How to Match the Knife in CNC Cars: 7 Steps for Accurate Tool Setting
Tool setting decides whether a program runs clean or scraps the first part. This guide covers how to match the knife in CNC cars: zero reference, offsets, test cuts, and the checks we run on 127 machines before a batch is released. Written for machinists and process engineers who need a repeatable setup, not a list of tips.

In this article
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
Key takeaways
What matching the knife in CNC cars actually means
Matching the knife in CNC cars is the setup work that ties a physical tool tip to a coordinate value the control can trust. The machine does not know where the edge sits. You tell it, and every later move depends on that number. Get it wrong by 0.02 mm and a Ø12 mm bore becomes a rework ticket.
The job has three parts. First, establish the machine zero so the control has a fixed reference. Second, measure the distance from that reference to each tool tip and write it into the offset table. Third, build the part coordinate system so the program origin and the workpiece origin are the same point in space.
People sometimes call this tool adjustment or tool alignment. The name changes by shop. The sequence does not. Zero, measure, compensate. Skip one and the other two are wasted effort.
This matters most on short runs, where tool setting eats a large share of the cycle. On a 5-part order, a 40-minute setup is the real cost driver, not the 6 minutes of cutting. A repeatable procedure is the only way to keep that number down.
- 1Machine zeroA fixed reference on the machine, set once per setup or after a crash.
- 2Tool offsetThe stored distance between machine zero and each tool tip.
- 3Part zeroThe program origin placed on the workpiece, usually a corner or a bore center.
What you need before the first tool touches metal
Clean the spindle taper and every tool holder. A chip on the taper is worth 5–15 μm of runout at the tip. On a Ø10 mm end mill, that shows up as a tapered wall and a poor surface finish. Wipe the holder with a lint-free cloth and check it under light before you load it.
Confirm the program origin on the drawing. Face-milled stock usually takes Z0 at the finished face and X0/Y0 at a corner or a bore center. Write the choice on the setup sheet. If the operator and the programmer pick different corners, the first part will be symmetrical and wrong.
Know the material. Aluminum 6061 cuts soft and springs back little. 316L stainless work-hardens, so light test cuts can rub instead of cut. Titanium Ti-6Al-4V needs a stiffer setup and slower test feeds. Match your test parameters to the material, not to habit.
Have the measuring tools out before you start. A 0–150 mm caliper for outside circles, an edge finder or a 3D probe for part zero, and a dial indicator for runout checks. A probe is faster and removes reading error, but a caliper still works on older controls.
- 1Clean taperWipe the spindle and holder. Chips are the most common cause of runout.
- 2Setup sheetRecord X0, Y0, Z0 and the datum choice so the next shift repeats it.
- 3Right toolsCaliper, edge finder or probe, indicator. Nothing else is strictly required.
Zero reference and the coordinate system behind it
Power up the machine and return it to the reference point. Most controls call this zero return or home. This sets a known position in machine coordinates. Until it is done, the control has no idea where the table is, and any offset you enter sits on sand.
Once the machine is homed, the distance between the tool position and the machine origin is fixed and repeatable. That fixed distance is the base for every tool offset you will write. If the machine loses home mid-shift, all offsets become suspect and the setup has to be re-verified.
Next, define the part coordinate system. Touch off X and Y with an edge finder at 400–600 rpm, or use a probe. For a bore center, sweep the bore with an indicator and split the reading. On a Ø400 mm rotary table, re-check the center after any indexing move.
Record the origin in the offset page and, more important, on paper. Controls lose power. Batteries die. A written X0/Y0/Z0 and the datum type take two minutes and save a full re-setup.
- 1Home the machineReturn to reference before entering any offset value.
- 2Touch off X and YEdge finder at 400–600 rpm, or probe the corner and the two edges.
- 3Write it downPaper record of X0, Y0, Z0 and the datum type.
Measuring tool length and radius without guessing
Tool length is measured from the machine reference to the tip, or to the gauge line, depending on how the shop sets up. Pick one method and stay with it. Mixed methods are the fastest route to a Z crash.
The test cut method still works well. In manual or MDI mode, bring the tool down in Z and take a light cut on the top face. Hold Z, record the machine Z reading, and enter it as the tool length offset. A 0.1–0.2 mm depth of cut is enough. Deeper cuts load the tool and change the reading.
For radius, test-cut an outside circle. Keep X steady, feed in Z, and cut a short cylindrical land. Measure the diameter with a caliper and compare it with the programmed value. The difference divided by two is your radius correction. Enter it into the wear offset and re-cut.
A 3D probe or a tool presetter removes the caliper step and typically holds 2–5 μm repeatability. On our 16 simultaneous 5-axis centers, probing is the default for parts that carry a ±0.005 mm tolerance. Manual test cuts remain the fallback for simple 3-axis work.
- 1Light cuts only0.1–0.2 mm in Z. Heavy test cuts change the reading.
- 2Radius math(programmed Ø − measured Ø) ÷ 2 goes into the wear offset.
- 3Probe when tight2–5 μm repeatability beats hand measurement on tolerance-critical parts.
Part zero, end facing, and total length
After the round is cut, the part still needs an accurate total length in Z. Face the end at center, keep the tool motionless in Z, and record the total length along the axis. That reading becomes the Z origin for the program.
Cutting at the center rather than the edge matters. A face cut at the edge sees interrupted load and pushes the tool. At center, the load is balanced and the recorded number is stable. On thin parts, take two light passes instead of one heavy pass.
Check the result before you run the program. Move to the programmed Z0 position with the spindle stopped and bring a gauge block or a shim to the face. If the gap is uneven, the part is tilted in the vise or the vise jaw is not parallel.
For turned parts on mill-turn centers, the same logic applies to the turret tools. Set the Z offset on a faced end, then verify X with a light cut on the outside diameter. Two measurements, one offset page, no surprises.
- 1Face at centerBalanced load, stable Z reading. Avoid edge-only face cuts.
- 2Verify with a shimGauge block at Z0 should touch evenly across the face.
- 3Mill-turn toolsSet Z on a faced end, confirm X on the outside diameter.
Step by step: matching the knife in CNC cars
Run these in order. Skipping ahead is where setups go wrong.
- 1Home the machinePower up and return all axes to the reference point. Confirm the home lamps or the position readout before you continue.
- 2Clean and load the first toolWipe the spindle taper and the holder. Load tool 1 and check runout with an indicator. Keep runout under 5 μm at the tip for finishing tools.
- 3Touch off X and YUse an edge finder at 400–600 rpm or a 3D probe. Locate the two edges and the corner, then enter X0 and Y0 in the work offset page.
- 4Set Z with a test cutIn MDI or manual mode, take a 0.1–0.2 mm cut on the top face. Hold Z and record the machine reading as the Z origin.
- 5Enter the tool length offsetWrite the recorded Z value into the offset table for tool 1. Repeat for every tool in the program before cutting the first part.
- 6Test-cut the outside circleCut a short cylindrical land with X steady. Measure the diameter with a caliper and halve the error into the radius wear offset.
- 7Face the end and set total lengthFace at center, hold Z, and record the total length along the axis. Use that as the Z origin for the part.
- 8Re-check the first toolAfter the last tool is set, come back to tool 1 and re-verify runout and length. Thermal drift over a long setup shows up here.
Tool setting methods compared
Pick the method that matches your tolerance, batch size, and machine.
| Method | Typical accuracy | Best for | Main risk |
|---|---|---|---|
| Manual test cut | ±0.02–0.05 mm | One-offs, simple 3-axis work | Operator reading error |
| Edge finder plus caliper | ±0.01–0.02 mm | Small batches on older controls | Chip on the taper |
| 3D probe on machine | ±0.002–0.005 mm | Tight-tolerance, 5-axis parts | Probe stylus wear |
| Offline tool presetter | ±0.002–0.005 mm | High-volume, quick changeover | Transfer error to machine |
| Automatic tool setting | ±0.002–0.005 mm | Lights-out and repeat runs | Setup and calibration cost |
Frequently asked questions
How often should tool offsets be re-checked?
Re-check at the start of every shift and after any crash, tool change, or spindle warm-up cycle. On long runs, verify the first tool again at the mid-point of the batch.
Thermal growth over a 4-hour run can move Z by 10–20 μm on an aluminum job. A two-minute re-check is cheaper than a rework batch.
Can I set tools without a probe or presetter?
Yes. A test cut, a caliper, and an edge finder handle most 3-axis work and hold ±0.01–0.02 mm in steady hands.
The limit shows up on 5-axis parts and on tolerances at ±0.005 mm, where hand measurement error consumes most of the tolerance band.
Why does the first part run oversize after a correct setup?
The usual cause is a radius wear offset that was entered in diameter terms, or a tool that was measured cold and then grew when it warmed up.
Check the offset sign first, then re-measure the part at cutting temperature. Aluminum moves a few microns between cold and warm states.
Do I need to re-home the machine after a power dip?
Yes. Any loss of position means the machine origin is no longer trusted, and every offset built on it is suspect.
Re-home, then re-verify X0, Y0, and the first tool Z before restarting the program.
What runout should I accept at the tool tip?
For finishing tools, keep runout under 5 μm. For roughing tools, 10–15 μm is usually acceptable.
Above those numbers you get uneven flute loading, poor surface finish, and shorter tool life. Check the holder before you blame the tool.
How does tool setting change on 5-axis work?
The tool tip must be defined in the rotary center as well, not just in X, Y, and Z. Any error in the rotary pivot distance shows up as a taper on the walls.
We verify pivot distance on the 16 simultaneous 5-axis centers before a tolerance-critical job runs.
Send us your drawing and setup notes
We review the print, flag the tolerance-critical features, and return a quote with a DFM analysis within 12 hours.
12-hour quote±0.005 mm tolerance100% inspection