Separate size, accuracy and repeatability
A nominal dimension identifies the intended size of a feature. Limits specify its permitted maximum and minimum. The difference between those limits is the tolerance. A drawing marked only “3 mm” leaves important questions unanswered: must the part slide, seal, locate another component or transmit torque? Each function can justify a different limit interval.
Repeatability concerns agreement among repeated measurements under specified conditions. Accuracy concerns closeness to the reference value, although metrology treats it as a qualitative concept rather than a number by itself. An instrument can repeat a biased result very consistently. Calibration, technique and a suitable uncertainty evaluation are needed before a displayed value becomes persuasive evidence.
Calculate the fit at both extremes
Consider an illustrative shaft allowed between 2.990 and 3.000 mm, paired with a hole allowed between 3.010 and 3.020 mm. The smallest diametral clearance occurs with the smallest hole and largest shaft: 3.010 minus 3.000, or 0.010 mm. The largest occurs with the largest hole and smallest shaft: 3.020 minus 2.990, or 0.030 mm.
| Feature | Minimum | Maximum |
|---|---|---|
| Shaft diameter | 2.990 mm | 3.000 mm |
| Hole diameter | 3.010 mm | 3.020 mm |
| Diametral clearance | 0.010 mm | 0.030 mm |
These calculations assume ideal geometry for the size comparison. They do not account for a bent shaft, a tapered hole, roughness peaks, dirt or operating temperature. Radial clearance is half the diametral difference for an ideal concentric pair. Confusing these two descriptions can create a factor-of-two error before manufacturing even starts.
ISO 286 provides a system of limits and fits for linear sizes. ISO 1 establishes 20 °C as the standard reference temperature for geometrical and dimensional properties. Refer to current standards and appropriate application guidance rather than treating a fit label as a substitute for understanding load, material, speed and lubrication.
Choose a measurement that can answer the question
A caliper is useful for many general checks, but narrow jaws, contact alignment and the operator’s force can become limiting for small features. A suitable micrometer may provide better control for an external diameter, while a tiny bore may need a different method altogether. The measurement method must access the actual functional surface without distorting it.
Resolution is the smallest indicated increment. It is not the same as uncertainty. Temperature, reference standards, contact force, alignment, instrument condition and repeated readings all contribute to uncertainty. NIST’s measurement guidance and the international Guide to the Expression of Uncertainty in Measurement offer a framework for describing those contributions rather than hiding them behind extra decimal places.
Before measuring, remove loose contamination safely, allow components and instruments to stabilize where practical, and check the instrument according to its instructions. Measure a shaft at more than one axial location and orientation if taper or out-of-roundness matters. Do not average away a shape defect that will still interfere with assembly.
Size alone does not guarantee motion
Two bearing holes may each have acceptable diameters but fail to share an axis. A flat mounting face can be tilted relative to a guide. A miniature mechanism can therefore bind even when every individual size appears compliant. Functional requirements must address geometry as well as size; standards such as ISO 1101 provide a language for geometrical tolerancing.
Surface finish is another separate property. A polished shaft may be undersized, tapered or bent. Conversely, an accurate diameter can have a texture unsuitable for a seal or bearing. Specify only the finish needed by the contact, and consider the process that will produce it. Unnecessary polishing can remove critical material and turn a controlled fit into an uncontrolled one.
Technical note: why tolerance stacks grow
When several dimensions contribute to one functional gap, their variations combine. A worst-case stack considers combinations of allowed extremes. Statistical methods need defensible process assumptions and do not automatically guarantee every assembly. A simpler locating scheme can reduce the stack before any individual tolerance is tightened.
Inspect the complete assembly
Begin with clean components and a known locating scheme. Tighten fasteners in an appropriate sequence and observe whether motion changes as clamping load rises. Thin frames can distort. Bearings that rotate freely before installation may become misaligned once constrained by a housing. These are system-level effects, not necessarily defective purchased parts.
Record dimensions, instrument identification, conditions and the observed behavior. If a fit is adjusted by selective assembly or lapping, document which parts belong together. Interchangeability and a one-off matched pair are different manufacturing achievements. Both can be legitimate, but the distinction should remain visible to the next person who services the machine.
The practical goal is not the tightest possible tolerance. It is enough dimensional and geometrical control to deliver the required movement reliably, with a measurement method capable of confirming it. Precision becomes useful when it is connected to function.
Bench references
- BIPM: SI Brochure — units and prefixes.
- JCGM publications — the GUM and international vocabulary of metrology.
- NIST Physical Measurement Laboratory — measurement science resources.
- ISO 286, ISO 1101 and ISO 1 — consult current editions and scope.