CNC Detection Setup Guide: 7 Steps to a Repeatable Probe
This CNC detection setup guide is written for machinists and process engineers who need to run a spindle probe in production, not just watch it move on a demo. It covers calibration, stylus choice, skip values, drift checks and the numbers that tell you the setup is drifting. Read it before your next first-article run.
In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
Key takeaways
What a CNC detection setup actually measures
A touch-trigger probe does not measure a surface. It measures the moment an internal contact breaks, then the control adds a compensation value to that machine position. Every number you get out of the cycle is really that compensation value. Get the compensation wrong and the probe will be repeatable and wrong at the same time, which is the worst failure mode on a shop floor.
Two calibrations matter. The first finds the effective stylus ball radius, which is not the nominal ball size. The second finds the lobing, the variation in trigger point as the stylus sweeps around a circle. Both are normally handled in one ring gauge cycle that touches the ring at several angles, usually 0°, 90°, 180° and 270°, then repeats.
The ring gauge should be at least 20 °C stable and clamped to the table, not held in a vise. Its certified diameter should be within 1 μm of the value you type into the control. If you type a nominal 50.000 mm for a ring that is actually 49.998 mm, you have just built a 2 μm error into every subsequent measurement on that machine.
- 1Trigger pointWhere the internal contact opens; drifts with stylus bending and temperature.
- 2Ball radius offsetThe value added to every hit; comes from the calibration cycle.
- 3LobingDirectional variation; usually largest on long carbon stems.
Stylus and probe selection before you touch the control
Stylus choice sets the ceiling on everything else. A 4 mm ruby ball on a 50 mm carbon fiber stem is a good default for aluminum and steel parts in the 100–400 mm range. A 2 mm ball reaches into small pockets but has less contact area, so surface finish and swarf matter more. Ceramic stems resist thermal drift better than carbon but are heavier and brittle on a hard crash.
Stem length is the trade you have to make consciously. A 100 mm stem gives you reach into a deep cavity, but any side hit bends it further, and bending shows up as a shift in the trigger point. If your part needs 100 mm of reach and ±0.01 mm measurement, run the calibration cycle at the working length, never at a shorter setup length and then swap.
Keep the stylus clean. A chip sitting on the ruby ball adds its own thickness to every hit in that direction. We wipe the ball with a lint-free cloth and isopropyl alcohol at the start of each shift on jobs that hold tighter than ±0.02 mm. On castings and rough stock, blow the surface off before the probe cycle, not after.
- 1Short stem, big ballBest repeatability; use when the feature is reachable.
- 2Long stem, small ballUse only when reach demands it; recalibrate at that length.
- 3Ruby on aluminumFine. Avoid ruby on cast iron and some composites, which wear it.
Reading the numbers: what good and bad look like
Repeatability and accuracy are separate. Repeatability is the spread of five hits on one point. Accuracy is how close the average of those hits sits to the true value. A probe with 1 μm repeatability and 20 μm of error is worse than one with 4 μm repeatability and 2 μm of error, because the first one will fool you on every part.
Temperature is the quiet variable. A 10 °C swing in the shop moves a 300 mm aluminum part by roughly 70 μm and moves the machine structure too. Probing a part that came off a warm machine and sat on a cold bench will give you a number that is honest about the part and wrong about the process. Let parts equalize before final inspection if the tolerance is tighter than ±0.02 mm.
Watch the direction of the error. If every measurement in X reads high by the same amount, it is a compensation value. If the error flips sign depending on which side you touch from, it is lobing or a loose stylus. If the error grows through the shift, it is thermal. Each one has a different fix, and adjusting the wrong one costs you a day.
- 1Same-direction spreadRepeatability. Should sit at 2 μm or below.
- 2Four-direction spreadLobing. Under 5 μm is workable on a 50 mm stem.
- 3Slow drift through the shiftThermal. Check shop temperature and part soak time.
Common setup errors and how they show up
The most expensive mistake is calibrating once and trusting it for months. Probe values drift with stylus wear, crashes and temperature. A shop that recalibrates every run catches a 5 μm shift the same morning. A shop that calibrates once a quarter ships a batch before anyone notices.
The second is using a skip value copied from a CAM template. Defaults are written for a clean block on a demo machine. On a casting with 1 mm of scale, a 2 mm skip drives the probe into the part before it looks for contact, and the alarm is the good outcome. The bad outcome is a bent stylus and a scrapped setup.
The third is measuring a part that is still clamped in a way that loads it. A thin plate held by two edge clamps will read differently before and after you release it. If you need the in-process number to match the final inspection number, probe it in the same state the customer will see it, or probe it after the fixture is relaxed.
- 1Stale calibrationRecalibrate per run, after crashes and after big temperature swings.
- 2Copied skip valuesSet per feature from the drawing and the stock condition.
- 3Probing a loaded partClamp state changes the number; match the final inspection state.
Step by step: CNC detection setup guide for a spindle probe
Work through these in order. Skipping step 2 is the most common cause of a probe that repeats well but measures 15 μm off.
- 11. Clean the taper and seat the probeWipe the spindle taper and the probe shank with a lint-free cloth. Insert the probe and check runout at the stylus ball with a dial indicator: keep total indicated runout under 10 μm. Above that, the ball radius calibration will vary with spindle orientation and your measurements will follow the spindle, not the part.
- 22. Calibrate on a clamped ring gaugeClamp a certified ring gauge to the table within 200 mm of the work zone. Enter its certified diameter, not the nominal. Run the calibration cycle at four angles plus a repeat pass. Record the ball radius and lobing values the control reports. If lobing exceeds 5 μm on a 50 mm carbon stem, something is loose or bent.
- 33. Verify with a second artifactMeasure a gauge block stack or a second ring of different diameter. The error should be within 3 μm of the certified value. If it is not, recheck the gauge temperature and the entered diameter before you touch the machine parameters. Do not adjust the probe to match one artifact; that hides the problem.
- 44. Set skip and overtravel values per featureSkip value is how far the probe travels before it looks for contact; overtravel is how far it can travel after contact before alarming. For a ±0.05 mm position on a machined surface, a 5 mm skip and 2 mm overtravel is a sane start. On raw stock with 1 mm of scale, raise the skip to 8–10 mm so the probe does not alarm on a bump.
- 55. Set feedrate for the touchUse a fast approach and a slow measuring feed. Typical values are 1,000–2,000 mm/min approach and 100–200 mm/min for the final touch. The slow feed is what determines repeatability; a probe touched at 500 mm/min will scatter more than one touched at 150 mm/min.
- 66. Run a five-hit repeatability checkTouch the same point on a hardened surface five times in the same direction. The spread should be 2 μm or less on a healthy setup. Then touch the same point from four directions. The spread here is your real lobing figure, and it will be larger than the same-direction spread.
- 77. Log the values and re-check on a scheduleWrite the ball radius, lobing, date and stylus serial into the job setup sheet. Re-run the ring calibration at the start of each production run, after any crash, and after a temperature swing of more than 5 °C in the shop. Drift is gradual, so the log is what catches it.
Which detection setup fits your part
Match the part and tolerance to the setup. Do not spend a ring-gauge cycle on a job that only needs a stock check.
| Part situation | Setup to use | Skip / overtravel | Expected result |
|---|---|---|---|
| Rough stock, ±0.5 mm | Single-point touch, fast feed | 8–10 mm / 3 mm | Presence check and stock offset |
| Machined face, ±0.05 mm | Calibrated probe, slow touch | 5 mm / 2 mm | Reliable to about ±0.01 mm |
| Bore or pocket, tight tol. | Calibrated probe, 4-direction | 3 mm / 1.5 mm | Best with a 4 mm ball, short stem |
| Deep cavity, reach limited | Long carbon stem, recalibrated | 5 mm / 2 mm | Lobing rises; verify before running |
| Thin-wall or flexible part | Low touch force, slow feed | 2 mm / 1 mm | Expect deflection; probe after support |
| First article on a new job | Full ring calibration plus artifact | Per feature | Do this before any production hit |
Frequently asked questions
How often should a spindle probe be recalibrated?
At the start of each production run, after any crash or hard bump, and whenever the shop temperature moves more than 5 °C from the last calibration. On jobs held tighter than ±0.02 mm, we also recalibrate at the start of each shift.
A probe that is recalibrated often is not a sign of a bad probe. It is the cheapest way to keep a measurement you can defend.
What causes a probe to repeat well but measure the wrong size?
Almost always the compensation values: the ball radius offset or the entered ring gauge diameter. Repeatability only proves the trigger is consistent, not that the offset is right.
Check the certified ring diameter first. A ring entered as 50.000 mm when it is 49.998 mm builds a 2 μm error into every hit.
Can I use one stylus for rough stock and finished surfaces?
You can, but expect to clean it more often. Scale, cast skin and chips on the ruby ball add thickness to the hit in that direction.
On castings, blow the surface off before the cycle and wipe the ball at the start of each shift. If the job holds tighter than ±0.02 mm, keep a second stylus for finished surfaces.
How do I know if the error is thermal or mechanical?
Watch the pattern. A mechanical problem gives the same error all shift. A thermal problem grows or shrinks with time and shop temperature.
Log the reading and the shop temperature together for a few days. The correlation usually shows up fast, and it tells you whether to look at the probe or at the environment.
What touch feedrate should I use for tight tolerances?
Use a fast approach at 1,000–2,000 mm/min and a slow final touch at 100–200 mm/min. The slow touch is what sets repeatability.
Going below 100 mm/min rarely helps and adds cycle time. Going above 200 mm/min on a tight job usually shows up as scatter in the five-hit check.
Does probing replace final inspection?
No. In-machine probing is a process control tool. It catches a bad offset, a shifted fixture or a missing feature before you cut the next part.
Final dimensional sign-off still needs a CMM or a calibrated bench instrument. We run 100% inspection before shipment and provide reports on request.
Send us the drawing and the tolerance callouts
Tell us which features you plan to probe in-process and which ones need bench inspection. We will review the setup, flag the features that are hard to reach, and come back with a quotation and a DFM analysis within 12 hours.
12-hour quote100% inspectionNo minimum order quantityNDA on request
More from GreatLight
We publish setup notes, tooling trials and inspection data from the factory floor.
