Tolerance
Tolerance is the permissible deviation of a dimension from its nominal value, that is, how much a part may “play” and still count as acceptable. It is defined as the total permissible variation, the difference between the upper and the lower acceptable dimension limit, and it is always an absolute value.
Example: for a part specified as “20 ±0.1 mm”, every piece from 19.90 to 20.10 mm is within spec. These two extremes are the limits, the nominal value (20 mm) is the basic size, and the tolerance is their difference: 20.10 − 19.90 = 0.2 mm.
Depending on how it is distributed around the basic size, the tolerance is:
Bilateral: in both directions, e.g. 20 ±0.1.
Unilateral: in one direction only, e.g. 20 +0.2/−0 or 20 +0/−0.2, common in precision fits.
In practice tolerances are not set arbitrarily· they follow standardised systems (e.g. ISO 286, with codes such as H7/g6).
Tolerance is essentially a compromise: the tighter it is, the better the part functions, but the more expensive it is to make. So it is set as loose as the function allows, no tighter.
Tolerance concerns the part and the drawing (what I allow to be manufactured), not the instrument. Accuracy and uncertainty concern how well I measure. Practical rule: the instrument must be significantly more accurate than the tolerance it checks· typically the measurement uncertainty should be a small fraction of the tolerance.
In production, the quick “in/out of tolerance” check is often done with go/no-go limit gauges: the “go” side must pass, the “no-go” must not· that confirms the dimension sits between the two limits without even reading a number.
Related guides











