CNC Machining Tolerances: Which Limits Earn Their Place on a Drawing

Tolerance arguments usually start with a number and end with a price. The useful question comes earlier: which dimensions on this part actually carry the fit, and what limit does each one need before the assembly stops working. Everything else can sit under a general note.

Which tolerances genuinely need to be on the drawing

Only the ones your assembly reads.

A drawing earns its tight limits from function, not from habit. The dimensions that carry a bearing, seal a face, locate a mating part or set a rotational relationship are worth controlling individually; free surfaces, clearance holes and cosmetic features are not. Blanket tolerance blocks applied to an entire drawing move cost onto every feature without improving any of them.

The practical test is to ask what happens if a dimension drifts to the edge of its band. If the answer is that nothing changes, the limit can be relaxed to the general tolerance and the part becomes cheaper and no worse. This is also why a tolerance strategy belongs in a design review rather than in a drawing note added at the end.

How the setup, not the machine, sets the limit

Machining tolerance is limited by how a part is held as much as by the machine that cuts it. A diameter turned between centres is naturally coaxial with itself; the same diameter cut in a second clamping inherits whatever error the second location introduced. Thin walls deflect under cutting load, deep pockets force long tools that bend, and hard materials wear the edge during the pass.

Feature	What limits the tolerance	How to make it easier
Diameter turned in one setup	Machine and tool condition	Keep the feature inside one clamping
Feature cut after re-clamping	Locating accuracy of the second setup	Name the reference; design a dedicated fixture
Thin wall or tall rib	Deflection under cutting force	Thicker section, or reduce unsupported height
Deep narrow pocket	Tool length and holder clearance	Wider corners, shallower floor, relieved pocket
Rectangular pocket corner	Radius left by a round cutter	Draw the radius the tool naturally leaves

 

Reading a tolerance table without over-specifying

Published tolerance tables are reference points for what a process family can normally hold, not targets to copy onto every drawing. Their value is comparative: they show how much of the cost curve sits between a general note and a precision limit, and they show which features are inherently harder than others. Used that way, a table helps you decide where to spend.

Where an international drafting standard applies, the general tolerance note on the drawing usually governs everything not individually dimensioned, and the individual callouts override it. That structure is worth preserving: it lets a supplier concentrate effort on the features that matter and stop chasing dimensions nobody will measure. Terminology for these callouts is standardised through ASME, and dimensional verification practice is covered by the NIST Manufacturing Extension Partnership.

Geometry and material that make a tolerance harder than it looks

Two parts can carry identical callouts and behave completely differently. Aluminium moves with temperature and cuts freely; stainless work-hardens and pushes back against the tool; engineering polymers deform elastically and spring back after the cutter passes. Geometry adds its own difficulty: an interrupted cut, a very small internal radius or a feature at the end of a long overhang all reduce what can be held consistently.

Material behaviour also interacts with finishing. Plating and anodizing add thickness on external surfaces, so a tight diameter needs to state whether the limit applies before or after coating, and a thread needs allowance for whichever process it will see. Surface condition callouts follow the terminology in ASTM Committee B08.

Proving a tolerance was held

A tolerance that cannot be measured cannot be defended, so the inspection plan and the drawing should be written together. State the datum each dimension references, say whether the check applies before or after coating, and decide whether the result will be reported as a dimensional record or a capability study. Where a feature is awkward to reach, agreeing the measurement method before production avoids a dispute about the report rather than the part.

For repeated batches, the useful question shifts from whether one part passed to whether the process is stable. Two identical parts from the middle of a run tell you less than a record showing where the measurements sat within the band across the batch. That record is also what makes a later process change visible before it becomes a rejected shipment.

How to tell a tolerance problem from a fixture problem

When a feature repeatedly misses its limit, the cause is usually holding rather than cutting. Fixture problems show a pattern: the same feature drifts on every part, or the error grows towards one end of a batch. Machine or tool problems tend to be distributed differently, appearing as surface deterioration or a gradual change across a run. Distinguishing the two changes the fix, because one is solved by redesigning the workholding and the other by changing the cutting conditions.

The fastest diagnostic is to measure the feature at more than one position and compare it with the holding points. Where the deviation follows the clamp, the fixture is deforming the part during cutting and releasing it afterwards. Where the deviation follows the tool entry, the cutter is deflecting. Drawing-verification guidance is published through ASME, machinability data for the material in question through ASM International, and inspection practice through the NIST Manufacturing Extension Partnership.