Tolerance stack-up basics: worst case and a worked example
Method
- Draw the chain from one end of the gap to the other.
- Mark each dimension with its direction: plus if it grows the gap, minus if it shrinks it.
- Add the nominals with their signs for the nominal gap.
- Add the tolerance magnitudes for the variation.
Three features, made-up numbers
Housing depth 20.0 plus or minus 0.2, minus part A 12.0 plus or minus 0.1, minus part B 7.5 plus or minus 0.1. The nominal gap is 20.0 - 12.0 - 7.5 = 0.5. The variation is 0.2 + 0.1 + 0.1 = 0.4, so the gap lies between 0.1 and 0.9. If the gap must stay above zero, the stack passes at worst case.
Common mistakes
- Adding signs wrongly.
- Using a reference dimension as a link.
Worst case, and why it is pessimistic
Worst-case analysis assumes that every part is at its extreme in the worst direction at once. This is safe and also pessimistic: for a chain of several dimensions it is very unlikely that they all sit at their limits together. Statistical methods combine the tolerances differently, usually as a root sum of squares, and give a narrower expected spread. They rest on assumptions about how the process behaves, so they should be used only where the customer allows it and the process data supports it.
A second example: fit
A shaft is 10.00 plus 0 minus 0.05, a hole is 10.10 plus 0.05 minus 0. The smallest clearance is the smallest hole minus the largest shaft: 10.10 - 10.00 = 0.10. The largest is the largest hole minus the smallest shaft: 10.15 - 9.95 = 0.20. The clearance is therefore between 0.10 and 0.20, and the spread of 0.10 is the sum of the two tolerances, 0.05 + 0.05.
Where reference dimensions and basic dimensions fit
Reference dimensions are never links in a stack, because they carry no tolerance of their own. Basic dimensions contribute through the tolerances of the frames that control them.
What to hand an inspector
A stack-up lives with the designer. An inspector needs only the toleranced dimensions. If a result for a gap or a fit is wanted in the report, record it as its own characteristic with the limits the designer states.
Mistakes to avoid
- Adding a tolerance that is a diameter as if it were a radius.
- Forgetting that an assembly gap is a result, not an input.
Worked example from the sample part
Imagine the sample bushing assembled into a housing. The housing's bore is 70.0 with a tolerance of plus 0.03 over zero, the same as the sample's row 4 diameter, and the mating shaft is 69.9 plus or minus 0.02. The smallest clearance is the smallest bore minus the largest shaft: 70.00 minus 69.92, which is 0.08. The largest is the largest bore minus the smallest shaft: 70.03 minus 69.88, which is 0.15. The clearance is therefore between 0.08 and 0.15, and its spread of 0.07 equals the sum of the tolerances, 0.03 plus 0.04.
| No. | Requirement | Min | Max | Result | Verdict |
|---|---|---|---|---|---|
| 1 | 20.0 ±0.2 | 19.8 | 20.2 | 20.08 | IN TOL |
| 2 | 70.0 ±0.3 | 69.7 | 70.3 | 69.91 | IN TOL |
| 3 | ⌀110.0 ±0.1 | 109.9 | 110.1 | 110.04 | IN TOL |
| 4 | ⌀70.0 +0.03/0 | 70.0 | 70.03 | 70.02 | IN TOL |
| 5 | ⌀30.00 +0.02/0 | 30.00 | 30.02 | 30.01 | IN TOL |
| 6 | 4X ⌀9.0 ±0.1 THRU | 8.9 | 9.1 | 9.05 | IN TOL |
| 7 | ⌀90.0 ±0.1 B.C. | 89.9 | 90.1 | 90.03 | IN TOL |
Row 4 of the table is the 70.0 diameter used above. The shaft is an invention for the example: the sample drawing has no mating part.
One-dimensional chains
A stack-up in one direction is simple arithmetic: signed nominals for the nominal gap, summed tolerances for the variation. Draw the loop from one surface of the gap to the other, going through each part and its dimensions, and mark each dimension plus when it grows the gap and minus when it shrinks it.
Position tolerances and stack-ups
A position tolerance contributes to a stack as a diameter of variation. In a radial stack, include half of it on each side. Material condition modifiers change the allowance with the feature's size, so a stack-up that uses MMC needs the actual sizes, which is why a stack-up at design time uses the worst case. See the true position calculator.
When worst case is too strict
If a worst-case stack-up fails but the product still has to be made, the designer may use a statistical method, which gives a narrower expected spread under assumptions about the process. That is a design decision with its own data requirements, and it is not something an inspector applies on their own.
Mistakes people make
- Mixing diameters and radii in one loop.
- Using a reference dimension as a link.
- Forgetting that a general tolerance applies to a link with no printed tolerance.
- Treating a stack-up result as a measurement: it is a prediction.
- What is stack-up analysis?
- A way to find the combined effect of several tolerances.
- What is tolerance stack-up?
- It is the combined effect of the tolerances of several dimensions on one result, such as a gap or a fit. In worst-case analysis the tolerances are added, so the result varies by their sum.
- How do I calculate a worst-case tolerance stack-up?
- Draw the chain from one end of the gap to the other, sign each dimension by its direction, add the signed nominals for the nominal result and add the tolerance magnitudes for the variation.
- What is the difference between worst case and statistical stack-up?
- Worst case assumes every part is at its extreme at once and gives the widest spread. Statistical methods assume the variations combine more gently and give a narrower spread, but only if the process data supports it.
- Does an inspector need to do stack-ups?
- Usually not. The designer does the stack-up. An inspector measures the toleranced dimensions, and records any result the designer asks for as its own characteristic.