Reading a tolerance stack before it reads you

One gap, many contributors

A lid that rattles or a button that sticks is rarely one bad part. It is usually the sum of several good ones. Each component sits comfortably inside its own tolerance, and together they eat the clearance the design depended on.

A stack-up makes that sum visible. You list every dimension in the chain between two surfaces, note its nominal value and its tolerance, and add them in the direction they act.

Worst case or RSS?

Worst case assumes every part arrives at the limit that hurts you most, all at once. It is safe and often pessimistic. Root sum square treats the variations as independent and adds them statistically, which reflects real production better once volumes grow.

• Worst case for safety-critical fits and very low volumes

• RSS when there are more than four contributors and stable suppliers

• Report both, and say which one the design is built on

What we change when it fails

When the gap can close, we rarely tighten every tolerance. We look for the biggest contributor and take it out of the chain: a locating feature on the housing, a datum moved to the part that matters, or a compliant feature that absorbs the variation.

The result is a design that works with ordinary tolerances, which is what keeps part prices and scrap rates sensible.