Physics · Article

Calibration, standards, and traceability

Read a calibration relation, use it within its stated conditions, and examine how a measurement result is connected to a reference.

Previous article: Error propagation and uncertainty budgets

Back to Physics

After reading, you can

Distinguish calibration, adjustment and verification.
Explain the role of a measurement standard and its associated uncertainty.
Invert a supplied calibration equation while preserving units and its allowed domain.
Interpolate only under an explicitly supplied rule and separate that rule from evidence about an instrument.
Identify the documented calibration links needed to support metrological traceability of a result.
Check a named measurement requirement without treating traceability as sufficient for every purpose.

Establish a relation between reference values and indications

An indication is the value displayed or otherwise supplied by a measuring system. Calibration connects indications to values supplied by measurement standards, retaining the uncertainties on both sides. Under specified conditions, that information is then used to establish how an indication yields a measurement result.

The relation might be given as a function, table or correction. It answers a concrete question: with this instrument, procedure and set of conditions, what result should be assigned to this indication? Comparing numbers without their units, quantity meanings and uncertainty information leaves that question incomplete.

Our examples use fictional length-calibration information. Their central numbers are stipulated for arithmetic, and their uncertainty statements are separately supplied assumptions. They are neither observations from a laboratory nor proof that a real instrument follows the chosen relation.

Separate calibration, adjustment and verification

Calibration establishes the relation. Adjustment changes the measuring system so that it supplies prescribed indications for given quantity values: changing a zero or scale factor setting is an example. The VIM treats calibration as a prerequisite for adjustment and says that recalibration is usually needed afterward.

Verification supplies objective evidence that a specified requirement is met. The requirement might concern a measurement procedure's target uncertainty. When relevant, uncertainty must enter the assessment; merely obtaining an attractive central number does not settle the requirement.

Applying a correction in a calculation need not change the instrument. Keep that operation distinct from turning a control or changing a stored setting. NPL's guide illustrates this practical separation between calibration corrections and equipment adjustments.

Three operations ask different questions; the requirement for verification must be stated.
Operation	Question
Calibration	What relation connects indications, reference values and uncertainties?
Adjustment	What change makes the system provide the prescribed indications?
Verification	What evidence shows that the specified requirement is met?

Identify the measurement standard

A measurement standard is a realization of a quantity definition used as a reference, with an assigned value and associated uncertainty. It can be a measuring system, a material measure or a reference material. Its role is not an assertion that its assigned value is an exactly known true value.

The word standard can also mean a written specification or technical recommendation. Here, measurement standard means the reference used in measurement. A document describing requirements and the physical or material reference that helps establish a quantity value play different roles.

For a length comparison, identify the particular reference, the length it represents and the relevant date and conditions. A measurement standard's uncertainty contributes when it is used to obtain another result. The reference's name alone is insufficient information for using it.

Read the conditions and domain before calculating

Specify what is measured, which instrument and procedure are used, and the applicable range and conditions. Australia's Measurement Standards Unit, for example, asks clients to supply instrument identification, range and special conditions when requesting calibration. That is concrete service guidance, rather than a universal certificate template.

Suppose a fictional report covers the length between two specified end faces along one marked direction. It applies only with the instrument horizontal, its zero setting unchanged, and temperature T between 19 °C and 21 °C inclusive. Its central relation is r = 1.05ℓ + 0.30 mm, for 4 mm ≤ ℓ ≤ 12 mm. Here r is the indication and ℓ is the reference-length value; both are lengths. The coefficient 1.05 has no unit.

The table's central pairs follow that supplied relation. Each reference value has standard uncertainty 0.010 mm and each corresponding indication has standard uncertainty 0.020 mm. Standard uncertainty expresses uncertainty on a standard-deviation scale, not a guaranteed maximum error. Separately, the fictional report supplies u = 0.040 mm for length results obtained by its stated use procedure throughout the allowed interval. That result uncertainty is an assumption here, not something derived from these three pairs.

Original stipulated calibration information; central arithmetic is exact within the model, while the supplied uncertainties remain nonzero.
Reference length ℓ	Reference standard uncertainty	Corresponding indication r	Indication standard uncertainty
4.00 mm	0.010 mm	4.50 mm	0.020 mm
8.00 mm	0.010 mm	8.70 mm	0.020 mm
12.00 mm	0.010 mm	12.90 mm	0.020 mm

Invert the relation and check the correction

To obtain a length estimate from an indication, subtract 0.30 mm from both sides of the central equation, then divide by the nonzero coefficient 1.05. The resulting central estimate is ℓ̂ = (r − 0.30 mm)/1.05. The hat marks an estimate; rearranging an equation does not remove the uncertainty of the information used.

Because the supplied slope is positive, the permitted reference interval maps to 4.50 mm ≤ r ≤ 12.90 mm. Check this interval as well as the conditions before applying the inverse. The subtraction combines lengths, and division by a dimensionless number leaves a length.

Define an additive correction for this model as C(r) = ℓ̂ − r, so that r + C(r) = ℓ̂. Its sign and magnitude depend on r. NPL stresses checking the direction of a calibration correction; VIM allows corrections expressed through additions, factors or tables. A correction compensates an estimated effect and is not a declaration of zero residual uncertainty.

Interpolate with a specified rule

Interpolation assigns a value between supplied points using a stated rule. Our central relation is affine: it has a constant slope and an offset. Use the two pairs (r₁, ℓ₁) = (4.50 mm, 4.00 mm) and (r₂, ℓ₂) = (8.70 mm, 8.00 mm). For an indication between them, set λ = (r − r₁)/(r₂ − r₁). The ratio λ is dimensionless and lies between zero and one.

The interpolated central value is ℓ̂ = ℓ₁ + λ(ℓ₂ − ℓ₁). It agrees with the inverse equation in this stipulated model. That agreement checks our arithmetic; two endpoint comparisons alone would not establish a real instrument's behavior between them or determine its uncertainty there.

Beyond the supplied range, the same algebra would be extrapolation. The report does not authorize that use. A different orientation, altered setting or temperature outside the stated interval also requires further applicable information. Being able to evaluate a formula is different from having a justified measurement result.

Follow the documented reference chain

Metrological traceability belongs to a measurement result. It connects that result to an identified reference through documented, uninterrupted calibration links, with every calibration contributing to the result's uncertainty. The reference can be a unit realization, a specified measurement procedure or a measurement standard.

An illustrative route could run from reference R to a laboratory reference, then to a working standard, and then through calibration of the measuring instrument to the result obtained with the stated use procedure. A working standard is used for routine calibration or verification of instruments or measuring systems. NPL describes reference and working standards in practical hierarchies. This is one possible route, not a required number of levels.

For each calibration link, retain the identities, relevant dates, conditions, assigned values or relations, units and uncertainty information. Explain how the links and use procedure support this particular result. NIST places responsibility for supporting a traceability claim on the result provider and for assessing it on the user. A sticker, serial number or laboratory name cannot replace the missing relation and uncertainty records.

Check fitness for the specified use

A valid reference chain does not guarantee that a result is useful for every task or free from mistakes. NIST's traceability policy and the VIM both distinguish traceability from adequate uncertainty for the intended purpose. Define the task before deciding that a result is suitable.

For our fictional report, a task requiring standard uncertainty no greater than 0.050 mm passes that one numerical check because 0.040 mm ≤ 0.050 mm. A task requiring no greater than 0.030 mm fails it. Neither comparison establishes product conformity or covers unstated requirements; the report's range, conditions and valid supporting records are still needed.

Read calibration information as an evidence package: what was related to what, under which conditions, with which uncertainty, and through which reference chain? Preserve the applicable procedure and corrections when using it. Calculating a full uncertainty budget is a further task; this lesson's supplied result uncertainty does not invite adding every uncertainty printed in a chain a second time.

From a reference relation to an applicable result

1. Establish the relation
Reference values and corresponding indications carry uncertainty. The fictional central relation is r = 1.05ℓ + 0.30 mm.
2. Use it within scope
Check 4.50 mm ≤ r ≤ 12.90 mm, 19 °C ≤ T ≤ 21 °C, the horizontal orientation and unchanged zero setting. Then use ℓ̂ = (r − 0.30 mm)/1.05.
3. Retain the reference links
Document the reference and each applicable calibration relation, date, conditions and uncertainty contribution. A sticker alone does not establish the result's metrological traceability.
4. Check the named requirement
The fictional result uncertainty is supplied as u = 0.040 mm within the stated use procedure. It passes u ≤ 0.050 mm and fails u ≤ 0.030 mm; traceability does not change those comparisons.

Four questions organize calibration use. The arithmetic is an original stipulated model, and the reference chain must support the particular result.

Worked examples

Use an indication without changing the instrument

Under all of the fictional report's stated conditions, r = 10.80 mm. Find the central length estimate and its additive correction. Retain the supplied result standard uncertainty.

The indication is inside 4.50 mm ≤ r ≤ 12.90 mm, and the question stipulates the applicable conditions.
Subtract the offset: 10.80 mm − 0.30 mm = 10.50 mm. Divide by the dimensionless slope: 10.50 mm/1.05 = 10.00 mm.
C = ℓ̂ − r = 10.00 mm − 10.80 mm = −0.80 mm. Check the sign: 10.80 mm + (−0.80 mm) = 10.00 mm.
The fictional report separately assigns u = 0.040 mm under the use procedure. The calculation neither changes the instrument's settings nor makes the estimated length exact.

The central estimate is 10.00 mm, the additive correction is −0.80 mm, and the supplied result standard uncertainty remains 0.040 mm.

Interpolate within the supplied central model

For r = 5.55 mm under the allowed conditions, interpolate between the first two table pairs. Then assess whether the report supports r = 13.95 mm.

The first indication lies between 4.50 mm and 8.70 mm. Its fraction of the indication interval is λ = (5.55 − 4.50)/(8.70 − 4.50) = 1.05/4.20 = 1/4; the common millimetre units cancel.
The length interval is 8.00 mm − 4.00 mm = 4.00 mm. Thus ℓ̂ = 4.00 mm + (1/4)(4.00 mm) = 5.00 mm.
The inverse equation agrees: (5.55 mm − 0.30 mm)/1.05 = 5.00 mm. This is a check of the supplied affine model, not evidence of real interpolation performance.
For r = 13.95 mm, the formal inverse gives 13.00 mm, beyond the stated 12.00 mm upper reference limit. The report supports neither that use nor assigning its 0.040 mm result uncertainty there.

The supported central interpolation is 5.00 mm, with the supplied 0.040 mm standard uncertainty under the procedure. The formal 13.00 mm extrapolation is outside the report's scope.

Assess the use as well as the reference chain

Assume the fictional result's three calibration links C1, C2 and C3 are valid and documented, each with its uncertainty contribution, and reference R is explicitly identified. For r = 6.60 mm, compare task A at T = 20 °C requiring u ≤ 0.050 mm, task B at T = 20 °C requiring u ≤ 0.030 mm, and task C at T = 24 °C requiring u ≤ 0.050 mm. All other stated conditions are met.

The indication is within the allowed interval. At the allowed temperature, the central inverse gives (6.60 mm − 0.30 mm)/1.05 = 6.00 mm, with the separately supplied u = 0.040 mm.
For task A, 19 °C ≤ 20 °C ≤ 21 °C and 0.040 mm ≤ 0.050 mm. The stated range, temperature and uncertainty checks pass under the assumed valid documentation.
For task B, 0.040 mm > 0.030 mm. A valid traceability chain does not overcome this failure of the named uncertainty requirement.
For task C, 24 °C is outside 19 °C ≤ T ≤ 21 °C. The report does not justify applying its relation and result uncertainty under that condition; the task requires further applicable evidence.

Task A passes the specified checks; task B fails the uncertainty criterion; task C is outside the supplied conditions. This is a scoped assessment of those requirements, not a blanket conformity certificate.

Try it yourself

A laboratory establishes a relation with uncertainty information, changes the instrument's zero setting, and then checks whether a stated target uncertainty is met. Name these three operations.
Show the answer and reasoning
Calibration, adjustment, and verification, respectively.
The first establishes the relation, the second changes the measuring system, and the third evaluates a specified requirement. Recalibration is usually needed after adjustment.
Within the fictional report's allowed conditions, r = 6.60 mm. Find ℓ̂ and C(r).
Show the answer and reasoning
ℓ̂ = 6.00 mm and C(r) = −0.60 mm; the supplied result standard uncertainty remains 0.040 mm.
(6.60 mm − 0.30 mm)/1.05 = 6.00 mm, and 6.00 mm − 6.60 mm = −0.60 mm. Adding the correction recovers the estimate.
For r = 7.65 mm under the allowed conditions, use the first two table pairs and their explicitly affine interpolation rule. Find λ and the central length estimate.
Show the answer and reasoning
λ = 3/4 and ℓ̂ = 7.00 mm.
λ = (7.65 − 4.50)/(8.70 − 4.50) = 3.15/4.20 = 3/4. Hence ℓ̂ = 4.00 mm + (3/4)(4.00 mm) = 7.00 mm. The inverse equation gives the same central value; the supplied procedure still carries uncertainty.
At T = 20 °C, r = 13.95 mm. The inverse equation produces a number. May you report it using this report's 0.040 mm standard uncertainty?
Show the answer and reasoning
No. The formal central value is 13.00 mm, outside the stated range.
13.95 mm exceeds the allowed upper indication 12.90 mm. The algebra remains evaluable, but the supplied calibration relation and uncertainty are authorized only within the declared domain.
An instrument carries a sticker naming a national laboratory. Its records identify all three calibration links, but one link has no available uncertainty information. Does the sticker and sequence of names establish metrological traceability of the present result?
Show the answer and reasoning
No. The evidence does not establish the required contribution from that calibration link.
Metrological traceability requires documented calibration links contributing to the result's uncertainty. The missing information must be obtained and its applicability assessed; a laboratory name cannot supply it. This is a conclusion about insufficient evidence, not proof that the actual measurement error is large.
A documented result under all allowed conditions has standard uncertainty 0.040 mm. A task specifies u ≤ 0.030 mm. Does traceability make it meet this requirement, and is 0.040 mm a guaranteed bound on the error?
Show the answer and reasoning
It fails the specified requirement, and the standard uncertainty is not a guaranteed error bound.
0.040 mm > 0.030 mm. Traceability concerns the supported reference connection; standard uncertainty expresses dispersion on a standard-deviation scale. Neither makes the result automatically fit the task or supplies a maximum error.

What to carry forward

Calibration establishes a relation with uncertainty; adjustment changes the system; verification checks specified requirements.
A measurement standard supplies an assigned reference value with uncertainty, rather than an automatically exact true value.
Use a calibration equation with the correct units, inverse direction, conditions and domain.
An interpolation rule and its range must be supplied or justified; endpoints alone do not establish behavior between them.
Metrological traceability connects a particular result to a reference through documented calibration links contributing uncertainty.
A valid chain still leaves the intended use, adequate uncertainty, applicable procedure and possible mistakes to be assessed.

References

Sources for the concepts and teaching approaches in this article.