CNC system control software and the work process behind a stable cut
A practical walkthrough for manufacturing engineers and buyers who need to know where the control software ends and the shop process begins. Read it to set offsets, prove a program and hand off a run that repeats on the next shift.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
What CNC system control software actually controls
The control software sits between the CAM output and the servo drives. It parses G-code, looks up tool offsets, applies cutter compensation, plans acceleration and deceleration, and sends position commands to each axis. It does not decide whether the part is correct. It only executes what the program and offset tables say.
That distinction matters when a part is out of tolerance. Engineers often blame the control when the real cause is a wrong wear offset, a tool holder with 0.03 mm runout, or a post-processor that emits an arc in the wrong plane. The control did its job. The input was wrong.
Three data sets feed every cut: the part program, the tool table, and the work offset. The program defines geometry. The tool table defines length and radius. The work offset defines where the part sits in machine coordinates. Change any one without updating the others and the first article moves.
- 1ProgramG-code from CAM, verified against the drawing revision.
- 2Tool tableLength and radius values measured, not copied from the tool label.
- 3Work offsetG54 to G59 set from a probe or edge finder, then logged.
Setting up a work process that repeats
A repeatable process starts before the first chip. Clamp the stock so the locating face is the same face used in CAM. If the program assumes the bottom face is Z zero, the vise jaw must touch that face and nothing else. Mixing datums between setup and programming is the most common cause of a scrapped first article.
Measure every tool on the machine or a presetter, then write the value into the tool table. A 12 mm end mill with 0.02 mm runout will cut an oversize slot even with a perfect program. On finishing passes below Ra 0.8 μm, tool runout shows up as chatter marks that no offset change will remove.
Log the offset values in a setup sheet. When the job returns next month, the operator needs the same G54 position, the same tool numbers and the same coolant settings. Without the sheet, the second run becomes a new setup, and the first article check repeats from zero.
- 1One datum per operationLocating face, program zero and inspection datum must match.
- 2Measure, do not assumeTool length and radius come from the machine or presetter.
- 3Write it downSetup sheets turn a one-off setup into a repeatable process.
Reading alarms and drift on the shop floor
Servo alarms, following errors and overtravel messages point to motion problems. Check the obvious first: a chip packed under a way cover, a loose coupling, or a tool that pulled out of the holder. Software rarely invents an alarm on a machine that was cutting fine an hour earlier.
Gradual drift is different. If the first ten parts measure 0.004 mm over nominal and part fifty measures 0.012 mm over, the cause is usually thermal growth or tool wear, not the control. Let the spindle warm up for 15 to 20 minutes, then re-check the first article. Add wear compensation in small steps and record each change.
Surface finish that degrades mid-run points to tool wear or chip recutting. Check the load meter and the chip shape. A change in chip color from silver to blue means heat is building, and the feed or speed needs attention before the next part.
- 1Alarm first, theory secondCheck mechanics before rewriting the program.
- 2Drift over timeThermal growth and wear, not the control software.
- 3Finish changeLook at the chip and the load meter.
What changes between one part and a 10,000 part run
A single prototype tolerates a slow setup. A production run does not. The process has to survive shift changes, tool replacements and operator turnover. That means the setup sheet carries more weight than the program itself, because the program is fixed and the people running it are not.
For runs above a few hundred parts, consider a probing cycle for work offset verification at the start of each shift. A 30 second probe pass catches a fixture that moved overnight and saves a batch of scrap. On five-axis work, verify the rotary table center once per setup, not once per batch.
Tool life management matters more than most shops admit. Track the number of parts per edge, replace on schedule, and keep a spare pre-measured tool at the machine. Waiting for a finish to degrade before changing the insert costs more than the insert.
- 1Probe per shiftCheap insurance against a moved fixture.
- 2Rotary center checkVerify once per five-axis setup.
- 3Edge countersReplace tools on count, not on feel.
Six steps to prove a program before cutting metal
- 11. Match the drawing revision to the CAM fileConfirm the program was posted from the current drawing. Check the revision letter on the setup sheet against the file name. Most wrong-part events start here, not at the machine.
- 22. Load and measure every toolMeasure length and radius on the machine or a presetter. Enter values into the tool table and check runout with a dial indicator. Keep runout under 0.01 mm for finishing tools.
- 33. Set the work offset from one datumUse a probe or edge finder on the locating face defined in CAM. Write the G54 value on the setup sheet. Confirm Z zero with a 0.01 mm shim or the probe.
- 44. Dry run with rapid override downRun the program with rapid override at 5 to 10 percent and feed hold ready. Watch the distance-to-go display. Any unexpected rapid move into the part means stop and re-check the offset.
- 55. Cut air, then cut the first articleRun once with the tool 20 mm above the stock, then cut the first part at reduced feed. Measure the features the drawing calls out before releasing the run.
- 66. Record and hand offWrite offset values, tool numbers, spindle loads and inspection results on the setup sheet. Attach it to the job. The next operator should be able to repeat the run without asking questions.
Which control feature matters for which job
Match the feature to the part, not to the brochure.
| Job type | Feature to rely on | What it buys you |
|---|---|---|
| 3-axis plate work | Work offsets and cutter comp | Fast setup, easy rework |
| 4-axis milled slots | Rotary indexing and tool table | Fewer setups, tighter position |
| 5-axis contoured surfaces | Rotary center and TCP | Correct geometry at any angle |
| High-mix low-volume | Probe cycles and setup sheets | Repeatable handoff between shifts |
| Long production runs | Tool life management and load monitoring | Predictable tool changes, less scrap |
| Thin-wall parts | Feed override and look-ahead | Less chatter on finishing passes |
The control is not the process
A clean program on a warm machine with measured tools and a logged offset will repeat. Fix the setup sheet before you rewrite the G-code.
Questions engineers ask before releasing a run
Do we need a probe to run a stable process?
No, but a probe makes the first article faster and catches a moved fixture before the run starts. Without one, the operator sets the offset with an edge finder and a shim, which works fine at ±0.02 mm and gets slow below ±0.01 mm.
For five-axis work, a probe or a calibrated test bar is the practical way to verify the rotary center each setup.
Why does the same program cut differently on two machines?
Machine geometry, thermal state and tool holders differ. A machine that has been running all day is warmer than one started ten minutes ago. Spindle growth of 0.01 to 0.02 mm over a shift is normal.
Warm up both machines for the same time, check tool runout, and compare the first article before assuming the control is at fault.
How often should offsets be re-checked?
Re-check the work offset at the start of each shift and after any fixture change. Re-check tool length after every tool change or insert replacement. Log each value on the setup sheet so drift is visible, not guessed.
Can we hold ±0.005 mm on a production run?
Yes, with the right machine, tooling and thermal control. The tolerance depends on fixture rigidity, tool runout, coolant and warm-up, not only on the control. On a 10,000 part run, plan for periodic first-article checks rather than a single check at the start.
GreatLight inspects 100 percent of parts before shipment and reports are available on request.
What belongs on a setup sheet?
Drawing revision, program name, work offset values, tool numbers with length and radius, spindle speed and feed, coolant setting, inspection points and the operator who ran the first article. One page is enough if it is complete.
When should a job move off a 3-axis machine?
When the part needs features on more than one face and the setup count starts driving cost or position error. Four-axis indexing removes several setups. Five-axis simultaneous cutting helps on contoured surfaces where a 3+2 setup would need long reach tools that chatter.
Send the drawing, get a process review
Quotation and free DFM analysis within 12 hours, from one prototype to 10,000+ part runs. No minimum order quantity, and your files stay confidential with an NDA on request.
12-hour quote100% inspectionNo MOQISO 9001 / IATF 16949
More from the shop floor
We publish setup notes, tooling trials and inspection data from the factory floor.
