Is RSS always less conservative than worst case?

For independent positive tolerance magnitudes, RSS is no larger than their arithmetic sum. That does not make it valid automatically; its statistical assumptions still require evidence.

Should nominal dimensions be negative in a stack?

Use a negative sign when a dimension acts opposite to the chosen closure direction. Keep the entered tolerance magnitude positive.

Can RSS guarantee every assembly will pass?

No. A simple RSS result estimates combined variation under statistical assumptions; it does not guarantee every permitted extreme combination.

What if contributors are correlated?

A simple root-sum-square model assumes independence. Correlated contributors require covariance terms or another model supported by process data.

Worst-Case vs RSS Tolerance Stack-up — When to Use Each Method

Build a signed dimension chain first

A tolerance stack is a one-dimensional closure equation. Dimensions that add available space are positive; dimensions that consume that space are negative. Keep the sign on each nominal dimension, but enter every bilateral tolerance as a positive magnitude.

For a gap made from 20.0 ±0.1 mm minus 8.0 ±0.2 mm, the nominal closure is 12.0 mm. The sign convention makes the physical question visible before any statistical method is chosen.

Worst case: guarantee against stated limits

Worst-case stack-up adds the absolute tolerance of every contributor. In the example, ±0.1 and ±0.2 become ±0.3 mm, so the assembly range is 11.7 to 12.3 mm.

Use worst case when the requirement must hold for every permitted combination, when production distributions are unknown, or when the consequence of failure does not justify a statistical assumption. It is conservative because it assumes all contributors reach their most unfavourable limit together.

RSS: an estimate with statistical conditions

Root-sum-square combines independent contributors as the square root of the sum of their squared tolerances. The same example gives √(0.1² + 0.2²) = 0.224 mm, producing an estimated 11.776 to 12.224 mm band when the entered tolerances are treated consistently.

RSS is not a weaker version of worst case. It is a different model. Use it only when contributors are reasonably centred, stable and independent, and when the meaning of each entered tolerance relative to process variation has been defined. Correlation, drift or mixed confidence levels can make a simple RSS result misleading.

Which method should decide pass or fail?

Use worst case as the default design screen. If worst case fails but the design is commercially important, gather process evidence before using a statistical allocation. That review may justify RSS, Monte Carlo simulation or a redesigned tolerance scheme, but the assumptions must travel with the result.

Use worst case for guaranteed interchangeability and unknown distributions.
Use RSS for a supported estimate of typical variation, not an automatic drawing guarantee.
Investigate sensitivity: the largest contributors offer the greatest redesign opportunity.
Include assembly, thermal, geometry and measurement effects when they materially affect closure.

Worked requirement check

Suppose the acceptable gap is 11.8 to 12.2 mm. The nominal 12.0 mm is centred, but the worst-case 11.7 to 12.3 mm range fails by 0.1 mm on both sides. The simple RSS band is narrower, yet it also extends beyond the requirement in this example. Dimvo therefore keeps worst case as its pass/fail basis and shows RSS as a separate estimate.

A passing calculation still depends on correct contributors and signs. Check the drawing loop, datum scheme, unilateral tolerances, geometric effects and assembly sequence before release.

Worst-case vs RSS tolerance stack-up