Resources What's New Support Download Contact Sales: Arthur-Industrial@outlook.com
Home › Resources › How a control chart works

How a control chart works

Center line, control limits, why 3 sigma, and the difference between control limits and specification limits.

A control chart is a time-ordered plot of a process statistic with a center line and upper and lower control limits.

The parts of a chart

  • Center line (CL): the process average (or average range, proportion, and so on).
  • Upper and lower control limits (UCL, LCL): usually the center line ± 3 standard deviations of the plotted statistic.
  • Points in time order: each point is one reading or one subgroup statistic.
  • Zones A, B and C: bands at 2 to 3, 1 to 2 and 0 to 1 sigma used by the pattern rules.

Why 3 sigma?

Shewhart chose 3-sigma limits for economic reasons. A stable, normal process puts a point outside them only 0.27% of the time, so false alarms are rare (about one every 370 points), while real shifts of moderate size are still caught quickly.

Control limits are not specification limits

  • Control limits come from the process data. They are the voice of the process.
  • Specification limits come from the customer or the drawing. They are the voice of the customer.

Never draw spec limits on an Xbar chart: the chart plots averages, which vary much less than individual parts.

Within-subgroup sigma

Control limits use the short-term (within-subgroup) variation, estimated from the average range, average standard deviation or average moving range. Using the overall standard deviation would widen the limits whenever the process shifts, hiding the very signals the chart is for.

Pairs of charts

Variables charts come in pairs: one for location (Xbar or individuals) and one for spread (R, S or moving range). Read the spread chart first: if it is out of control, the location chart's limits are not reliable.

Questions and answers

Can I use the limits from the drawing tolerance instead of calculating them?

No. Limits must reflect what the process actually does. Tolerance-based limits make the chart either blind or full of false alarms.

How often should I sample?

Often enough to catch changes before they produce many bad parts, and at least as often as the process is known to change (tool changes, lots, shifts).

In RS-1: SPC tab: every chart shows its limits, zones and flagged points.

← All articles