Alternative calibration data-entry modes

Step 2 of the calibration wizard, Input data, opens with a dropdown offering four ways of delivering a calibration's data; the rest of the step renders immediately below according to the current selection:

• Reference vs Measured — the default: (reference, measured) point pairs, the reference standard's reading against the channel's own raw output signal, which Open Gauge fits to a curve.
• Reference vs Indicated — the reference standard's reading against what the channel's own display already shows, converted to the same physical quantity — no known transfer function between them.
• Model (transfer function) — the lab/manufacturer certificate states the calibration model itself, with no raw data points at all. Only offered for OEM/External Accredited Lab/Customer's Asset calibrations, and not offered at all when Calibration purpose is After Repair.
• Frequency response — a sweep of frequency/reference/measured points for sensors that deliver a signal in the frequency domain (accelerometers, microphones, etc), reduced to a sensitivity ratio at a chosen baseline frequency. Offered for every Calibration purpose, including After Repair. See Frequency response for the full breakdown.

If the calibration's Calibration purpose is After Repair, Reference vs Measured and Reference vs Indicated both automatically split into two comparisons instead of one — see below. Model isn't offered at all for After Repair, since a single lab-stated model can't represent a before/after comparison; Frequency response is unaffected by purpose and has no as-found/as-left split.

Model (transfer function)

Use this when the certificate states the calibration model directly rather than raw data points — common for OEM/External Accredited Lab/Customer's Asset calibrations. Step 2 becomes a model-entry form:

• Model type — a Polynomial (order 1–5, coefficients entered in the same descending-power convention Open Gauge's own regression engine uses — highest degree first, constant term last), or a Custom formula: write a formula shape in x (the measured/output value) with any other letters as free parameters, e.g. a*x + b, and Open Gauge detects them automatically and asks you to enter a value for each one (from the certificate/datasheet that gave you the model) — see Custom formula syntax for the full allowed syntax and worked examples. What's actually stored is the fully-resolved formula (parameters replaced with their values). A custom formula is fully evaluable — it participates in the saved calibration's fitted-curve chart and the Health tab's drift/curve comparison exactly like a polynomial, just evaluated from the formula instead of polynomial coefficients.
• The valid range (min/max) the model applies over.

Step 3 then collects the same uncertainty budget inputs a Reference-vs-Measured calibration collects (reference standard uncertainty, resolution, sensor nominal accuracy, distribution/confidence, decision rule) — because there's no measured data, conformity is assessed by treating the model as trusted-as-declared (an assessed error of exactly zero) and comparing only the combined uncertainty budget against the channel's tolerance spec, at both ends of the declared valid range.

Reference vs Indicated

Use this when a lab provides the reference standard's measurement alongside the channel's own display reading, already converted to the same physical quantity — but the transfer function itself (how one becomes the other) isn't known or wasn't provided. Since there's no function to fit, Open Gauge can't produce a fitted curve or a polynomial equation for this calibration — but every statistic that only needs a residual (the direct difference between the two values) is still computed exactly as usual: R², RMSE, max error, %FS error, repeatability, hysteresis, the full uncertainty budget, and a real decision-rule conformity statement. The saved calibration's detail view shows these statistics and a residuals chart in place of the usual fitted-curve chart, and labels the second column Indicated — a physical-quantity reading, not a raw sensor signal — with no separate "Fitted" column, since under this mode it would just duplicate the Indicated column.

After Repair: As-Found/As-Left data

When Calibration purpose is After Repair, choosing Reference vs Measured or Reference vs Indicated on Step 2 replaces the usual single data-entry table with two tables side by side — one for data taken before the repair (as-found, left) and one for data taken after (as-left, right) — sharing a single Reference/Measured unit row above both, since it's the same channel either side of the repair.

Step 3 then shows the uncertainty budget inputs and the decision rule/tolerance once, shared across both sides — again, the same channel and spec either side of a repair. Below that, two columns (as-found left, as-left right) each show that side's own conformity criteria — the actual [error] ± [uncertainty] ≤ [tolerance] expression with real numbers plugged in, plus a CONFORMS/ DOES NOT CONFORM badge — followed by that side's own model (if fitted) and statistics:

• Starting from Reference vs Measured, each side is independently curve-fit — as-found and as-left can even end up with different polynomial degrees or a different fit method, since the instrument's behavior may have genuinely changed across the repair.
• Starting from Reference vs Indicated, each side is a residual-only comparison, exactly as described above, computed independently for as-found and as-left.

The as-left result is this calibration record's official result: it's what feeds the next due date, the approval workflow, and the Health tab's history. The as-found result is kept purely as read-only diagnostic context — visible on the saved calibration's detail view, but it's never used for due-date or drift calculations, and doesn't need to conform to any spec on its own.

What's the same either way

Whichever mode you use, the record you get is a normal calibration — it appears in the channel's history, can be voided or go through the approval workflow, and downloads a certificate. A Model or Reference-vs-Indicated calibration's certificate still shows every statistic Open Gauge was able to compute for it; only the specific rows that genuinely don't apply (e.g. a fitted-curve equation when there's no known transfer function) are left out rather than fabricated.

See Example 4 for a worked Model case, including what gets saved and displayed.