Tolerance Stack-Up Calculator
Add the dimensions of your chain, choose a distribution for each one, and get the closing dimension with worst-case and statistical (RSS) results. Free, no sign-up, and nothing leaves your device.
Sign “+” means the line makes the closing dimension larger, “−” makes it smaller. The calculated closing dimension is positive when there is a gap between the neighbouring parts and negative when they interfere. Type tolerances with their own sign, for example −0.05 and −0.01 or +0.1 and +0.05. A comma or a point both work as decimal separator.
Results
CD = closing dimension. “Variation” is the distance from CD to the limits, “extreme values” are the limits themselves. ppm = assemblies per million outside the limits. ±kσ: every block uses the limits mean ± kσ and shows how many ppm that leaves outside according to its own method. ppm outside: you give the ppm and every block finds its own limits (and how many σ they are).
Chain sketch
Dimension chain
Schematic drawing, not to scale: a longer line means a larger dimension. The loop is closed by the closing dimension (CL).
Distribution of the closing dimension
How it works
Each line is one dimension of the chain, with a nominal value and a lower and upper tolerance. The closing dimension is calculated from all lines. Four methods are shown. Worst case adds the extreme values of every line. RSS combines the standard deviations of the lines and assumes the result is normal. Monte Carlo simulates a million assemblies with random values for every line. Numerical convolution calculates the distribution of the closing dimension from the distributions of the lines (for independent lines), so it stays reliable far out in the tails, up to about 7σ. You can define the limits by a sigma level (3σ, 4σ, 5.5σ or any value), and every method then shows its own ppm outside, or by a ppm, and every method then finds the limits (in σ) that leave that ppm outside.
The standard deviation of each line comes from its distribution: Normal (set by Cp, with an optional shift of the mean), Trapezoidal (the flat top runs from b % to c % of the tolerance width) or Uniform, or Custom, where you type the mean, σ, skewness and kurtosis yourself (a flexible one-humped distribution is fitted to them; very unusual combinations are refused with a message). The boxes at the right of each line show these four numbers for the ready-made distributions. The mean of a line is the middle between its limits, so tolerances like −0.05 / −0.01 are handled correctly.
With the temperature option on, each line can also have a thermal expansion coefficient (α, in ×10⁻⁶ per °C). By default the factor 1 + α·10⁻⁶·(T − T_ref) multiplies the nominal value and also the tolerances, the mean and σ of that line; with “Mean only” the line just moves by its nominal value × α·10⁻⁶·(T − T_ref) and its tolerances stay as given. The results are then shown at a low, the reference and a high temperature, so you can see how materials that expand differently change the closing dimension.
New to tolerance stack-ups? Read the beginner tutorial with worked examples. The formulas, assumptions, limits and public sources are on the methods and sources page.
This calculator is an aid for engineers. Always check results that matter against your own methods and standards.