Tolerance

Tolerance is the amount of variation a dimension is allowed — the band, wide or narrow, within which the actual size of a feature may fall and the part still does its job. No manufacturing process makes a part exactly its nominal size: heat, tool wear, deflection and measurement error all move the real size a little, and the drawing’s tolerance is the agreement about how much movement is acceptable. A dimension of twenty millimetres with a tolerance of plus or minus a tenth means the feature may measure anywhere from 19.9 to 20.1 millimetres, and every one of those sizes is a good part; a dimension with a tolerance of plus or minus two hundredths means the same feature must sit in a band a tenth as wide. Tolerance is the language in which the designer tells the machinist what the part must be, and it is the number that decides how the part is made, measured and priced. This entry sets out what tolerances are, why they exist and why choosing them is one of the most consequential decisions on any drawing.

Why tolerances exist

Tolerance is not a concession to clumsy manufacturing — it is the necessary condition of making parts that assemble. A shaft that must turn in a bearing cannot be exactly the hole’s size in either direction, and a piston sliding in a bore needs a controlled gap; the tolerance is what states that gap as a range, and the branch of that idea that pairs a shaft and a hole is the subject of this wiki’s entry on fits and limits. Tolerances also exist because perfect interchangeability of identical parts depends on them: if every example of a part is guaranteed to fall within its tolerance band, then any one will assemble with any other, without fitting and filing at the bench — the foundation of mass production. And tolerances exist because they are the drawing’s statement of function: the dimension that matters for assembly carries a tight tolerance, the dimension that only locates a hole carries a loose one, and the tolerance is how the designer says which surfaces the part really lives by. A part dimensioned without tolerances is not a part at all — it is an unanswered question about how much variation the design can bear.

Reading tolerance on a drawing

The tolerance is read from the dimension itself, and it appears in a few standard forms. The nominal size is the round number the designer intends — the twenty millimetres — and the tolerance is the allowed departure from it, written as plus and minus deviations. A bilateral tolerance spreads the band either side of nominal — plus or minus a tenth; a unilateral tolerance allows departure in only one direction — plus zero, minus a twentieth — used where the size must not exceed the nominal but may fall short. A limit dimension drops the nominal altogether and writes the two acceptable sizes directly, twenty point one and nineteen point nine, leaving no arithmetic to the reader. Where a single tolerance governs many ordinary dimensions, it may appear as a note in the title block — the general tolerance of plus or minus a couple of tenths that applies to every dimension without its own — and the machinist who reads the drawing knows that the general note sets the default and every explicit tolerance overrides it. The meaning of any of them is the same: a promise about the range the real feature will be allowed to occupy.

Tolerance and the process that makes it

Because tolerance states what a process must be able to hold, it dictates how the feature is made. Machining holds its everyday sizes to a few hundredths of a millimetre with ordinary care, and the entries on accuracy and repeatability describe how a capable machine holds its position; but each tightening of the band asks more of the process. A feature held to a few hundredths may be made by a standard CNC machining cut and measured with ordinary instruments; a feature held to a few micrometres asks for a stable machine, a sharp and consistent tool, controlled temperatures, a finishing operation such as reaming or grinding, and measurement equal to the task. The drawing’s tolerance therefore selects the operations: the dimension that must be true to a micron is ground or honed and inspected with a micrometer or a CMM, while the dimension that only locates a hole is drilled and accepted with a calliper. And because a single part carries many tolerances, the part is made by the union of its requirements — the geometry controlled by GD&T, the sizes by their limits, and every feature by the process its tolerance demands.

The cost of tightness

Tolerance is where function meets cost, because tightness is bought with time, equipment and scrap. The relationship is not gentle: as the tolerance band narrows, the cost of holding it climbs steeply. A tight feature may need extra operations, slower feeds and speeds, a better machine, a special tool, more frequent inspection and the first-article and in-process checks that guard it; and because a tight band has less room for the normal scatter of the process, more parts fall outside it and are reworked or scrapped. The skilled designer tolerances only what the function needs — tight where the part mates or locates, loose where it merely exists — and the machinist sees the wisdom in the parts that machine quickly and assemble perfectly. The tolerance that is tighter than the function requires is not quality, it is waste; the tolerance that is looser than the function allows is a part that fails in service. Between them lies the tolerance that says what the part must do and no more, and that is the tolerance worth machining to. Tolerance, in the end, is the meeting place of design and manufacturing — the designer’s statement of what a part must be, the machinist’s statement of what a process can hold, and the agreement between them about where the part may live. The machinist who reads a tolerance understands the function behind it and the process that will hold it; the drawing that carries honest tolerances is a drawing that can be made, and the part that meets them is a part that works.