Glossary term
Gain Error
Engineering definition of gain error covering span error, sensitivity error, scale-factor error, calibration slope, full-scale impact and release evidence.
Definition
metricGain error is the error caused by an incorrect measurement slope, sensitivity or scale factor.
Gain error makes the reported value grow too quickly or too slowly as the measurand increases. It is often called span error, sensitivity error or scale-factor error. Unlike a constant offset, gain error is small near zero and grows across the range. It can come from calibration slope, amplifier gain, excitation voltage, ADC reference, ratiometric mismatch, sensor sensitivity or software scaling.
Gain error is the error caused by an incorrect measurement slope, sensitivity or scale factor. It makes a reported value grow too quickly or too slowly as the measurand increases.
Gain error is different from offset error. Offset error can dominate near zero. Gain error grows with measured value and often matters most near the upper part of the range.
Slope Error
If a calibration uses indicated slope (m_i) and the reference slope is (m_r), fractional gain error can be written as:
The percent gain error is:
The same idea applies to sensor sensitivity (S), amplifier gain, ADC scale factor or software conversion coefficient.
Sensitivity Error
For a sensor specified by sensitivity, the same calculation can be written as:
where (S_i) is indicated or fitted sensitivity and (S_r) is reference sensitivity. This form is useful for load cells, pressure transducers, accelerometers, photodiodes and bridge sensors.
Resulting Measurement Error
For a measurement value (x), the approximate gain-related error is:
This is why gain error is often less important near zero but important near full scale.
Worked Example
A pressure transmitter should have reference slope:
Calibration finds:
The fractional gain error is:
so:
At (x=8.0\ \text{bar}), the gain-related error is approximately:
If the allowed error is (0.10\ \text{bar}), the gain error alone is too large near this operating point.
Offset and Gain Together
A simple linear measurement error model is:
where (b) is offset error. Two instruments can have the same error at one point but very different behavior across the range. A single zero check cannot prove gain accuracy, and a single high-span check cannot isolate offset.
Relation to Linearity Error
Gain error changes the slope of the best linear relation. Linearity error describes residual curvature or departure from that chosen line. If all points lie on a wrong-slope line, the main issue is gain error. If no straight line fits the points well, the issue is nonlinearity or model form.
That distinction matters because gain error can often be corrected with a scale factor, while nonlinearity may require a narrower range, piecewise correction, different sensor physics or design change.
Ratiometric Systems
In bridge and ADC systems, ratiometric measurement can reduce some gain error. If the same excitation or reference affects both the sensor output and ADC conversion, slow reference drift may cancel.
The cancellation is not automatic. Amplifier gain drift, bridge self-heating, cable resistance, reference distribution, firmware scaling and non-ratiometric auxiliary channels can still create span error.
Diagnosis
Gain error should be separated from offset, nonlinearity and saturation. A zero check mainly tests offset. A span check tests scale. Several intermediate points test whether one slope is adequate. A saturation check confirms that the apparent slope change is not caused by clipping, range limiting or common-mode violation.
If the same percentage error appears across the range, gain correction may be appropriate. If error changes sign, bends, or appears only near one end, linearity error or saturation is more likely. If the gain changes with temperature, time or excitation, the problem is drift or reference control rather than a one-time calibration coefficient.
Software Scaling
Modern measurement systems often hide gain error in firmware coefficients, unit conversion, ADC reference assumptions or digital filtering. A corrected calibration certificate is not enough if the deployed software uses an old coefficient or a different engineering-unit scale.
The release record should therefore connect the physical calibration to the active coefficient set, firmware version, data-acquisition range and displayed engineering units.
Evidence for Release
A useful gain-error review states the slope or sensitivity reference, calibration points, fitted coefficient, reference standard, uncertainty, range, temperature, excitation, signal-conditioning gain, ADC reference, software coefficient and whether the installed configuration matches the calibration setup.
For safety, quality or clinical decisions, gain error should be checked near the operating point that controls the decision, not only over a convenient laboratory range.
Limits and Common Mistakes
Common mistakes include checking only zero, correcting offset while leaving span error, using datasheet gain accuracy without installed verification, changing amplifier gain after calibration, assuming ADC bit depth proves scale accuracy and ignoring excitation drift in non-ratiometric systems.
Another mistake is treating gain error and linearity error as interchangeable. A wrong slope, curved response and saturated front end need different corrective actions and different release evidence.