Quick Answer: A2, D3, and D4
A2, D3, and D4 are control chart constants used in Statistical Process Control to calculate control limits for X-bar and R charts. A2 defines the ±3σ width around the process mean on the X-bar chart: UCLX̄ = X̄̄ + A2·R̄ and LCLX̄ = X̄̄ − A2·R̄. D4 sets the R-chart upper control limit: UCLR = D4·R̄. D3 sets the R-chart lower limit: LCLR = D3·R̄, where D3 = 0 for n ≤ 6. All three constants decrease as subgroup size n increases, because larger subgroups produce more stable range estimates. The constants are derived from the expected value (d2) and standard deviation (d3) of the relative range distribution for normally distributed data.
X-bar / R Chart Control Limit Calculator
D3 = —
D4 = —
CL = —
LCL = —
CL = —
LCL = —
What Are Control Chart Constants A2, D3, and D4?
Control chart constants are fixed numerical values derived from statistical theory. They convert easily measured quantities — the average subgroup mean and the average subgroup range — into statistically valid control limits. A2, D3, and D4 are the constants associated with the X-bar/R chart pair, the most widely used control chart combination in manufacturing and quality engineering.
The three constants each serve a specific role. A2 connects the average range to the X-bar chart control limits. D4 sets the upper boundary for the R chart. D3 sets the lower boundary for the R chart. Each value changes with subgroup size n because the statistical properties of the sample range depend on how many observations make up each subgroup.
Source note: The constants in this table are based on values tabulated in Montgomery, D.C. (2020), Introduction to Statistical Quality Control (8th ed.), Wiley, and cross-referenced against the NIST/SEMATECH e-Handbook of Statistical Methods. Values given to three decimal places are consistent with ASQ and Shewhart's original work.
Control Chart Constants Table: A2, D3, D4 by Subgroup Size
The table below gives A2, D3, and D4 for subgroup sizes n = 2 through n = 25. Click any row to load those constants into the calculator above. For X-bar/S charts, see the related A3, B3, B4 table in the section below.
D3 = 0 for n = 2 through n = 6 because the lower 3-sigma limit on the range falls at or below zero. This is statistically correct, not an error. Values sourced from Montgomery (2020) Introduction to Statistical Quality Control, 8th ed., and NIST e-Handbook of Statistical Methods. Click any row to load constants into the calculator.
What Is A2? Definition, Formula, and Interpretation
A2 is the control chart constant that translates the average subgroup range into X-bar chart control limits. It is defined as A2 = 3 / (d2 × √n), where d2 is the expected value of the standardized range for subgroup size n, and the 3 represents the ±3-sigma width of statistical control limits.
X-bar Chart Formulas Using A2
X̄̄ = grand mean (average of subgroup means). R̄ = average subgroup range. A2 = constant from table for chosen n.
| n | A2 | Effect on X-bar UCL/LCL Width |
|---|---|---|
| 2 | 1.880 | Widest limits — small subgroup, high range variability |
| 5 | 0.577 | Industry standard — good balance of precision and simplicity |
| 10 | 0.308 | Tighter limits — near the X-bar/S chart crossover point |
The key relationship to understand: as n increases, A2 decreases. Larger subgroups provide a more precise estimate of the process standard deviation through the range, so the control limits can be set closer to the center line without sacrificing the ±3σ guarantee.
What Is D3? R-Chart Lower Control Limit Constant
D3 is used to compute the lower control limit on the R chart. Its formula is:
R̄ is the average subgroup range calculated from your baseline data. D3 is the constant from the table for the chosen subgroup size n.
Why D3 = 0 for n ≤ 6
D3 = 0 for subgroup sizes 2 through 6. This is not a data problem or an error — it reflects the shape of the sampling distribution of the range at small n. The range can never be negative (it is always ≥ 0), but for small subgroups the 3-sigma lower limit on the range falls below zero. Setting D3 = 0 correctly handles this: the lower control limit is simply zero, meaning any non-negative range value is acceptable on the lower side. D3 becomes positive starting at n = 7 (D3 = 0.076).
| n | D3 | LCLR Interpretation |
|---|---|---|
| 2–6 | 0 | LCL = 0 × R̄ = 0. The lower bound does not constrain R at these subgroup sizes. |
| 7 | 0.076 | The 3-sigma lower limit on R first becomes positive here. |
| 10 | 0.223 | A range falling below D3 × R̄ signals unexpectedly low variation — investigate for data entry errors or measurement problems. |
What Is D4? R-Chart Upper Control Limit Constant
D4 is the constant that sets the upper control limit on the R chart:
D4 is always greater than 1, which makes sense: the upper control limit must lie above the average range R̄. D4 decreases toward 1 as n increases because the range distribution becomes less spread out at larger subgroup sizes. For n = 2, D4 = 3.267; for n = 25, D4 = 1.459.
Practical meaning of D4
A subgroup range that exceeds UCLR = D4 × R̄ signals that the within-subgroup variation has increased beyond what the process normally produces. This is a special cause in the variation dimension — it does not automatically mean the product is out of specification, but it does indicate something has changed in the process. Common causes include tool wear, operator change, incoming material variation, or measurement issues. The R chart should always be interpreted before the X-bar chart, because out-of-control R values invalidate the X-bar control limits.
How to Use the Control Chart Constants Table
The procedure below takes you from raw data to a complete set of X-bar and R chart control limits.
Step 1 — Choose Subgroup Size (n)
Decide how many observations will make up each subgroup before data collection begins. Subgroups of n = 4 or n = 5 are the most common choice in manufacturing — they provide a reasonable sensitivity to mean shifts while keeping data collection manageable. The subgroup size must remain consistent throughout the study.
Step 2 — Look Up A2, D3, and D4
Find the row in the control chart constants table that matches your chosen n. Record all three values: A2 for the X-bar chart, and D3 and D4 for the R chart. Double-check that you are reading across the correct row — using the wrong n is the most common lookup error.
Steps 3–5 — Collect Data, Compute Means and Ranges
Collect at least 20 to 25 subgroups from a period when the process is believed to be stable. For each subgroup i: calculate the subgroup mean X̄i (average of the n observations) and the subgroup range Ri = maximum − minimum within the subgroup. Then compute the grand mean X̄̄ = average of all X̄i, and the average range R̄ = average of all Ri.
Steps 6–7 — Calculate Control Limits
Step 8 — Plot and Interpret
Plot R values on the R chart first. Check whether any R points fall above UCLR or exhibit non-random patterns (runs, trends, cycles). If the R chart shows control, the process variation is stable and the X-bar limits are valid. Then plot X̄i values on the X-bar chart using the same time order. Points outside the X-bar control limits, or non-random patterns, indicate that the process mean has shifted — triggering investigation and corrective action.
Complete Worked Example: X-bar / R Chart Calculation
Scenario: A machining operation produces shaft diameters measured in millimeters. A quality engineer collects 6 subgroups of n = 5 observations each to establish baseline control limits.
| Subgroup | Obs. 1 | Obs. 2 | Obs. 3 | Obs. 4 | Obs. 5 | X̄i | Ri |
|---|---|---|---|---|---|---|---|
| 1 | 10.22 | 10.18 | 10.30 | 10.25 | 10.20 | 10.230 | 0.12 |
| 2 | 10.28 | 10.24 | 10.19 | 10.27 | 10.22 | 10.240 | 0.09 |
| 3 | 10.15 | 10.26 | 10.21 | 10.23 | 10.20 | 10.210 | 0.11 |
| 4 | 10.30 | 10.22 | 10.26 | 10.19 | 10.28 | 10.250 | 0.11 |
| 5 | 10.20 | 10.24 | 10.18 | 10.22 | 10.21 | 10.210 | 0.06 |
| 6 | 10.25 | 10.21 | 10.27 | 10.23 | 10.19 | 10.230 | 0.08 |
| Averages | X̄̄ = 10.228 | R̄ = 0.095 | |||||
Step-by-Step Calculation (n = 5)
| Step | Action | Result |
|---|---|---|
| 1 | Look up constants for n = 5 | A2 = 0.577, D3 = 0, D4 = 2.114 |
| 2 | Grand mean | X̄̄ = (10.230 + 10.240 + 10.210 + 10.250 + 10.210 + 10.230) / 6 = 10.228 |
| 3 | Average range | R̄ = (0.12 + 0.09 + 0.11 + 0.11 + 0.06 + 0.08) / 6 = 0.095 |
| 4 | X-bar UCL | 10.228 + 0.577 × 0.095 = 10.228 + 0.055 = 10.283 |
| 5 | X-bar LCL | 10.228 − 0.577 × 0.095 = 10.228 − 0.055 = 10.173 |
| 6 | R-chart UCL | 2.114 × 0.095 = 0.201 |
| 7 | R-chart LCL | 0 × 0.095 = 0 (D3 = 0 for n = 5) |
Summary of Limits
X-bar CL = 10.228
X-bar LCL = 10.173
R CL = 0.095
R LCL = 0.000
Interpretation: All 6 subgroup ranges are below UCLR = 0.201, so the R chart shows control. All 6 subgroup means fall between 10.173 and 10.283, so the X-bar chart also shows control. The process appears to be stable during this baseline period.
Where Do A2, D3, and D4 Come From?
The constants originate from the statistical theory of the sample range. For a sample of size n drawn from a normal distribution with mean μ and standard deviation σ, the relative range W = R/σ follows a distribution whose mean is d2 and whose standard deviation is d3. These two quantities, d2 and d3, are the building blocks for all X-bar/R chart constants.
| Constant | Formula from d2 and d3 | What It Represents |
|---|---|---|
| A2 | 3 / (d2 × √n) | 3-sigma limit factor for X-bar chart using range-based σ estimate |
| D3 | max(0, 1 − 3×d3/d2) | Lower 3-sigma factor for R chart; zero when theoretical lower limit < 0 |
| D4 | 1 + 3×d3/d2 | Upper 3-sigma factor for R chart; always > 1 |
| d2 | E[W] = E[R/σ] | Expected value of relative range; used to estimate σ from R̄ as σ̂ = R̄/d2 |
| d3 | SD[W] = SD[R/σ] | Standard deviation of relative range; governs the spread of the R-chart limits |
The values of d2 and d3 for each n were computed by Tippett (1925) using numerical integration of the range distribution for normal samples. They appear in Montgomery (2020) and the NIST e-Handbook as standard reference values.
X-bar/R Chart vs X-bar/S Chart: When to Use Which
The choice between an R chart and an S chart for the variation component depends primarily on subgroup size. Both charts monitor process variability — they simply use different statistics to do so.
| Feature | X-bar/R Chart | X-bar/S Chart |
|---|---|---|
| Variation measure | Range (max − min) | Sample standard deviation s |
| Typical subgroup size | n = 2 to 10 | n > 10 preferred |
| X-bar chart constants | A2 | A3 |
| Variation chart constants | D3, D4 | B3, B4 |
| Computational ease | High — range is quick to calculate by hand | Lower — standard deviation requires more calculation |
| Statistical efficiency | Adequate for n ≤ 10 | Better for n > 10 (s uses all observations) |
| Common application | Shop-floor SPC, manual charting | Automated data collection, large subgroups |
The standard recommendation from both ASQ and Montgomery (2020) is to use the R chart when n ≤ 10 and the S chart when n > 10. For n = 10, either chart is defensible. Never mix the constants — do not apply A2 to an S-chart calculation or B3/B4 to an R-chart calculation.
Related Control Chart Constants: A3, B3, B4, c4, d2
A2, D3, and D4 are the constants for X-bar/R charts. Other chart types use a parallel set of constants. The table below shows the most frequently referenced constants across different chart types.
| Constant | Main Chart | Purpose | Limit Formula |
|---|---|---|---|
| A2 | X-bar/R Chart | X-bar chart limits using range | X̄̄ ± A2·R̄ |
| D3, D4 | R Chart | R-chart lower and upper control limits | D3·R̄ and D4·R̄ |
| A3 | X-bar/S Chart | X-bar chart limits using sample std dev | X̄̄ ± A3·s̄ |
| B3, B4 | S Chart | S-chart lower and upper control limits | B3·s̄ and B4·s̄ |
| c4 | S Chart | Unbiasing factor for estimating σ from s̄ | σ̂ = s̄ / c4 |
| d2 | R Chart / All | Estimating process σ from range: σ̂ = R̄/d2 | σ̂ = R̄ / d2 |
| E2 | Individuals / MR | X-chart limits using moving range (n=1) | X̄ ± E2·MR̄ |
Control Limits vs Specification Limits: A Critical Distinction
Confusing control limits with specification limits is one of the most common errors in Statistical Process Control. The two concepts address completely different questions and must never be used interchangeably.
Control Limits (UCL, LCL)
Calculated from actual process data using control chart constants (A2, D3, D4). They answer: What variation does this process routinely produce? A point outside control limits signals that something has changed in the process — a special cause is present. Control limits are set by the process, not by requirements.
Specification Limits (USL, LSL)
Set by engineering drawings, customer contracts, or design requirements. They answer: What range of values is acceptable for this product? Specification limits define product conformance. They are set by external requirements, not by process behavior. Setting spec limits equal to control limits is always incorrect.
The Four Possible Situations
| Process State | Control Chart | Specifications | Action |
|---|---|---|---|
| In control, meets specs | All points in limits | Cp/Cpk ≥ 1 | Continue monitoring |
| In control, fails specs | All points in limits | Cp/Cpk < 1 | Process improvement needed — design or capability problem |
| Out of control, meets specs | Point(s) outside limits | Output currently conforming | Investigate special cause — problem may worsen |
| Out of control, fails specs | Point(s) outside limits | Defective output present | Immediate investigation and containment |
Common Mistakes When Using A2, D3, and D4
These errors appear frequently in practice and in academic submissions. Each one produces incorrect control limits.
Reading the wrong row
The most common error: looking up constants for n = 4 when subgroups actually contain 5 observations. Always verify n before reading across the table. Double-check the subgroup size used in the data collection, not just what was planned.
Applying A2 to the R chart
A2 belongs on the X-bar chart only. Using UCLR = X̄̄ + A2 × R̄ produces a nonsense result. The R-chart limits require D3 and D4 applied to R̄, not to the grand mean.
Treating D3 = 0 as an error
For n ≤ 6, D3 = 0 and LCLR = 0. This is correct and intentional. Setting LCLR to some positive arbitrary value "to be safe" is statistically incorrect and produces false signals.
Confusing R with R̄
R is the range of a single subgroup. R̄ (R-bar) is the average of all subgroup ranges across the baseline period. Control limit formulas use R̄, not the range of any individual subgroup.
Mixing X-bar/R and X-bar/S constants
Using A3 (an S-chart constant) in place of A2, or vice versa, produces limits that do not correspond to ±3σ for the chosen variation measure. Each constant set is derived specifically for its associated chart type.
Setting control limits equal to specification limits
This is covered in detail above. Control limits describe process behavior; specification limits describe product requirements. Equating them makes it impossible to detect real process problems and typically leads to either over-adjustment or missed signals.
Practice Problems: A2, D3, D4 Calculations
These problems cover the full range of skills needed to work confidently with control chart constants. Click any question to reveal the solution.
Solution: From the control chart constants table, for n = 4: A2 = 0.729, D3 = 0, D4 = 2.282.
D3 = 0 because n = 4 falls below the threshold (n = 7) at which the lower 3-sigma limit on the range becomes positive.
Given: n = 5 → A2 = 0.577. X̄̄ = 25.0, R̄ = 1.2.
UCLX̄ = 25.0 + 0.577 × 1.2 = 25.0 + 0.692 = 25.692
LCLX̄ = 25.0 − 0.577 × 1.2 = 25.0 − 0.692 = 24.308
Given: n = 5 → D3 = 0, D4 = 2.114. R̄ = 1.2.
UCLR = 2.114 × 1.2 = 2.537
LCLR = 0 × 1.2 = 0
A subgroup range greater than 2.537 mm would fall above UCLR and signal increased process variation.
Given: n = 7 → D4 = 1.924. R̄ = 3.5.
UCLR = 1.924 × 3.5 = 6.734
The observed range of 7.2 exceeds UCLR = 6.734, so this subgroup is out of control on the R chart. The within-subgroup variation is greater than the process normally produces. Investigate before interpreting the X-bar chart.
D3 = 0.136 is derived from the sampling distribution of the subgroup range, not the subgroup standard deviation. For an X-bar/S chart, the variation measure is s (sample standard deviation), and the S-chart lower limit uses B3, not D3. B3 for n = 8 is 0.185. Using D3 on an S chart produces a control limit that does not correspond to ±3σ in the s-distribution and will give incorrect signals.
No — on the basis of that single point alone, the R = 5.1 subgroup is in control, since 5.1 < 8.2. However, in practice, you would also examine the R chart for non-random patterns (such as a sustained run above R̄, or a trend upward), not just individual points beyond the control limits. A process can exhibit special cause behavior through patterns even when all individual ranges remain below UCLR.
Control limits first (n = 5, A2 = 0.577):
UCLX̄ = 50.0 + 0.577 × 4.0 = 52.308
LCLX̄ = 50.0 − 0.577 × 4.0 = 47.692
Process capability: σ̂ = R̄ / d2 = 4.0 / 2.326 ≈ 1.720 (d2 = 2.326 for n = 5).
Cp = (USL − LSL) / (6σ̂) = (53 − 47) / (6 × 1.720) = 6 / 10.32 ≈ 0.58
Cp = 0.58 < 1.0, so the process is not capable. Even if statistically in control, the process produces output across a range that exceeds the specification window. This is a process improvement problem, not a control limit problem.
For n = 10: D4 = 1.777.
UCLR = 1.777 × 2.1 = 3.732
R = 4.8 > UCLR = 3.732. Yes, this is an out-of-control signal on the R chart. The within-subgroup variation for this subgroup is greater than the upper control limit. The cause should be investigated before the corresponding X-bar point is interpreted.
Frequently Asked Questions
A2 is used exclusively on the X-bar chart within the X-bar/R chart pair. It multiplies the average subgroup range (R̄) to set the ±3σ control limits around the grand mean. The formulas are UCLX̄ = X̄̄ + A2 × R̄ and LCLX̄ = X̄̄ − A2 × R̄. A2 is not used on the R chart; D3 and D4 handle R-chart limits.
No. Each constant has a specific value for each subgroup size n. Using the wrong n gives wrong control limits. The values must be read from the table for the actual subgroup size used in data collection. A2 for n = 2 is 1.880; for n = 5 it is 0.577; for n = 10 it is 0.308, showing a substantial difference as subgroup size changes.
Once you have R̄, use the d2 constant: σ̂ = R̄ / d2. For n = 5, d2 = 2.326, so if R̄ = 0.095, then σ̂ = 0.095 / 2.326 ≈ 0.041. This estimate of process σ applies within subgroups and represents short-term variation. It is the same variation that A2 incorporates when setting X-bar control limits at ±3σ/√n.
Process capability compares the natural variation of a process, typically represented by ±3σ, with the specification limits required for the product or process. The common capability index is Cp = (USL − LSL) / (6σ̂), where σ̂ = R̄/d2 when estimated from control chart data. Control limits describe the expected behavior of the process, while capability indices assess whether that process can meet customer or engineering specifications.
Yes. A simple approach is to create a lookup table in Excel with subgroup size n in column A, A2 in column B, D3 in column C, and D4 in column D. You can then use INDEX/MATCH or VLOOKUP to retrieve the appropriate constant. For example, =INDEX($B$2:$B$26, MATCH(n_cell, $A$2:$A$26, 0)) retrieves A2 for the selected subgroup size. The control limits can then be calculated using formulas such as =Xbarbar + A2_value * Rbar.
A common recommendation is to collect at least 20 to 25 subgroups when the process is believed to be operating under stable conditions. Fewer subgroups can produce less reliable estimates of X̄̄ and R̄, which can make the resulting control limits less dependable. When possible, collecting more subgroups provides a better basis for estimating normal process variation and identifying unusual signals.
References and External Sources
Montgomery, D.C. (2020). Introduction to Statistical Quality Control, 8th ed. Wiley. — The standard academic reference for SPC, including complete tables of control chart constants and their derivation. The constants in this page are consistent with Appendix VI of that text.
NIST/SEMATECH e-Handbook of Statistical Methods. Section 6.3.2: Control Chart Constants. National Institute of Standards and Technology. itl.nist.gov — U.S. government reference covering the full derivation and tabulation of SPC constants.
American Society for Quality (ASQ). Control Chart — Learn About Quality. asq.org/quality-resources/control-chart — ASQ overview of control chart types, usage, and the role of control chart constants in quality engineering practice.
Tippett, L.H.C. (1925). "On the Extreme Individuals and the Range of Samples Taken from a Normal Population." Biometrika, 17(3–4), 364–387. — The original paper that computed the expected values and standard deviations of the range distribution for normal samples, which underlie d2, d3, and all derived control chart constants.
Shewhart, W.A. (1931). Economic Control of Quality of Manufactured Product. Van Nostrand. Republished by ASQ Quality Press (1980). — The foundational text introducing statistical control charts and the concept of control limits derived from process data.
Related Statistical Tables & Resources
Statistical Context: Why Control Chart Constants Work
The ±3σ Basis
Shewhart chose ±3σ control limits as a practical tradeoff between sensitivity (detecting real process changes) and specificity (avoiding false alarms). Under a normal distribution, ±3σ contains 99.73% of the data. The probability of a false alarm on any single subgroup is about 0.27%. Control chart constants translate this ±3σ standard into the range scale using the relationship σ = R̄/d2. For foundational grounding, see Statistics Fundamentals' normal distribution guide.
Why the R Chart Comes First
The X-bar control limits use A2 × R̄ to estimate process spread. If the R chart is out of control, R̄ is not a stable estimate of σ, and the resulting X-bar limits are unreliable. Always verify R-chart control before interpreting X-bar chart signals. This sequencing — R chart first, X-bar chart second — is standard practice in all SPC references, including ASQ's guidance on control charts.
Range vs Standard Deviation as a Variation Measure
For small subgroup sizes, the range (max − min) captures nearly as much information about process spread as the full standard deviation, with far less calculation. At n = 10, the range uses only the two extreme values and ignores the other eight. This efficiency loss is why the S chart becomes preferable above n = 10. The X-bar/R chart's computational simplicity was particularly valuable before computers, but the constants remain in wide use today. See the standard deviation guide for the comparison.