Tiza Group

Home / Tools / Calibration and uncertainty

Calibration and measurement uncertainty

Type the readings of the instrument, the value of the standard, its certificate, the resolution and the tolerance: the tool builds the uncertainty budget, calculates k from the effective degrees of freedom and the expanded uncertainty, and decides pass or fail with simple acceptance or with a guard band.

JCGM 100:2008§ 4.2, 4.3, 5.1.2, G.4GUM: measurement uncertaintyEA-4/02 M:2022§ S9.7, anexo Euncertainty in calibrationILAC-G8:09/2019§ 4.2decision rulesJCGM 106:2012§ 7.6, 8.3.2probability of conformanceNASA-HDBK-8739.19-4§ 3-28test uncertainty ratio

Example cases
Calibration and measurement uncertainty
ALearn

E1Instrument readings

Only to display the results: V, A, Ω, °C, bar…

At least two, separated by “;” or spaces. A decimal comma is accepted.

V

What the standard delivers or indicates at that point.

E2Standard and resolution
V

The one stated on the certificate, usually 2.

V

The last digit or the division you read. ±resolution/2 is taken, rectangular.

E3Another contribution
V

Hysteresis, stability of the standard, uniformity… 0 = none.

E4Tolerance and decision
V

The maximum permissible error at this point.

Use a decimal point. The calculation runs in your browser and updates as you type.

Decision

—

Instrument error

—

Expanded uncertainty U

—

Coverage factor k

—

Effective degrees of freedom

—

Combined uncertainty uc

—

Mean of the readings

—

Standard deviation of the readings

—

Type A uncertainty of the mean

—

Tolerance / uncertainty ratio

—

Probability of conformance

— %

Instrument connected to the standard
The standard delivers the reference value and the instrument reads it several times. Each reading falls on the error axis; with them and the type B sources U comes out, and the lamp shows the decision. Change the unit to see the set-up for each quantity.
Uncertainty budget
Each row is a source. The bar is its weight in uc²: the longest one is the source to reduce first. νeff is truncated to the next lower integer to look up k, and U is reported with two significant digits.
Conformity decision
The wide band is the tolerance; the inner one, the acceptance limits. The dot is the error with its ±U bar. The 4 to 1 line is a convention, not an acceptance criterion.

Budget as a table

Step-by-step calculation

Result not what you expected? We check it with you on site.

What it is for

When you calibrate an instrument against a standard, the error alone is not enough: you need to know how far you can trust it. This tool builds the uncertainty budget of one calibration point from the repeated readings (type A), the certificate of the standard, the resolution and one extra limit (type B), and calculates the expanded uncertainty with the coverage factor k that matches the degrees of freedom.

With the tolerance of the instrument it decides pass or fail, with simple acceptance or with a guard band, and shows the probability of conformance and the ratio between tolerance and uncertainty. It works for voltmeters, clamp meters, pressure gauges, temperature indicators, insulation testers and any instrument compared with a standard at one point.

How it is calculated

JCGM 100:2008§ 4.2.2-4.2.3type A

Repeatability: type A

uA = s / √n · ν = n − 1

s is the experimental standard deviation of the n readings. Halving uA takes four times as many readings.

JCGM 100:2008§ 4.3.3, 4.3.7, 4.3.9type B EA-4/02 M:2022§ S9.7resolution

What does not repeat: type B

ucert = U/k · ures = (res/2)/√3 · ulim = a/√3, a/√6 or a/√2

The certificate is divided by its own k, not by √3. The resolution and limits with no known distribution are treated as rectangular; the triangular and U-shaped distributions are used when the value is known to gather at the centre or at the ends. Type B sources enter with infinite degrees of freedom.

JCGM 100:2008§ 5.1.2, G.4.1uc and νeff

Combined uncertainty and effective degrees of freedom

uc2 = Σ (ci·ui)2 · νeff = uc4 / Σ (ci·ui)4/νi

The sources are added in quadrature (no correlation, ci = 1). νeff comes from the Welch-Satterthwaite formula and is truncated to the next lower integer before looking up k.

JCGM 100:2008§ 6.2.1, G.3, 7.2.6U and rounding EA-4/02 M:2022§ Annex Ek factor

Expanded uncertainty

U = k · uc · k = t95.45 %(ν)

k is the Student t factor for a coverage of 95.45 % with the truncated degrees of freedom; with many degrees it equals 2. U is reported with two significant digits and the error with the same decimals.

ILAC-G8:09/2019§ 4.2decision rule JCGM 106:2012§ 8.3.2acceptance limit

Decision with a guard band

AU = TU − w · AL = TL + w · w = U (guard band) or 0 (simple)

The result passes if the error lies between AL and AU; otherwise it fails. With a guard band, an error inside the tolerance but closer than U to the limit fails. With simple acceptance, the risk of accepting an out-of-tolerance instrument at the limit can reach 50 % and is shared with the customer.

JCGM 106:2012§ eq. (4) and (6)probability of conformance NASA-HDBK-8739.19-4§ 3-28TUR

Probability of conformance and TUR

pc = Φ((TU − E)/uc) − Φ((TL − E)/uc) · TUR = T / U

pc assumes a normal distribution around the error E: with few degrees of freedom it is optimistic. TUR is the half-width of the tolerance divided by U; the 4 to 1 reference is a convention of ANSI/NCSL Z540.3 (not consulted), not an acceptance criterion.

Limits of the calculation

  • One calibration point and one measurand with the same unit across the budget: the sensitivity coefficients are 1 and there is no correlation between sources.
  • The sources are the readings, the certificate, the resolution and one extra limit. Drift of the standard, temperature, load or wiring enter through that limit if you estimate them.
  • The probability of conformance uses a normal distribution with uc; with few degrees of freedom the real one has wider tails and the probability is lower than shown.
  • It does not calculate the global consumer or producer risk (JCGM 106:2012, 9.5) or guard bands other than w = U.
  • You set the tolerance and the decision rule (manufacturer specification, instrument standard or agreement with the customer). The tool does not replace the certificate of an accredited laboratory.

Frequently asked questions

Why is the coverage factor k not always 2?

k = 2 gives a coverage close to 95 % only when the combined uncertainty has many degrees of freedom. If repeatability weighs a lot and few readings were taken, the GUM (JCGM 100:2008, G.4) calculates the effective degrees of freedom with the Welch-Satterthwaite formula, truncates them to the next lower integer and takes k from the Student t factor for 95.45 %. With five dominant readings k can exceed 2.5; with three, 3. Reporting k = 2 in that case underestimates the uncertainty.

What is a guard band and when is it used?

It is a margin w subtracted on each side of the tolerance to obtain the acceptance limits. ILAC-G8:09/2019 describes the rule with w = U: the result passes only if the error is more than U away from the limit. The risk of declaring an out-of-tolerance instrument conforming then stays below 2.5 % at the limit. With simple acceptance (w = 0) that risk can reach 50 % and is shared with the customer. The rule is agreed with the customer before calibrating.

Is the 4 to 1 ratio between tolerance and uncertainty mandatory?

Not in the sources the tool uses. The test uncertainty ratio (TUR) is the half-width of the tolerance divided by U (NASA-HDBK-8739.19-4). The figure of 4 to 1 comes from ANSI/NCSL Z540.3, a paid standard that was not consulted, and JCGM 106:2012 (7.6) warns that this ratio is often ill defined. The tool shows it as a reference: what decides is the agreed decision rule.

We test your equipment on site

We carry out the electrical tests of transformers, protection systems and switchyard equipment, and deliver the record of each test with its results.

Related guides

Other tools

All tools
WhatsApp