CMM probe radius compensation converts a measured probe-ball center location into an intended part-surface point using the active qualified probe radius and an applicable direction or normal. For CNC part approval, the essential question is not whether compensation is “on,” but whether the reported point type, normal, coordinate frame, CAD revision, and evaluation method all match the drawing’s inspection intent.
Why a probe-ball center is not automatically a surface point
A tactile CMM registers a position associated with the probing system. When a spherical stylus contacts a part, the center of that sphere and the physical contact location are separated by the probe radius. Radius compensation is the software-controlled geometric step that relates the measured center to a surface point.
That distinction becomes important whenever an inspection report exports individual points, evaluates a profile, constructs a feature, or compares results from different programs. A coordinate can look credible while describing a different entity than the drawing or receiving system expects.
For a simple, well-defined surface, the conceptual relationship is straightforward: a compensated surface point is derived by offsetting from the ball center along the applicable surface direction by the active probe radius. The direction is not a decorative field. It determines where the software places the compensated point. If its origin, orientation, or sign is wrong, the reported surface result can be wrong even when the physical contact was repeatable.
This is separate from machining-tool radius compensation. CNC cutter compensation controls a machining path. CMM probe radius compensation interprets a measurement contact. One does not validate or replace the other.
It is also separate from stylus qualification. Qualification establishes the probe data that software uses; inspection compensation applies that active probe information to measured contacts. A disagreement in results may involve qualification, but changing a radius or disabling compensation merely to obtain a passing result is not a valid disposition. A metrologist should investigate the measurement definition and system condition.
Surface normals, vectors, and the meaning of IJK
A surface normal is a direction perpendicular to the intended surface at a particular location. On a flat plane, it is consistent across the plane. On a curved surface, it varies from point to point. Compensation needs a defensible direction because an offset taken along one direction is not equivalent to an offset taken along another.
In PC-DMIS documentation, the I, J, and K fields for vector-point, surface-point, and high-point features represent the supplied normal vector used for probe compensation; the documentation also states that the vector should point away from the surface and is normalized by the software. Hexagon: IJKT Surface boxes
That statement should not be generalized into “every IJK value is the contact normal.” Its meaning depends on feature type: for example, circles and cylinders use a feature centerline. Hexagon: IJKT Surface boxes
For a buyer reviewing evidence, this means a report field labeled IJK is insufficient by itself. Ask what feature was evaluated, what the vector means in that software and mode, and whether the vector is used during compensation, feature construction, or presentation.
Normal direction must belong to the approved geometry
A practical review separates four questions:
- Is the measured item a ball-center point or a compensated surface point?
- Which normal or vector was used, and what feature or nominal geometry supplied it?
- In which alignment or coordinate system is the result reported?
- What subsequent feature construction, fitting, or tolerance evaluation used the point?
These are linked but not interchangeable. A correct compensated point can still be evaluated against the wrong datum alignment. A correct CAD model can still be paired with an old part-program revision. A feature fit can produce a result that differs from raw point deviations without either result being inherently invalid.
Do not rely on an always-add or always-subtract rule
Buyers sometimes ask for a universal diameter adjustment such as “add the probe radius for an internal feature” or “subtract it for an external feature.” That rule is too crude for approval work.
The correct interpretation depends on the active probe definition, contact direction, feature type, nominal geometry, alignment, software mode, and the later evaluation method. An internal cylinder, external cylinder, angled cone, freeform surface, and a constructed feature do not create the same decision. The sign cannot be approved from a diameter label alone.
The better decision framework is to classify the result before comparing it:
- Raw center data: useful for traceability and debugging when explicitly labeled. It is not automatically a part-surface result.
- Compensated point data: intended to represent a surface location after an applicable radius offset. Approval requires the normal and active probe definition to be traceable.
- Constructed feature result: a circle, cylinder, plane, or other feature derived from data. Its reported size or location also depends on the construction method and included data.
- CAD or nominal comparison result: a deviation evaluated relative to a particular model revision, alignment, and tolerance strategy.
This classification prevents a common failure mode: comparing a raw point export from one system with a compensated profile result from another and calling their difference a machining error.
Manual scans require version- and mode-specific review
Manual scan behavior deserves special care because it can be documented differently by software version and scan mode. The PC-DMIS manual-scan guidance for the cited release says that, for the described supported hard-probe manual scans, CAD surfaces can be used with FINDNOMS to seek nominal values for measured points. When nominal data is found, the point is compensated along the found vector; otherwise it remains at ball center. The same page says that without CAD surfaces, probe compensation does not occur and data remains at ball center. Hexagon: Rules for Manual Scans in General
That is not a rule that every CMM inspection needs CAD, nor that every manual mode is uncompensated. It is a documented condition for a particular manual-scan context. A buyer should request the program version, scan mode, probe type, nominal-data method, and export labeling before applying that behavior to a project.
This distinction matters most for freeform and blended CNC surfaces. A scan may contain a mix of points where nominal matching succeeded and points that remained as ball centers. Treating the entire export as one consistent type can corrupt a downstream comparison.
Three hypothetical approval failures that look like part failures
Ball-center exports mistaken for surface data
A supplier sends point coordinates to a customer who overlays them on CAD. The customer sees an apparent offset and rejects the lot. Later review shows the export contained probe-ball centers while the receiving comparison assumed compensated surface points.
The buyer’s corrective question is not “can you shift the coordinates?” It is: “What does each exported coordinate represent, and where is that status labeled?” Preserve the original export, the compensated result if available, and the program context. Do not overwrite evidence with a manually shifted file.
A normal or reference was assigned to the wrong geometry
A shallow curved wall is measured near a blend. The point is compensated using a direction tied to an adjacent nominal face, or the report is transformed using an unintended workplane. The resulting deviation may be systematic but not correspond to the drawing’s controlled surface.
Ask for the nominal surface identifier, local normal origin, alignment sequence, and graphical or data evidence sufficient to establish which geometry controlled the result. On a curved surface, a direction that is reasonable at one location may be wrong a short distance away.
Imported CAD revision differs from the released drawing
A program executes successfully but uses a model revision that does not match the released drawing or the machined part. The CMM can then produce internally consistent results against the wrong nominal.
Require CAD revision, drawing revision, program revision, and part identification to be reviewed together. Revision control is an approval prerequisite, not clerical cleanup after a result is disputed.
Numbered approval workflow for compensated CMM evidence
- Identify the drawing revision, CAD revision where used, part identifier, and acceptance characteristic before reviewing a result.
- State whether each relevant dataset is ball-center data, compensated surface-point data, or a constructed/evaluated feature result.
- Confirm the active probe and qualified probe configuration used by the program. Record configuration identity rather than assuming the nominal stylus size from a setup sheet.
- Identify the coordinate system, datum alignment, and units used for acquisition, reporting, and comparison. Resolve any frame conversion before comparing coordinates.
- For compensated points, document the normal or direction source, its orientation, and whether it originates from CAD, nominal feature geometry, or another defined method.
- For manual scans, record software version, scan mode, probe type, whether nominal matching was used, and how unmatched points were handled or labeled.
- Identify the evaluation method after acquisition: point-to-CAD comparison, profile analysis, feature construction, fitting, or another approved method. Do not infer it from a screenshot title.
- Compare like with like. Use the same units, frame, model revision, point definition, and tolerance basis before deciding whether results agree.
- If results conflict, preserve raw files, program settings, alignment information, and exports. Escalate for metrology review rather than altering compensation, probe radius, or vector sign to force agreement.
- Approve the inspection package only when its reported result is traceable to the drawing intent and its measurement definitions are unambiguous to the receiving quality team.
Compact approval inputs for buyers
Before approving a CMM-based CNC part result, request these items in the inspection package:
- Drawing and CAD revision identifiers, plus the part and program revision.
- A clear label for raw ball-center, compensated surface-point, and constructed-feature outputs.
- Probe configuration identity and evidence of the configuration used, not an assumed probe radius.
- Alignment, datum reference frame, units, and normal/vector source for critical surface results.
- Software name, version, mode, and scan method when scans support acceptance.
- The feature construction or CAD-comparison method used after point collection.
- A disposition path for mismatches that preserves original data and assigns review to qualified metrology personnel.
This is an inspection-definition checklist, not a demand that every part receive the same report depth. The agreed documentation should follow critical features, drawing requirements, and the project’s acceptance plan. For broader planning, see SUUXIANG’s quality assurance approach.
How this changes the RFQ and supplier-quality conversation
A drawing-defined CNC project benefits from defining inspection intent before machining and final approval. If a surface is function-critical, specify the relevant datum scheme, CAD authority where applicable, required output type, and whether a point report, feature result, or surface comparison is the acceptance evidence.
SUUXIANG supports drawing-based CNC parts and inspection planning where machining, datum strategy, and the documentation needed for approval must be considered together. The useful objective is not a generic “pass” report; it is evidence that the inspected entity matches the design entity being accepted.
For a concise technical review, submit an inspection requirement with the drawing revision, 3D model when applicable, critical surfaces, datum scheme, and requested inspection output.
Buyer FAQ
Is probe radius compensation the same as probe qualification?
No. Qualification establishes the probe information used by the measuring system. Radius compensation uses the active qualified probe information and an applicable direction to interpret a contact relative to the part surface. Both need traceability, but they answer different questions.
Does every IJK vector represent the surface normal at contact?
No. Ask the metrologist to identify the feature type and explain that field in the actual software mode before using it as evidence. Hexagon: IJKT Surface boxes
Does a manual CMM scan always require CAD for compensation?
No universal rule follows from one manual-scan workflow. In the cited PC-DMIS documentation, FINDNOMS can use available CAD surfaces for the described scan context, while points without found nominal data remain ball centers. Review the actual version and mode used. Hexagon: Rules for Manual Scans in General
Why can two CMM reports disagree when both name the same feature?
They may use different point definitions, alignments, probe configurations, CAD revisions, scan modes, fitting methods, or evaluation methods. Start by comparing those definitions before attributing the difference to machining variation.
Should a team reverse a normal or edit the probe radius to make a result pass?
No. A forced change destroys the evidentiary value of the result. Preserve the original data and have a metrologist determine whether the issue is the normal, alignment, probe setup, nominal geometry, program logic, or part condition.