machine conditionBacklash and runout show up in the part regardless of the program.
• 3
  InspectionProbing collects data. Judging acceptance is a separate step.