Beta-Tester gesucht: Prüfpläne mit calScript schreiben und den Messplatz mit dem calServer Runner steuern.Bei der Beta mitmachen

Measurement analysis

Passed is not an answer.

Two analyses that turn measured values into a decision: how close was it, and where is the device heading?

Not just passed, but how close

Measurement uncertainty belongs in the picture, not in the footnote.

A calibration certificate says "passed". It doesn't say whether the measured value passed at 12 percent of the tolerance or at 93. But that is the number a lab uses to decide whether a device can run another year or needs adjusting now.

The tolerance chart plots each measuring point as a percentage of the tolerance, with the uncertainty bar around it. Points whose bar reaches beyond the limit are marked by calServer as close. There, the decision rule per ILAC-G8 decides, not the measured value alone. In the picture: seven measuring points, highest use 93 percent, one of them close, none outside.

The basics are in our wiki (in German): Measurement uncertainty explained simply and Decision rule.

calServer tolerance chart of a calibration: deviation per measuring point as a percentage of the tolerance, with uncertainty bars and a marked close point

How much longer will the device last?

Four calibrations of a voltage reference standard, with the trend line up to the next due date. Move the measured value and see when it gets close.

i.T.KonformDie Annahmegrenze rückt um U nach innen: konform bis ±0,080 V um den Sollwert. Die Abweichung beträgt 0,066 V.Wahrscheinlichkeit, dass der wahre Wert in der Toleranz liegt: über 99,9 %. Risiko einer falschen Annahme: unter 0,1 %.
Derselbe Messwert nach allen drei Regeln
RegelAnnahme bisUrteil
Nichtbinär mit Schutzband±0,080 Vi.T. Konform
Binär mit Schutzband±0,080 Vi.T. Konform
Einfache Akzeptanz±0,100 Vi.T. Konform
Verlauf und Trend bis zur Fälligkeit
!i.T.!10,10010,0009,900202320242025202616.03.2027

Trend: +0,018 V pro Jahr. Bei gleichem Verlauf liegt der Wert zur Fälligkeit am 16.03.2027 bei 10,084 V (?, in der Toleranz, Konformität nicht nachweisbar). Die Annahmegrenze wird voraussichtlich um den 27.12.2026 erreicht.

Werte als Tabelle
DatumMesswertUUrteil
14.03.202310,012 V0,020 Vi.T.
12.03.202410,031 V0,020 Vi.T.
18.03.202510,048 V0,020 Vi.T.
17.03.202610,066 V0,020 Vi.T.
  • i.T.konform
  • ?in der Toleranz, Konformität nicht nachweisbar
  • !?außerhalb, Nichtkonformität nicht nachweisbar
  • !nicht konform

Rechenannahme für die Wahrscheinlichkeit: Normalverteilung um den Messwert mit u = U / 2. Das Beispiel ist symmetrisch; calServer rechnet obere und untere Grenze getrennt und bewertet einen Wert genau auf einer Grenze zugunsten der inneren Zone.

See drift before it becomes a finding

Eight calibrations, one measuring point, one statement.

A single calibration certificate is a snapshot. The question a lab really cares about is: is this device drifting away? Anyone who wants to read that from eight PDF certificates ends up typing numbers into Excel.

calServer analyzes the same measuring point across all calibrations of the device: deviation against calibration date, the tolerance band around it, a trend line inside. From four usable calibrations on, a t-test checks whether the slope is more than scatter. Only then does calServer call it detectable and quantify it per calendar year. That is the difference between "the curve looks like it is rising" and a number an auditor accepts.

More in our wiki (in German): Drift and Setting and adjusting calibration intervals.

calServer measurement history of an instrument: drift of one measuring point over eight calibrations with trend line and tolerance limits

What the analysis does

Both work on the same measured values that are created during calibration anyway.

Tolerance chart

Tolerance use instead of yes/no

Every measuring point shows how much of the tolerance it uses, not just that it fits.

  • Deviation as a percentage of the tolerance
  • Uncertainty bar per point
  • Close marking per ILAC-G8

Measurement history

Drift with proof

The same measuring point across all calibrations, with a tested slope instead of guesswork.

  • Trend line from four calibrations
  • t-test for significance
  • Rate of change per calendar year

Point matching

The same point, even after years

Measuring points are matched by nominal value, type and range, even if the description changes.

  • Measuring point key instead of line number
  • Recognizes renaming
  • Reports what doesn't match

Data hygiene

Also says what it doesn't know

Whatever can't be analyzed is named instead of being silently included.

  • Limits and units are checked
  • Discarded rows are counted
  • No silent assumptions

Measurement analysis

Your measured values already exist. Use them.

We show you the tolerance chart and drift analysis on your own calibration data.