Understanding Instrument Accuracy Specs

8 min read · updated 2026-07-13

"±0.1% accuracy" sounds like one number. On a datasheet it can mean five different things — and the differences decide whether your measurement is fit for purpose. Here's how to read the fine print.

The two reference bases: span vs reading

Percent of span is a fixed absolute error: ±0.1% of a 0–100 kPa range is ±0.1 kPa everywhere — a small relative error at full scale, a large one near zero. At 10% of range, that same spec is ±1% of what you're actually measuring. Percent of reading scales with the measurement: ±0.1% of reading is ±0.01 kPa at 10 kPa and ±0.1 kPa at 100 kPa — a constant relative error. Same headline number, radically different behaviour at the bottom of the range. The error unit converter translates any spec between % span, % reading, engineering units and mA — the fastest way to see what a claim really promises at your operating point.

Why turndown quietly destroys accuracy

Range a 0–250 kPa transmitter for 0–50 kPa (a 5:1 turndown) and a ±0.05%-of-URL spec is now ±0.125 kPa on a 50 kPa span — 0.25% of the working span. Deep turndown on flow is worse, because DP flow squares it: at 25% flow the DP is 6.25% of range. This is why "install one big transmitter and re-range it later" has limits, and why datasheet accuracy statements often carry a turndown-dependent formula rather than a single number. Read that formula at your actual range, not the flattering reference range.

Composite specs: the "% reading + % FS" trap

Calibrator and precision-meter specs usually combine terms: "±0.02% of reading ± 0.005% of full scale." Evaluate both at your test point and add them — the floor term dominates at low readings. A spec that looks ten times better than your transmitter can shrink to barely adequate at the 0% test point. Which leads to the ratio that formalises "adequate":

TUR: is your standard good enough?

The test uncertainty ratio compares the tolerance you're verifying against the uncertainty of the standard you're using; the traditional criterion is 4:1. Check any transmitter-plus- calibrator pairing with the TUR calculator. Below 4:1 you aren't necessarily stuck — but you must either widen acceptance limits (guard-banding) or accept rising false-accept risk, consciously.

Accuracy is a bundle, not one number

A reference accuracy figure typically bundles linearity, hysteresis and repeatability at reference conditions — and excludes temperature effects, static pressure effects, and long-term drift, which are specified separately and can each exceed the headline number in real installations. Two habits keep you honest: measure hysteresis and repeatability yourself from up/down calibration data (the hysteresis & repeatability calculator turns your readings into % of span), and total the installed error budget — reference accuracy + ambient temperature effect + drift since last cal — before declaring a measurement fit for a 1%-tolerance duty.

The checklist

When a datasheet says ±X%: of span or of reading? At what turndown? Including which effects, at what temperature? Verified against what standard, at what TUR? Five questions, thirty seconds each — and the difference between an accuracy claim and an accuracy budget.

Calculators used in this guide: error unit converter · TUR calculator · hysteresis & repeatability · calibration table generator