Calibration overview

Calibration is the core of what Open Gauge does: comparing a sensor's readings against a trusted reference, fitting a correction curve, evaluating measurement uncertainty, and issuing a pass/fail statement — all following JCGM 100:2008 (GUM) and ISO/IEC 17025:2017.

Two backend touchpoints

A calibration event involves two API calls:

1. POST /calibrations/analyze — ephemeral. Takes raw (reference, measured) point pairs plus fitting/uncertainty parameters, runs the full analysis, and returns the result. Nothing is persisted. The calibration wizard calls this on every input change (debounced) to give live feedback as you enter data.
2. POST /calibrations — atomic create. Takes the already-computed result (the wizard forwards whatever /analyze returned) plus metadata, and writes one calibrations row and its data-point rows in a single transaction. A certificate PDF is generated best-effort in the same request.

This split exists so you can freely explore different polynomial degrees, distributions, and decision rules before committing — the expensive, persisted step only happens once, when you save.

The calibration wizard

Start a calibration from an asset's channel. The wizard has three steps:

1. General info — calibration type (internal, against a reference asset you own, or external, from an outside lab), calibration date, who performed it, environmental conditions, and (for internal calibrations) the reference standard used and procedure followed. This step also has a "Coefficients only (no raw data)" checkbox for externally-issued certificates that state a polynomial directly — see Coefficients-only calibrations.
2. Raw data — enter (reference, measured) point pairs. See Curve fitting for what Open Gauge does with them.
3. Analysis / review — the live-computed fit, uncertainty budget, and conformity statement. See The uncertainty budget and Decision rules.

Saving is never blocked by a failed (non-conforming) result — a failed calibration is still an important, real record. If the result doesn't conform to the channel's spec, the save confirmation shows an explicit warning naming the specification and decision rule, and the save button relabels to "Save anyway."

Internal vs. external calibrations

• Internal — performed in-house, against one of your own assets flagged as a reference standard. When you pick a reference asset, Open Gauge automatically fetches that asset's own most recent calibration's expanded uncertainty and folds it into this calibration's uncertainty budget as a reference_standard contribution — see The uncertainty budget.
• External — performed by an outside calibration lab. You record the lab's name and certificate number; if the lab's certificate states a polynomial directly rather than raw data, use the coefficients-only path instead of re-deriving a fit from scratch.

Where results end up

Historical calibrations are viewable from the asset's detail page — every past calibration for every channel, most recent first, each with its full analysis, uncertainty budget, and certificate. Nothing is ever edited in place; a re-calibration is a new row, not a correction to an old one.