Reference temperature
20°C
ISO 1 standard
Min points — circle
3
Recommend 8+
Min points — plane
3
Recommend 9+
Min points — cylinder
5
Recommend 12+
3-2-1 alignment
6 DOF
Primary · Secondary · Tertiary
Type 01
Bridge CMM
The most common type. A rigid bridge straddles the granite table. The Z-ram carries the probe head and moves vertically. High accuracy and repeatability, suited to most dimensional inspection work in a temperature-controlled lab.
Type 02
Gantry CMM
An overhead rail system running on floor-mounted tracks. Built for very large or very heavy workpieces that cannot be lifted onto a table. Common in aerospace (engine cases, fuselage sections) and large automotive tooling.
Type 03
Horizontal Arm
The probe arm extends horizontally from a vertical column and moves in X. Ideal for large sheet metal parts, automotive body panels, and body-in-white inspection where the feature of interest faces sideways rather than upward.
Type 04
Portable Arm
A manually-operated articulated arm (e.g. FARO, Hexagon Romer). Each joint has an encoder that tracks angle. Highly flexible — can be used on the shop floor, directly on the machine, or in the field. Less accurate than fixed CMMs.
Touch Trigger Probes
When the stylus contacts the surface, an electrical signal triggers the controller to latch the current (X,Y,Z) machine position. The machine retracts and moves to the next point. Fast, reliable, and the standard for most dimensional inspection work.
Typical use cases
- — Diameter, distance, true position, angle
- — Datum setup and part alignment
- — High-throughput production inspection
- — Any feature where discrete point data is sufficient
Typical repeatability: 0.3–1.0 µm unidirectional; 1–2 µm 2D (after lobing correction). Probe examples: Renishaw TP20, TP200, OMP40.
Scanning Probes
The probe maintains continuous contact with the surface while moving along a defined path, collecting hundreds or thousands of points per second. A spring mechanism applies a controlled contact force; deflection is measured to determine the surface position. Gives far higher data density than touch trigger.
Typical use cases
- — Form measurement: flatness, circularity, cylindricity
- — Freeform surface inspection against CAD
- — Profile of a surface GD&T evaluation
- — Reverse engineering and first-article surface mapping
Probe examples: Renishaw REVO (5-axis scanning), Zeiss VAST, Hexagon HP-S. Scanning is slower to set up than touch trigger but captures form errors that discrete hits miss entirely.
Tip Materials
Ruby (Al₂O₃)
Standard material for most applications. High hardness, excellent sphericity, low wear rate. Avoid on aluminium — the affinity between ruby and aluminium causes pick-up (aluminium coating the ball), which inflates the effective tip diameter and produces false readings. Use silicon nitride instead for aluminium parts.
Silicon Nitride (Si₃N₄)
Recommended for aluminium and soft non-ferrous metals. Lower chemical affinity for aluminium than ruby, so pick-up is far less likely. Slightly less hard than ruby but still excellent for most general work.
Zirconia (ZrO₂)
Good wear and chemical resistance. Less common than ruby or Si₃N₄. Used in applications requiring thermal or chemical stability.
L:D Ratio Rule
Keep the stylus length-to-diameter ratio below 5:1 wherever possible. A long, thin stylus acts as a flexible beam — vibration and contact force cause bending that appears directly as measurement error. If a long stylus is unavoidable, reduce probing speed and apply additional pretravel correction.
Stylus Types
| Type | Best for | Notes |
|---|---|---|
| Ball | General purpose — planes, circles, cylinders, bores | Default choice. Well-characterised geometry, easy to qualify. Available from 0.3 mm to 12 mm diameter. |
| Disc | Undercuts, grooves, O-ring grooves, slots | Flat disc can reach geometry a ball cannot. The contact diameter must be factored into the measurement. |
| Cylinder | Blind bores, internal threads, narrow slots | Provides a defined contact line rather than a point. Approach direction is critical. |
| Star | Complex parts requiring access from multiple directions | Multiple ball tips on one shank. Qualify once, access multiple directions without changing probe. Reduces re-qualification time. |
| Hemisphere | Large-radius form measurement, surface mapping | Large contact area averages surface roughness. Better for smooth, large-radius surfaces. |
Probe Qualification
Before measuring, every probe configuration must be qualified against a certified reference sphere. Qualification determines the effective ball diameter (accounting for Hertzian deformation at contact) and the precise ball-centre offset relative to the machine datum. This corrects for tip runout, stylus deflection, and probe offsets.
Re-qualify after: any probe crash, changing styli, starting a new measurement session, or any significant temperature change. Some labs qualify at the start and end of every batch for traceability.
The 3-2-1 alignment constrains all six degrees of freedom (3 translation + 3 rotation) by referencing three datum features in a priority hierarchy. This directly mirrors the datum reference frame (DRF) defined on the engineering drawing, ensuring measurements are made in the same coordinate system the tolerances were designed in.
Primary Datum — 3 Points minimum
Constrains 3 DOF. Establishes the primary axis and locks the primary reference plane (e.g. Z=0).
Typically the largest, most stable surface. 3 non-collinear points are the mathematical minimum; use 9+ on a large surface. Points must be spread as far apart as possible to maximise angular sensitivity.
Secondary Datum — 2 Points minimum
Constrains 2 more DOF. Establishes the secondary axis and locks rotation about the primary axis.
Often a long face, edge, or bore axis. Minimum 2 points. Maximise the distance between them — a short secondary datum will amplify any angular error in the alignment.
Tertiary Datum — 1 Point minimum
Constrains the final DOF. Locks translation along the remaining axis (e.g. X=0).
A bore centre, a boss, or a single-point contact. Its position relative to the secondary datum must match the drawing's intent — this is what locates the part in the third direction.
Other Alignment Methods
Iterative / Best-Fit
Minimises the sum of squared distances between measured points and nominal CAD positions. Useful for castings, forgings, and freeform surfaces where no clean datum features exist. Not recommended when drawing datums are well-defined — best-fit can hide real geometric errors by distributing them across all features equally.
Iterative with Constraints
A best-fit that is constrained to honour specific datum priorities. Combines the flexibility of best-fit for freeform surfaces with the traceability of datum-based alignment. Used in aerospace forging inspection per AS9102.
Two-Touch Rough Alignment
A fast preliminary alignment taken from just a few points before the part is properly located. Used to safely approach the full datum features without collision risk. Always follow with a full 3-2-1 alignment before measuring.
Alignment Residuals
After alignment, check the reported residuals for each datum hit. Large residuals mean the part isn't sitting correctly on its datums, the fixture is unstable, or the datum feature itself has poor geometry. If multiple features fail by the same amount in the same direction, suspect the datum alignment before condemning the part.
The mathematical minimum defines the geometry but gives no information about form error. Always use significantly more points in practice — additional data reduces the effect of surface roughness, stylus lobing, and local form deviations on the fitted result.
| Feature | Min pts | Recommended | Point distribution | Notes |
|---|---|---|---|---|
| Point | 1 | 1 | Single perpendicular hit | No fitting — raw hit coordinates only. Useful for constructed features only. |
| Line (2D) | 2 | 5–8 | Evenly spaced along the full length; include points near both endpoints | More points capture straightness deviation in the fitted line rather than just the ends |
| Plane | 3 | 9–16 | Square or triangular grid; avoid collinear patterns; include near-edge points | 3-point minimum gives no form information. 9-point 3×3 grid is standard for a datum face. |
| Circle | 3 | 8–12 | Equal angular spacing (45° increments for 8 pts). Never use 4 hits at 0/90/180/270° only | 90°-only patterns can completely miss a 4-lobed oval error. At least 8 hits at 45° spacing catches it. |
| Arc | 3 | 5–8 | Distributed across the full arc span including near both endpoints | Arc must subtend >60° for numerically stable fitting. Short arcs are highly sensitive to point placement. |
| Sphere | 4 | 9–16 | Multiple latitudes: e.g. 3 equatorial + 3 mid-latitude + 1 pole. Never equatorial only. | Equatorial-only hits fit a circle, not a sphere. Include at least one axial or polar hit. |
| Cylinder | 5 | 12–20 | Minimum 2 cross-sections spread axially; 6–8 equally spaced hits per cross-section | A single cross-section fits a circle only — axis direction is undefined. Long bores need 3+ sections. |
| Cone | 6 | 12–16 | Minimum 2 sections at different heights; equal angular spacing per section; maximise axial spread | Cone half-angle is very sensitive to axial separation of cross-sections — spread them as wide as accessible. |
| Torus | 7 | 20+ | Multiple meridional and equatorial sections | Rarely fitted directly. Usually measured as a swept profile or surface deviation from CAD. |
| Slot | 4 | 10–14 | 2–3 pts per long wall + 3 pts per semicircular end | Width from wall-to-wall; length from arc centre to arc centre. Check the software's slot-fitting algorithm. |
Approach & Retract
The probe must approach along the surface normal — perpendicular to the expected surface. A tangential approach causes the stylus to ride along the surface before triggering, recording an incorrect position.
Search distance is used when the surface location relative to nominal is uncertain. The probe moves until it detects contact rather than expecting the surface at exactly the nominal position.
Probing Speed
Higher probing speed increases throughput but reduces accuracy. The probe triggers at a slightly late position due to controller response time and stylus bending under contact force. The faster the approach, the larger the dynamic error.
CMM manufacturers specify rated accuracy at a defined probing speed. Match your speed to the tolerance budget. If the feature has a tight tolerance (under 0.05 mm), always slow down.
Common Mistakes
Single cross-section on a cylinder
All points in one axial plane fit a circle, not a cylinder. The axis direction is undefined. Always spread points axially.
4-hit circle at 90° increments
Completely misses a 4-lobed oval aligned at 45°. Use 8 equally-spaced hits minimum.
Measuring near edges
Points within 0.3–0.5 mm of an edge are affected by burrs and chamfers. Keep clear unless the edge is the feature.
Skipping re-qualification after a crash
A crash can shift the stylus, chip the ball, or change the probe offset. Always re-qualify and re-run the last feature.
Measuring a warm part
Parts fresh from machining carry heat. Thermal gradients distort geometry. Always allow adequate soak time.
Probe Qualification
Measure a Circle
Measure a Cylinder
3-2-1 Alignment
True Position Dimension
Flatness & Straightness
Syntax Notes
- —
$$— comment line (ignored by the controller) - —
PTMEAS/CART, X, Y, Z, I, J, K— target point XYZ + approach vector IJK - — The approach vector (IJK) is the outward surface normal, pointing away from the part
- —
IN= inches,MM= millimetres (set globally in programme settings) - —
BONUS/0= RFS evaluation (no MMC/LMC bonus applied) - —
IN,0at end of FEAT line = inside feature (bore);OUT,0= outside (boss)
CMM software evaluates GD&T characteristics by fitting mathematical shapes to the measured point cloud and computing the specified tolerance criterion. The fitting algorithm used significantly affects the reported value — understand what your software is actually computing for each characteristic.
| Characteristic | Fitting algorithm | What is actually computed | Practical notes |
|---|---|---|---|
| Flatness | Minimum zone (Chebyshev) | Separation of two parallel planes that just contain all measured points | Use a 3×3 grid minimum. Least-squares fit reports a smaller (optimistic) value than true minimum zone — ASME Y14.5 requires minimum zone. |
| Straightness | Minimum zone | Smallest separation of parallel lines (or planes for axis straightness) enclosing all points | For axis straightness on a cylinder, extract the axis from multiple cross-sections; do not measure a 2D profile line only. |
| Circularity | Minimum radial zone (MZC) | Smallest radial separation between two concentric circles enclosing all points in the cross-section | Touch trigger with 8 hits is marginal — scanning gives far more reliable circularity values. High lobing may be missed entirely with few points. |
| Cylindricity | Minimum zone cylinder | Smallest radial separation between two coaxial cylinders enclosing all measured points | Requires multiple axial sections and high angular density. Always more than or equal to the circularity at any single cross-section. |
| True Position | Least squares centre → offset × 2 | ΔX, ΔY from nominal in the DRF. TP = 2√(ΔX² + ΔY²). With MMC bonus: tolerance increases by size deviation from MMC. | Must be evaluated in the correct datum reference frame. Verify the DRF setup matches the drawing before trusting the result. |
| Perpendicularity | Angular deviation → linear zone | Angle error between measured axis/surface and datum axis/plane, expressed as a linear tolerance zone width | Reported as a zone width in mm, not an angle in degrees. Feature length matters: a 0.1° error on a 100 mm feature = 0.17 mm deviation. |
| Parallelism | Distance between parallel planes | Separation of the datum plane and a parallel plane that just contains all surface points | The datum plane error adds directly to the measured parallelism. Always use more points on the datum face than the controlled face. |
| Circular Runout | Range of radial variation per cross-section | At each axial position: (max radial distance from datum axis) − (min radial distance) across one full revolution | Must be measured relative to the datum axis — not the feature's own fitted axis. The datum axis must be well-defined from real datum features. |
| Total Runout | Range across entire surface | Combined effect of form, taper, and coaxiality — range of all radial readings across the complete surface relative to datum axis | Always ≥ circular runout at any single section. Requires dense scanning across the full axial length for a reliable result. |
Reference Temperature
20°C
ISO 1 defines 20°C (68°F) as the international reference temperature for all dimensional measurement. Measurements at other temperatures must be corrected, or both the part and the CMM must be at the same stable temperature.
Even a 1°C difference between a steel part and 20°C introduces 11.7 µm of error per metre per degree. For a 300 mm steel shaft measured at 21°C: 0.3 m × 11.7 µm/m·°C × 1°C = 3.5 µm error — significant for tight tolerances.
Thermal Soak Time
Parts must reach thermal equilibrium with the measurement environment before measuring. Cold parts brought from storage expand as they warm, distorting geometry throughout the measurement run.
These are guidelines. Complex cross-sections and large thermal masses take longer. Parts fresh from a grinder or lathe carry process heat — wait for them to stabilise before measuring, not just soak.
Thermal Expansion Coefficients
Granite's low CTE is why CMM tables are made from it — the machine structure stays stable even with moderate temperature variation. Software temperature compensation corrects measured dimensions back to 20°C using the part's CTE and a measured temperature reading.
Before Measuring
-
01
Warm up the CMM — air bearings, motors, and controller electronics need 30–60 min to reach thermal stability
-
02
Qualify the probe. Inspect the reference sphere — remove any contamination before qualification
-
03
Verify the part has thermally soaked. Parts from a lathe, grinder, or cold store need time to stabilise
-
04
Clean the part. Coolant, chips, and oil cause probe pick-up errors and contaminate the reference sphere
-
05
Review the drawing before starting. Confirm all datums, tolerances, and any special inspection notes
-
06
Check fixture stability. Datum surfaces must contact fixture correctly — probe them to verify before measuring controlled features
During Measurement
-
07
Check alignment residuals before measuring controlled features. High residuals invalidate all downstream results
-
08
Do not lean on the CMM table or the part. Vibration and applied force distort measurements in real time
-
09
On new programmes, run slowly and keep a hand near the E-stop. Collisions damage the probe and can shift the part
-
10
After any probe crash: stop, re-qualify the probe, and re-run the last feature before continuing
-
11
If the part is moved or re-fixtured mid-run, redo the full alignment before continuing — the coordinate frame has changed
Reviewing Results
-
12
Sanity-check critical failures before raising a rejection. Re-measure the feature manually before acting on an unexpected result
-
13
If multiple features fail by the same offset in the same direction, suspect the datum alignment — not the part
-
14
Save the programme and results report. Traceability requires that measurements can be reproduced and audited
-
15
When a batch shows a systematic offset, look at the process — tool wear, fixture wear, datum surface condition — before adjusting the programme
-
16
Clean the probe tip and reference sphere after use. Cover the CMM when not in use to keep the table and optics free of contamination
COMING SOON
Gauge R&R Calculator
Analyse measurement system variation using a standard Gauge R&R study. Outputs %GRR, number of distinct categories, and ANOVA breakdown. Supports crossed and nested study designs.
COMING SOON
Measurement Uncertainty
Calculate combined measurement uncertainty per GUM. Input Type A and Type B uncertainty sources and output expanded uncertainty at a specified confidence level.
COMING SOON
Tolerance Stackup
Build a linear tolerance stackup chain and calculate worst-case and RSS gap or interference results with directional contribution inputs.