A two-term accuracy specification written as "±(a% of reading + b% of range)" is the dominant format on modern pressure, temperature and multifunction process calibrator datasheets, and it directly states the worst-case error at every point along the scale [S1][S2].
The first term scales with the value being measured, so a 100 Pa reading on a 0.01% of reading instrument carries only 0.01 Pa of proportional error, while a 100,000 Pa reading on the same instrument carries 10 Pa. The second term is a fixed offset tied to the full scale or the selected range, and it dominates when the reading is small [S1][S2].
Why two terms, not one: the mathematics behind the spec
Single-term "% of full scale" or "% of range" gives a flat absolute error across the entire range, so 0.05% FS on a 0 to 200 kPa module is a constant ±0.1 kPa whether you are reading 5 kPa or 195 kPa; as a fraction of the reading, that is 2% at the low end and 0.05% at the top [S5].
Single-term "% of reading" is the mirror image, with the error tracking the input proportionally but ignoring offset drift, so a 0.05% of reading module at 5 kPa reads 0.0025 kPa, yet the same module is rarely specified for a fixed floor of noise or zero stability [S2][S5].
Combining the two as "±(a% of reading + b% of range)" preserves the proportional behavior of the reading term and adds a guaranteed offset term from the range, which is why measurement laboratories prefer it: at full scale the total error is a% + b%, and at 10% of range it is 0.1·a% + b% [S1][S5].
Decoding the spec sheet, term by term
Manufacturers almost always publish the reading term first because it is the smaller, marketing-friendly number, with a typical high-end pressure calibrator showing 0.01% of reading plus 0.015% of full scale, and a mid-range hand-held showing 0.05% of reading plus 0.05% of range [S5].
The "range" in the second term has at least three common definitions that an engineer has to identify before any comparison: nominal or selection range (100% of the rms value of the active range, what the user picks on the dial), effective range (the span over which the published spec is actually valid), and maximum measurable peak value (used by some power analyzers rather than process instruments) [S1].
For a pressure calibrator the second term is most often full scale of the installed sensor module, so changing modules changes the offset term even if the reading term stays the same; the same logic applies to a dry block temperature calibrator, where the range is usually the block's rated span in degrees C [S5].
Worked example, three common spec formats side by side

Take a 0 to 200 kPa pressure module, then read 50 kPa (one quarter of full scale) on three hypothetical calibrators: A is 0.05% FS only, B is 0.05% of reading only, C is 0.025% of reading plus 0.025% FS [S5].
Calibrator A: absolute error is 0.05% × 200 kPa = 0.1 kPa flat, which is 0.2% of the 50 kPa reading, so the relative error at quarter scale is four times worse than at full scale [S5].
Calibrator B: absolute error is 0.05% × 50 kPa = 0.025 kPa, which is 0.05% of the reading, but there is no guaranteed floor; a 0.1 kPa zero shift on a 5 kPa test gives 2% error that the spec does not capture [S2][S5].
Calibrator C: absolute error is (0.025% × 50 kPa) + (0.025% × 200 kPa) = 0.0125 + 0.05 = 0.0625 kPa, or 0.125% of the 50 kPa reading, which is the best of the three at quarter scale and the same as B at full scale [S5].
Who the format is for, and who should avoid it
The two-term spec fits calibration work where the Test Uncertainty Ratio (TUR) must stay above 4:1 across the working range: instrument shops validating 4 to 20 mA loop calibrator measurements, deadweight tester cross-checks where the deadweight tester reference is itself two-term, and ISO/IEC 17025 labs that have to propagate both components through their uncertainty budget [S1][S3].
It is less useful for field troubleshooting where a single worst-case number is enough, for very high static pressures close to full scale where the offset term is negligible anyway, and for sensors where zero can be re-set immediately before every reading, because the range term is then double-counted and a "% of reading only" spec is honest about what remains [S2][S5].
Engineers comparing two datasheets should also confirm the range reference: a "0.02% of reading + 0.01% of range" spec on a 0 to 1000 mA module is not directly comparable to the same numbers on a 0 to 100 mA module, because the offset term scales with the chosen range, not the reading [S1][S5].
Limits, pitfalls, and what the spec does not cover

The two-term spec is normally given for reference conditions of 20 to 23 °C, 1 year stability period and a defined warm-up, so field use at 5 °C or 40 °C, or after 2 years since calibration, will exceed the published number; the temperature coefficient and long-term drift must be added separately to the uncertainty budget [S1][S3].
It also does not cover hysteresis between ascending and descending pressure, repeatability on cycling, and the extra error introduced when a signal calibrator sources mA/V while simultaneously reading, because sourcing and measuring share internal references whose cross-loading is not always included in the headline spec [S3][S5].
For accredited work, the published accuracy has to be combined with the calibration standard's own uncertainty into a Test Uncertainty Ratio, and the conventional rule of thumb is that the calibrator should be 4× more accurate than the device under test, with 10× preferred, so a 0.1% transmitter tested against a 0.025% reading + 0.025% range calibrator still passes the 4× rule only at the top of the transmitter's range [S3].
Sourcing and standards context
The "% of reading + % of range" wording is consistent across Yokogawa's power-analyzer guidance, AMETEK's pressure calibration knowledge base, Beamex's pressure calibrator explainer and Fox Valley Metrology's multimeter treatment, and it maps directly onto the way manufacturers publish spec sheets for temperature calibration bath units, decade resistance box standards and most digital multimeters [S1][S2][S3][S5].
No single ISO or IEC standard defines this format in the way IEC 61298 defines process instrument test procedures, so engineers should treat it as a manufacturer convention, read the small print for the range reference, and propagate the two terms separately into any GUM-style uncertainty calculation rather than summing them as one worst-case number [S1][S5].
Trackable signals to watch in the next calibration cycle: any move by major calibrator vendors to publish the two terms as separate rows rather than a single bracketed expression, which would simplify TUR calculations, and any update to ISO/IEC 17025 guidance that formally recognises the reading-plus-range format as a documented basis of statement, both of which are visible in vendor knowledge bases today but not yet in the standard itself [S1][S3][S5].