Measurement Results¶

Measurement results capture the actual calibration data of a test or measuring instrument. They are documented within a calibration and evaluated against Specifications.

Overview¶

Measurement results are accessible in the Calibration Detail View (/v2/calibration/[id]) under the Results tab. Each result is linked to the associated calibration via the CTAG (calibration tag).

Results Table¶

The results are displayed in a table. These columns are shown by default:

Column	Description
Row	Sequential number of the measurement point
Description	Name of the test step
Fixed Quantity	Setpoint of the measurement point
Variable Quantity	The captured measured value
Expanded Uncertainty	Measurement uncertainty U of the point
Status	Passed / Failed

Raw data and provenance of a measurement row¶

Each row carries an expander. It shows three blocks:

Raw data of this row — every field of the row that holds a value, regardless of which columns are currently shown. This is the block for checking an import: whatever is not listed here is empty. Without it, "the field is empty" could not be told apart from "the column is not displayed", and you had to guess which of the roughly 250 columns to enable in order to check.
Linkage and timestamps — the keys the row hangs off (calibration, device, and for imported rows the MET/TEAM UIDs of call sheet, call sheet result and point along with the instance), plus the source's timestamps and calServer's own. If an imported row differs from what MET/TEAM holds, the UIDs are how you find it over there.
Normalized measurement row — the base values, i.e. the same quantities without their SI prefix converted to the base unit (useful when rows in mV and kV sit next to each other), and the provenance: whether the row came from the test run, from manual entry or from a MET/TEAM import, when it was captured, which test step it belongs to and which engine version computed it.

If the third block reads "no normalized measurement row exists for this row", the row predates this record keeping. The measured value itself is unaffected, and the raw data above it is complete. The record is filled in on the next restart of the API container; to do it right away:

docker exec calserver-api-v2 php artisan measurements:backfill

A row without a single measured value is not an error, by the way: MET/CAL carries header and text lines as measurement points too.

More measurement fields¶

A measurement point has around 250 fields — the full metrology set inherited from MET/CAL and MET/TEAM. There are two ways to bring them into the grid:

For everyone, as the installation's default: in Field Management under the Measurement object, where fields are shown, hidden, renamed and reordered. This applies to every user and every calibration.
For yourself, at the table: the Columns button above the results table opens a searchable list of every measurement field. The choice is kept for the signed-in user; Reset restores the Field Management default.

The second way is meant for the occasional question about a single certificate ("what was the tolerance utilisation here?"). It does not change the tenant's default.

Switched-on columns while editing

Additionally shown columns stay visible in edit mode, but without an input field. Only what Field Management marks as editable is writable — an input on any other column would silently discard the typed value on save.

The most important fields:

Field	Meaning
Lower Limit / Upper Limit	Error limits, computed as setpoint ± tolerance, in the display digits of the step (resolution, otherwise at least three places) rounded towards the setpoint; 0 is a valid limit
Tolerance Positive / Negative	Allowable deviation upwards and downwards
Tolerance Reference	Nominal (the setpoint is the reference) or UUT Indicated (the instrument's reading leads)
Tolerance Error	Share of the tolerance used up, in percent
Error	Difference between the UUT reading and the reference value
UUT Indicated / Reference Value	The two values the error is derived from
Standard Uncertainty / Coverage Factor	Uncertainty u before expansion and the factor to U
Guardband Lower/Upper Limit	Decision limits, moved inwards by the expanded uncertainty
Guardband Method	Decision rule behind the conformity statement, e.g. `ILAC-G8 w=U`
UUT Range / UUT Resolution	Reference quantities for tolerance terms with "/" and "D"
Function Select Code	MET/CAL FSC of the test step
Accredited	Is the point inside the accreditation scope? Written at capture from the annex in the quality specifications, see Accreditation scope

Every measured quantity is stored as a triple of value, prefix and unit — the prefix is the SI prefix (m, k, µ; case matters), the unit sits next to it without a prefix (V, A, Z for ohms). In the grid the two companions carry their field's name plus the qualifier in brackets: "Fixed Quantity (prefix)", "Fixed Quantity (unit)".

Two objects, the same fields

Field management keeps Measurement (the view/edit mask) and Measurement (grid) (the columns of the results table) apart. The field names are identical in both; visibility is configured independently.

Recording Results¶

Individual Results¶

Open the calibration detail view
Switch to the Results tab
Click + Add Result
Enter the measurement point and the measured value
The evaluation is performed automatically against the stored specifications

Bulk Editing¶

For efficient entry of multiple measurement points, a bulk editing function is available:

Click Bulk Edit
All measurement points are displayed in an editable table
Navigate between input fields using the Tab key
Enter the actual values directly into the table
Click Save All to apply all entries

Test Steps as Template

If a Procedure has been assigned to the calibration, the test steps are automatically pre-filled as measurement points. You only need to enter the actual values.

Automatic Evaluation¶

The evaluation of measurement results is performed automatically based on the stored Specifications:

Passed — The actual value is within the defined tolerances
Failed — The actual value exceeds the tolerance limits

Example:
  Target value:  10.000 V
  Tolerance:     ±0.003 V (= 9.997 V to 10.003 V)
  Actual value:  10.002 V → Passed ✓
  Actual value:  10.005 V → Failed ✗

Overall Evaluation

The overall result of the calibration (field C2316) is derived from the combination of all individual evaluations. If at least one measurement point has failed, the entire calibration is considered failed.

Measurement uncertainty per GUM¶

If the calibration has a standard whose released specification covers the measuring point (for instance from an imported MET/CAL ACC file), calServer computes the uncertainty on capture, as a budget per GUM:

Contribution	Calculation
U1, standard	Accuracy A from the specification row divided by its confidence (sigma, default 2)
S1, repeatability	Standard deviation of the readings divided by √N (only with `repeat N`), with N − 1 degrees of freedom
S2, resolution	UUT resolution / 2 / √3
U3 … U10	further contributions: from the standard's specification (such as the homogeneity of a bath), from the laboratory's uncertainty model, from the plan (`contrib`), the hysteresis
u	√(U1² + S1² + S2² + U3² + … + U10²)
U	k · u

If more than eight further contributions come together, the remaining ones share U10 (which holds the root of their sum of squares); each stays in the budget on its own, none is dropped.

Coverage factor k. If the plan names k at the group, exactly that applies. Otherwise calServer determines k by default from the effective degrees of freedom (Welch-Satterthwaite, GUM annex G) for a coverage probability of about 95 %. Without repetitions k = 2; if you measure four times and the spread dominates, k grows (2.32 at about ten effective degrees of freedom). Finite degrees of freedom come from the repeatability (N − 1) and from a contribution of the uncertainty model that states them; all other contributions count as exactly known. Under Administration › Basic Settings › Rules › Measurement uncertainty you can fix k = 2 instead, like MET/CAL. The effective degrees of freedom are stored in the measurement row.

The specification row is found by unit, nominal value (within the range) and frequency (modifier within the MOD1 band). The contributions are stored in the measurement fields (system accuracy, U1, S1, S2, standard uncertainty, degrees of freedom), the method in "Uncertainty Method": GUM, GUM, CMC, GUM, Ersatz Normal (stand-in), Vorgabe (stated) or keine (none). If U falls below the smallest specifiable uncertainty (CMC) from the quality provisions, the CMC is stated and the method says so. The accuracy ratio (TAR) uses the standard's accuracy.

If the test plan names roles (such as source cal for the calibrator), you assign one of the chosen standards to each role in the preflight of the test. A point then uses the specification of the standard of its role, even if another standard of the calibration would cover it as well. If the role's standard does not cover the point, no other standard steps in. calServer records per point which standard and which specification row supplied the value.

If several standards are attached to the calibration, the test wizard offers “Standard for this point” at the measuring point. Left empty, the role or the one covering standard applies; a choice applies to this point only and takes precedence over the role. Below the result you can expand the uncertainty budget: method, k and where it comes from, degrees of freedom, u, U, CMC, model, one row per contribution with origin, distribution, standard uncertainty and degrees of freedom (ν, ∞ for an exactly known value), and the standard U1 came from.

A fixed uncertainty entered at the test step always wins (method Vorgabe).

When no specification matches¶

If a point gets no U1, calServer no longer carries on silently. The method is then keine, and the point names the reason: no standard on the calibration, several standards without a role, no released specification, no matching row (with standard, option and nominal value), several equally good rows, an unreadable tolerance, or a point without a measuring function. The test wizard shows the reason at the point, the preflight already before measuring.

If the point has a verdict, this is a conformity statement without measurement uncertainty: the decision rule has to take the measurement uncertainty into account (ISO/IEC 17025, 7.8.6.1). The completion check reports it under completeness and holds the release by default. Add the specification or the standard, or store a stand-in for the standard in an uncertainty model, then recalculate without measuring again. For factory calibrations without an uncertainty statement, turn the finding into a notice under Rules › Measurement uncertainty.

Recalculating without measuring again¶

In the completion check (tab Release), the finding “Conformity statement without measurement uncertainty” or “Measurement uncertainty with a stand-in for the standard” offers the button Recalculate measurement uncertainty. It recomputes exactly these points from their stored readings: U1 from the specification that matches now, the model contributions in their current revision, k, degrees of freedom and verdict. Nothing is measured, the readings stay as they are.

Only before release. While the certificate awaits release or is released, the button is disabled and names the reason; withdraw or reject the process first.
Not while the calibration is checked out to a workstation: the values are being made there.
A point with a budget from a standard stays as it was captured, even if the specification or the model changed since.
A standard chosen at the point stays chosen. Who captured and when stays recorded.
Every change is in the change log of the calibration, per point with method, U, k and verdict before and after.

Afterwards calServer reports how many points now take their U from the specification of a standard or from a stand-in and how many still have none; the completion check reloads. The permission is the same as for editing the results.

Uncertainty models¶

Under Administration › Master Data & Structure › Uncertainty models you store contributions that belong to the laboratory rather than to a standard: thermal EMFs, temperature influence, coupling. A model applies to the whole laboratory or to one device type and to one unit, optionally only to one measuring function and one range. You write the contributions like a tolerance:

Notation	Meaning
`0.0005%`	of the nominal value
`0.01/`	of the UUT range (without a range, of the nominal value)
`2U`	absolute in the model's prefix (with prefix µ, 2 µV)
`1D`	digits of the UUT resolution

Several terms separated by spaces add up. For each contribution you choose the distribution: rectangular, triangular, U-shaped or normal with k = 2. At most eight contributions per model.

Optionally a contribution states its degrees of freedom ν: how reliably the value itself is known (GUM G.4.2). Empty means exactly known (ν = ∞), as before. As a rule of thumb ν ≈ ½ · (Δu/u)⁻²: known to about 10 % gives 50, to 25 % gives 8, to 50 % gives 2. A finite ν lowers the effective degrees of freedom and can raise k above 2 where the contribution dominates the budget.

A model can also carry a stand-in for the standard, with a confidence like a specification. It only applies when no standard covers the point, and the method says so (GUM, Ersatz Normal); the completion check lists such points as a notice. The CMC is never a stand-in.

At each point the most specific matching model applies: device type before laboratory, with range before without, with function before without. If two match equally well, neither applies, and the point says so. Every change raises the model's revision and is logged; captured points keep their values, the model name and the revision they were computed with. Only points without U or with a stand-in can be recalculated on purpose.

Coverage in the preflight¶

Once the standards are chosen, the preflight shows how many measuring points would get a measurement uncertainty (from the specification, with a stand-in, as stated) and which would not, with the reason. It blocks nothing: you can choose another standard before you measure.

On the calibration certificate¶

The certificate states k per point. The default text on the measurement uncertainty follows the k that was used; with k = 2 everywhere it stays as before. The DCC carries k and coverage probability per value.

Edit Results¶

Previously recorded results can be corrected afterwards:

Click the edit icon next to the result
Change the actual value
The evaluation is automatically updated
Save the change

Delete Results¶

Individual results or all results of a calibration can be deleted via the action menu.

Traceability

In regulated environments, measurement results should not be deleted. Instead, mark the calibration as "invalid" and perform a new calibration. This ensures complete documentation is maintained.