A compound accuracy spec of ±2 °C or 2% of reading ties two error terms together by the greater-wins rule: the larger of the absolute floor and the proportional term governs at every measured point, as defined in standard industrial accuracy terminology [S2][S6].
That structure is the dominant format on infrared pyrometers, RTD transmitters, thermocouple indicators, and clamp meters, where manufacturers want low-end error to stop ballooning when the reading is small [S2].
What "2% of reading" really means
When a datasheet quotes % of reading, the error is computed against the value currently displayed rather than the instrument's full-scale range, a distinction explained in AMETEK Calibration's reading-vs-scale article [S2].
Mass-flow and flow-meter vendors use the same % RD convention because flow error usually scales with Reynolds number, not with the meter's maximum range, and a fixed 0.5% FS spec would overstate the error at the high end while still being optimistic at the low end [S3].
The Lorric flowmeter guide puts both conventions on one line: "% Full Scale (FS) means error is a fixed percentage of the maximum range, while % Reading (Rd/RD) means error scales with the displayed value" [S6], which is the binary choice behind every accuracy spec on a process instrument.
How the ±2 °C floor and the 2% term combine
At low temperatures, the fixed ±2 °C term dominates: a 30 °C reading under a 2% RD spec would only be ±0.6 °C, but the compound spec floors that at ±2 °C, so a hand-held pyrometer specified at "±2 °C or 2% of reading" delivers ±2 °C at 30 °C, not the tighter proportional figure [S2].
At high temperatures, the proportional 2% term takes over: at 600 °C, 2% of reading is ±12 °C, which dwarfs the 2 °C floor, so the proportional term is what the operator actually feels, and datasheet writers usually illustrate the crossover point right at the floor value of 100 °C, where 2% of rdg equals the absolute term [S6].
This greater-of formulation is why two pyrometers with the same headline "2%" number can perform very differently: one with a ±2 °C floor and 2% RD holds accuracy at low temperatures, while a competitor with a 2% FS spec inflates error at low temperatures and understates it near full scale, exactly the trap the AMETEK explainer warns against [S2].
% RD vs % FS: the actual error arithmetic
A 0 to 500 °C transmitter at 0.5% FS carries a flat ±2.5 °C error across the whole span, so at 50 °C that is 5% of reading, while the same instrument at 0.5% RD carries only ±0.25 °C at 50 °C, growing to ±2.5 °C at 500 °C [S2][S3].
In a 4-20 mA loop, a pressure transmitter with 0.1% FS will report 16 mA (25% of span) as 16.00 mA ±0.016 mA, which is 0.064% of reading at that point, and that divergence grows toward the bottom of the range, where a FS spec is most punishing [S2].
For custody-transfer flow, the trade-off matters financially: a Coriolis flow meter with 0.05% RD and no FS floor will out-perform a 0.1% FS unit across most of the operating envelope, which is why % RD has become the default for high-end fiscal metering under OIML R117-class applications [S3][S6].
When to pick ±2 °C or 2% RD, and when not to
Compound ±2 °C or 2% RD is the right spec for surface-temperature scanning, HVAC duct work, food-service probe thermometers, and field portable calibrators, where the probe sees values from near-ambient to several hundred degrees and the user cannot afford a low-end blow-out [S2][S6].
It is the wrong spec for laboratory-grade PRT readings under 50 °C, where a tighter absolute floor of ±0.05 °C and a smaller proportional term is standard, and for differential pressure transmitter work in flow elements, where zero stability and turndown matter more than the percentage slope [S2].
Engineers should also reject compound specs that hide a large % RD under a small floor, or vice versa: a spec written as "±0.5% of reading" with no floor is fine at high values but unprotected against offset error near zero, which is why most process instruments bolt on a fixed °C or mbar floor as insurance [S6].
Field failure modes engineers actually hit
Sensor self-heating in RTD probes is one of the dominant hidden error sources at low temperatures, and a 2% RD spec that looks great on paper will still drift if the excitation current is not trimmed, because the platinum element heats 0.1-0.3 °C above the measured medium at typical 1 mA excitation, an effect that drops as temperature rises [S2].
Reference-junction compensation on Type K thermocouples is a second trap: an uncompensated cold-junction adds roughly ±0.5 °C per 10 °C ambient swing, and that error is additive to whatever % RD the meter claims, so a "2%" spec can degrade to a 4% effective error in a 50 °C ambient without the operator noticing [S2].
Reading the datasheet the way a process engineer should
Always translate the spec into an error band at three points: 10% of range, 50% of range, and 100% of range, then plot it on the calibration sheet, because the difference between % RD and % FS shows up most starkly at 10% of range, where FS specs widen to 10x the proportional figure [S2][S6].
Finally, watch the format string for hidden units: % RD is dimensionless, % FS is dimensionless, and the °C or °F floor is in real temperature units, and a spec written "2% of full scale or 2 °C, whichever is greater" behaves identically to "±2 °C or 2% of reading" only when the floor is interpreted as an absolute, not a span-based, value [S2][S6].
Comparable specs across sensor families
Thermocouple Type K instruments in the portable class typically ship with ±2 °C or 0.75% RD, RTD (Pt100) handheld units with ±0.3 °C + 0.2% RD, and infrared pyrometers in the 0-1000 °C class with ±2 °C or 2% RD, all using the greater-wins convention [S2][S6].
For pressure sensor work, a 0.05% RD spec is the high-end norm for calibration-grade transducers, while 0.1-0.25% FS is typical of industrial pressure transmitter catalogues, and a 0.5% RD unit is the budget tier for HVAC and tank-level use [S2].
For flow, the comparable ladder runs 0.05-0.1% RD (Coriolis, fiscal), 0.5-1.0% RD (magnetic, vortex, ultrasonic), and 2-5% FS (turbine, orifice, older rotameters), with the flow meter class picked by the trade-off between turndown, installed cost, and the actual % of reading error the plant can absorb [S3][S6].
Engineers writing procurement specs should require the manufacturer to state the spec format explicitly, demand the floor and proportional terms be listed separately, and reject datasheets that bury the term under phrases like "typical" or "nominal," because the difference between ±2 °C or 2% RD and ±2% FS is the difference between a usable low-end measurement and a spec that only holds at one point on the dial [S2][S6].
The next signal worth tracking: IEC 60751 (RTD) and IEC 60584 (thermocouple) tolerance class updates through the IEC technical committees, and the broader move in process instrumentation toward % RD-only specs that drop the floor term entirely, which keeps accuracy tight at the top of the range but transfers offset responsibility to the calibration shop.
See also our earlier report, EL and Visual Inspection Gates in PV Module Production: Defect Coverage and Acceptance.