Every process varies: fill weights differ by a few grams, delivery times by a few minutes. Statistical process control separates this common-cause, random variation from special-cause variation – a worn tool, a new batch of material, an oven running hot. Control limits are the dividing line. They sit L standard deviations either side of the process mean, and Walter Shewhart’s classic choice of L = 3 means a stable, normally distributed process stays inside the limits 99.73% of the time.
A point above the upper control limit (UCL) or below the lower control limit (LCL) is a signal to look for a special cause. So is a long run of points on one side of the centre line, which suggests the mean has shifted even if no single point is extreme; the calculator checks for eight or more in a row.
Control limits describe what the process actually does; they are not specification limits, which say what the customer wants. When you paste measurements, the calculator estimates σ either with the sample standard deviation or, as on an individuals (I-MR) chart, from the average moving range divided by d₂ = 1.128, which is less inflated by shifts. For quantities that cannot be negative, the LCL can be set to zero.
How to use the Upper Control Limit Calculator
- Input
- Process mean (x̄)
- Process standard deviation (σ)
- Measurements in time order
- Estimate σ from
- Control limit width L (number of σ)
- Lower limit
Everything is calculated in your browser, so results appear instantly and nothing you type is sent to a server.
How it works
Enter the mean and standard deviation, or paste the measurements so the tool computes them. The limits are the mean plus and minus L standard deviations (with 1σ and 2σ zone lines for reference). With data, each point is compared with the limits, the longest run on one side of the centre line is counted, and everything is drawn on a control chart.
Worked example
A filling line averages 500 g with a standard deviation of 4 g. With 3-sigma limits, UCL = 500 + 3 × 4 = 512 g and LCL = 500 − 12 = 488 g; the 2σ warning lines are 508 g and 492 g. A stable normal process stays inside 488–512 g 99.73% of the time, so a false alarm happens on average once every 370 fills.
With the 20 default measurements (mean 501.25 g, sample SD 4.25 g), the limits are 488.49 g and 514.01 g; the 514 g reading on sample 19 is close to the UCL but inside it. Using the moving-range estimate (σ̂ = 4.99 g) widens the limits to 486.27–516.23 g because that reading also creates two large jumps.
Sigma level and false alarms
| L (sigmas) | Share inside the limits | False alarm about once every |
|---|---|---|
| 1 | 68.27% | 3 points |
| 2 | 95.45% | 22 points |
| 2.5 | 98.76% | 81 points |
| 3 | 99.73% | 370 points |
| 3.5 | 99.953% | 2,149 points |
| 4 | 99.994% | 15,787 points |
Formula
UCL = x̄ + L·σ · LCL = x̄ − L·σ · Individuals chart: σ̂ = M̄R ÷ 1.128 · Share inside limits (normal process) = 2Φ(L) − 1 · ARL₀ = 1 ÷ (1 − share inside)
Worked example
Inputs — Input: Mean and standard deviation; Process mean (x̄): 500; Process standard deviation (σ): 4; Control limit width L (number of σ): 3; Lower limit: Allow negative values.
Results — Upper control limit (UCL) = 512; Lower control limit (LCL) = 488; Centre line (mean) = 500; Standard deviation used (σ) = 4; Distance between the limits = 24; Upper 2σ warning line = 508.
Key terms
- Upper control limit (UCL)
- The centre line plus L standard deviations.
- Lower control limit (LCL)
- The centre line minus L standard deviations.
- Common-cause variation
- The natural, random variation of a stable process.
- Special-cause variation
- Variation from an identifiable, unusual source.
- Average run length (ARL)
- The average number of points plotted before a signal occurs.
Frequently asked questions
How do you calculate the upper control limit?
UCL = mean + 3 × standard deviation. A filling process with mean 500 g and σ = 4 g has UCL = 512 g and LCL = 488 g.
Why are control limits set at 3 sigma?
For a normal process 99.73% of points fall within ±3σ, so a stable process gives a false alarm only about once every 370 points, yet real shifts are still caught quickly.
What is the difference between control limits and specification limits?
Control limits come from the process data and show what it naturally does; specification limits come from customers or engineers and show what is acceptable.
What does a point outside the control limits mean?
That a special cause is likely – investigate what changed at that time (operator, material, machine, measurement).
Can the lower control limit be negative?
Mathematically yes, but for quantities that cannot be negative – counts, times, weights – the LCL is set to zero. Choose "Not below zero" for that.
Why use the moving range instead of the standard deviation?
The average moving range ÷ 1.128 reflects short-term, point-to-point variation and is less inflated by shifts in the mean, so the limits stay tight enough to detect them.
How many data points do I need?
At least 20 to 25 measurements from a period when the process was running normally before trusting the limits.
What other signals are there besides points beyond the limits?
Common run rules include 8 points in a row on one side of the centre line, 6 points steadily rising or falling, and 2 of 3 points beyond 2σ on the same side.