Learn · updated 2026-10-28
c Chart and u Chart: Control Charts for Defect Counts
A c chart plots the number of defects in equal-sized samples and a u chart plots defects per unit when the sample size varies; both take their limits from the Poisson distribution.
The choice comes down to one question: is every sample the same size? If yes, chart the raw count (c chart); if the amount inspected changes, chart the count per unit (u chart). Both rest on the Poisson distribution, so the limits come from the average alone.
c chartc̄ ± 3√c̄u chartū ± 3√(ū ÷ n), n = units in that samplec̄, ūtotal defects ÷ samples, total defects ÷ total unitsLCL0 if the formula gives a negative number
Defects or Defectives?
| You count | Same sample size | Sample size varies |
|---|---|---|
| Defects (flaws; one part can have several) | c chart | u chart |
| Defectives (bad parts; each part is good or bad) | np chart | p chart |
Worked Example: c Chart (NIST, Defects on 25 Wafers)
NIST counts the defects on each of 25 wafers. The average is 16 defects, so the limits are 16 ± 3√16 = 16 ± 12.
These are the values NIST prints: center 16, UCL 28, LCL 4. LimitLine runs this example as a regression test; see the validation page.
Worked Example: u Chart (Montgomery, Defects in Computers)
Twenty samples of 5 computers each. The average is ū = 1.93 defects per computer, so UCL = 1.93 + 3√(1.93 ÷ 5) = 3.79 and LCL = 0.07, the limits from Montgomery's center line. When the number of units changes from sample to sample, each sample gets its own limits (wider for small samples), which is why the u chart, not the c chart, is the right one for unequal samples.
Three Common Mistakes
- A c chart with different sample sizes. A bigger batch has more defects for no reason except its size; use a u chart.
- Mixing defects and defectives. Counting flaws needs c or u; counting rejected parts needs p or np.
- Huge counts with false alarms everywhere. When almost every point is out, the Poisson assumption is too narrow for your process; the Laney U′ chart widens the limits by the variation actually seen.
Sources
- NIST/SEMATECH e-Handbook, §6.3.3.1 Counts Control Charts
- D. C. Montgomery, Introduction to Statistical Quality Control, p. 318 (defects in computers), as reproduced in the IMSL JMSL 6.0 UChart documentation, which does not state the edition
The LimitLine Newsletter
About one email a month: new guides, tool updates and one practical SPC tip. Unsubscribe in one click.