Sensor Accuracy: Understanding Deviation Between True and Measured Values

ISO 5725-1 defines accuracy through 2 distinct components: trueness and precision. Trueness is the closeness of agreement between the mean of a large series of measured results and the true value. Precision is the closeness of agreement between independent results obtained under stipulated conditions.

A target analogy illustrates the distinction. A sensor that consistently hits the same point on the target shows high precision, whether or not that point is the bullseye. A sensor that hits the bullseye on average across many measurements shows high trueness. Accuracy, by the ISO 5725-1 definition, requires both: results that are tightly grouped and centered on the true value.

In industrial sensor applications, the distinction is practically significant. A 3D laser profile sensor with high precision but low trueness repeats its measurements reliably, yet all results carry a systematic offset from the true surface position. Calibration addresses trueness; hardware stability and noise reduction address precision.

Measurement error is the difference between the measured value and the true value of the measurand, expressed as:

e=xmeasured−xtruee=xmeasured​−xtrue​

3 standard forms of measurement error exist in metrology: absolute error, relative error, and percentage error.

Absolute error expresses the deviation in the same unit as the measurand — for example, ±15 µm for a 3D sensor measuring a surface profile. Relative error expresses the deviation as a ratio to the true value, expressed as erel=e/xtrueerel​=e/xtrue​. Percentage error expresses relative error as a percentage of the true value. Industrial sensor data sheets report absolute error most frequently, as it directly communicates the physical magnitude of deviation to manufacturing engineers.

The relationship between error and accuracy is inverse: as measurement error decreases, accuracy increases. Calibration reduces systematic error components; robust sensor design reduces both systematic and random contributions. Measurement uncertainty — a broader statistical concept defined by the GUM (Guide to the Expression of Uncertainty in Measurement) — captures the full range of possible error contributions and is the applicable metric for formal metrology assessments.

Systematic errors are measurement errors that remain constant or change predictably under repeated measurements under the same conditions. A sensor with a systematic error of +20 µm reports every measured value 20 µm above the true value. 4 common sources of systematic errors in industrial sensors include offset errors, gain errors, nonlinearity, and thermal drift.

Offset error is a constant deviation of the output signal from the true value across the entire measurement range. Gain error is a proportional deviation that scales with the magnitude of the measurand. Nonlinearity describes a systematic deviation whose magnitude varies with the measured value in a non-proportional pattern. Thermal drift describes a systematic shift in output caused by temperature changes within the sensor’s operating range.

Systematic errors are detectable and correctable. Calibration — the process of comparing sensor output against a traceable reference standard — identifies and compensates systematic deviations. Regular calibration against certified reference artifacts is the standard correction method for systematic errors in industrial measurement systems.

Random errors are measurement errors that vary in an unpredictable manner across repeated measurements under identical conditions. Unlike systematic errors, random errors cannot be corrected by calibration. 3 primary sources of random errors in optical and thermal sensors include electronic noise, mechanical vibration, and electromagnetic interference.

Electronic noise arises from thermal agitation of charge carriers in the detector and signal conditioning circuits. It produces fluctuating output values even when the measurand is constant. Mechanical vibration from the production environment introduces random positional uncertainty between the sensor and the measurement object. Electromagnetic interference from drives, switching power supplies, and RF sources couples into the sensor signal path and adds unpredictable signal components.

Random errors are characterized statistically by their standard deviation. Reducing random errors requires averaging multiple measurements, improving the signal-to-noise ratio through hardware design, or applying digital filtering. The precision specification in a sensor data sheet directly characterizes the magnitude of random error under defined test conditions.

Gross errors are measurement errors of exceptional magnitude that fall outside the expected statistical distribution of a measurement process. 3 common causes of gross errors include sensor malfunction, incorrect setup, and unexpected object surface conditions that exceed the sensor’s operating specifications.

Gross errors appear as outliers in measurement data sets and distort mean values and standard deviation calculations when included without filtering. Outlier detection methods — including the 3-sigma rule, Grubbs’ test, and interquartile range filtering — identify and exclude gross errors from measurement data before analysis. In automated industrial inspection systems, gross error detection is integrated into the measurement software as a data quality gate that prevents outlier-contaminated results from entering process control logic.

Accuracy and resolution are independent parameters that describe different aspects of sensor performance. Accuracy describes the closeness of a measurement result to the true value of the measurand. Resolution describes the smallest change in the measurand that the sensor can detect and report as a distinct output value.

A 3D laser profile sensor with a resolution of 1 µm detects surface height differences as small as 1 µm. If that sensor carries a systematic offset error of 50 µm, every reported value is 50 µm from the true surface position, despite the high resolution. The sensor resolves fine detail precisely but measures it inaccurately.

The relationship also operates in the opposite direction: a highly accurate sensor with coarse resolution reports the correct mean value across a measurement series but cannot detect small deviations from nominal. Both parameters are required to characterize sensor suitability for a given measurement task.

Repeatability is the closeness of agreement between successive measurement results for the same measurand, obtained under the same conditions of measurement within a short period. High repeatability means the sensor produces the same output value on repeated measurements of the same object. This corresponds to high precision in the ISO 5725-1 model.

Accuracy requires both high repeatability and high trueness. A sensor can exhibit high repeatability and low accuracy simultaneously when it consistently produces the same incorrect value — a situation characterized by low trueness and high precision. In manufacturing quality control, repeatability is a necessary but not sufficient condition for accuracy: a sensor must first be repeatable before calibration can establish its accuracy.

Temperature is the single most significant environmental factor affecting the accuracy of industrial sensors. 2 distinct temperature effects degrade sensor accuracy: the thermal offset drift and the thermal gain drift. Thermal offset drift shifts the zero-point of the output signal with temperature changes. Thermal gain drift scales the sensitivity of the sensor proportionally with temperature changes.

In 3D laser profile sensors, thermal expansion of the optical bench, lens assembly, and detector array introduces position-dependent errors that grow with temperature deviation from the calibration reference temperature — typically 20 °C per ISO 1 standard for dimensional measurements. In infrared cameras, detector sensitivity and dark current both vary with temperature, directly affecting radiometric accuracy.

Sensor specifications report thermal accuracy coefficients as ±X µm/K or ±X °C/K, expressing the accuracy change per Kelvin of ambient temperature deviation. Operating a sensor within its specified temperature range minimizes thermal accuracy degradation.

Object surface characteristics introduce 3 accuracy-relevant effects in optical 3D measurement: specular reflection, subsurface scattering, and low signal return. Each effect alters the optical signal received by the sensor and introduces deviations between the detected surface position and the true surface geometry.

Specular surfaces — including polished metals, chrome-plated parts, and mirror-like glass — reflect the laser line of a triangulation sensor away from the detector, causing signal dropout or reduced signal intensity that degrades positioning accuracy. Translucent and subsurface-scattering materials — including plastics, ceramics, and biological tissue — allow laser light to penetrate the surface and scatter from sub-surface layers, producing a broadened peak in the triangulation signal that shifts the detected surface position toward the interior of the material.

Low-reflectivity surfaces — including black rubber, carbon fiber, and matte-coated parts — return insufficient optical energy to the detector, increasing noise and reducing the accuracy of peak detection. AT Sensors addresses these surface-dependent accuracy effects through HDR measurement modes that adapt the sensor’s exposure settings within a single measurement cycle, and through multi-slope linearization that maintains peak detection accuracy across surfaces of widely varying reflectivity.

Mechanical vibration introduces random positional uncertainty between the sensor and the measurement object during data acquisition. In a laser triangulation sensor capturing a cross-sectional profile, vibration at a frequency above the sensor’s profile rate introduces random lateral and axial displacements that appear as profile noise and accuracy degradation.

3 installation factors determine vibration-induced accuracy degradation: the vibration frequency spectrum relative to the sensor’s measurement rate, the amplitude of displacement in the measurement axis, and the mechanical coupling path between vibration sources and the sensor mount. Direct coupling through rigid machine frames transfers vibration efficiently from drives, presses, and conveyor systems to the sensor.

Rigid mounting on a structure mechanically isolated from vibration sources — using vibration-damping elements, separate mounting frames, or granite surface plates in precision applications — reduces vibration-induced accuracy errors. Vibration resistance specifications define the sensor’s structural integrity limits, not the measurement accuracy under vibration.

Signal processing quality determines the effective accuracy of the measurement result extracted from the sensor’s raw output signal. 3 signal processing parameters directly affect accuracy: peak detection algorithm precision, averaging depth, and digital filter cutoff frequency.

In laser triangulation sensors, the detector captures a light intensity distribution across the image sensor. The sub-pixel peak detection algorithm determines the position of the laser line centroid with accuracy finer than one pixel pitch. More advanced algorithms, including weighted centroid and Gaussian fitting, achieve sub-pixel accuracy of 1/10 to 1/100 of a pixel pitch under optimal signal conditions.

Averaging multiple consecutive measurements reduces random noise and improves effective accuracy at the cost of dynamic response speed. Digital low-pass filtering attenuates high-frequency noise components in the output signal. The selection of averaging depth and filter settings requires balancing accuracy improvement against the required measurement bandwidth for the application.

Absolute accuracy, expressed as ±X µm or ±X °C, states the maximum permissible deviation between the sensor’s output and the true value across the specified measurement range under defined reference conditions. A 3D sensor with an accuracy specification of ±10 µm guarantees that no measured value deviates more than 10 µm from the true surface position under the specified test conditions.

Accuracy as a percentage of full-scale output (%FSO) expresses the maximum deviation as a fraction of the total measurement range. A sensor with a 100 mm measurement range and 0.1 %FSO accuracy has a maximum absolute deviation of 0.1 mm across the full range. This format is common in displacement and distance sensors where the absolute deviation scales with the range.

Accuracy as a percentage of reading (%RDG) expresses the maximum deviation as a fraction of the currently measured value. A 0.5 %RDG accuracy specification on a thermal sensor measuring 200 °C implies a maximum deviation of ±1 °C at that reading. %RDG specifications are common in infrared temperature measurement because detector response and atmospheric absorption both scale with the measured temperature. Data sheets report either peak accuracy (maximum single-point deviation) or RMS accuracy (root mean square deviation across the range), with peak accuracy being the more conservative and more frequently reported value.

The RMS deviation across the measurement range is expressed as:

e_RMS = √[(1/n) · Σ (x_i − x_true,i)²], for i = 1 to n

Measurement uncertainty and accuracy are related but distinct concepts in metrology. Accuracy describes the agreement between a measured value and the true value. Measurement uncertainty, defined by the GUM (JCGM 100:2008), quantifies the range of values that can reasonably be attributed to the measurand, combining all identified systematic and random error contributions into a single statistical estimate.

A sensor data sheet accuracy specification is a manufacturer-defined performance guarantee under specified test conditions. A measurement uncertainty budget is a site-specific calculation that accounts for the sensor’s accuracy plus all additional uncertainty contributions — including fixturing, environmental conditions, reference standard uncertainty, and operator effects. Formal measurement uncertainty analysis per GUM is the applicable methodology for calibration laboratories and accredited quality systems.

Comparing accuracy specifications across sensor types requires 3 conditions to be matched: the measurement range must be equivalent, the reference conditions must be identical, and the accuracy metric format must be the same. A ±10 µm accuracy specification on a sensor with a 10 mm range is not equivalent to ±10 µm on a sensor with a 300 mm range — the second sensor achieves the same absolute accuracy over 30 times the range.

Traceable calibration is the technical basis for valid accuracy comparison. Calibration performed with reference standards traceable to national metrology institutes — PTB in Germany, NPL in the UK, NIST in the USA — ensures that accuracy claims from different manufacturers use a common measurement reference. ISO/IEC 17025 accreditation of the calibration laboratory provides formal assurance of traceability.

Laser triangulation sensors determine 3D surface coordinates by measuring the position of a projected laser line on a CMOS or CCD image sensor. The accuracy of the Z-axis (distance) measurement depends on 5 parameters: the triangulation angle, the focal length of the receiving optics, the image sensor pixel pitch, the sub-pixel peak detection algorithm performance, and the surface reflectivity of the measurement object.

The triangulation angle — the angle between the laser projection axis and the camera axis — is the primary geometric determinant of Z-axis accuracy and resolution. Larger triangulation angles produce higher Z-axis sensitivity but introduce greater sensitivity to surface tilt and shadowing effects. AT Sensors designs triangulation angles to balance Z-axis accuracy against measurement robustness across the surface types encountered in industrial inspection applications.

Z-axis accuracy specifications for industrial laser profile sensors range from ±1 µm for short-range precision sensors to ±100 µm for long-range sensors with measurement ranges exceeding 500 mm. The DIN EN ISO 10360-10 standard provides test procedures for evaluating the accuracy of optical 3D measurement systems under defined conditions.

The fundamental geometric relationship governing Z-axis sensitivity in triangulation is:

Δz=ptan⁡(α)Δz=tan(α)p​

where $\Delta z$ is the Z-axis measurement increment, $p$ is the pixel pitch of the image sensor, and $\alpha$ is the triangulation angle.

Infrared cameras measure surface temperatures by detecting the thermal radiation emitted by objects across the 7–14 µm wavelength range for long-wave infrared detectors, or the 2–5 µm range for mid-wave detectors. Temperature measurement accuracy in infrared cameras is specified in 2 formats: absolute accuracy (±X °C or ±X K) and percentage-of-reading accuracy (±X % of reading), with the larger of the 2 values typically applying across the measurement range.

4 physical factors determine infrared camera temperature measurement accuracy: detector sensitivity expressed as NETD (Noise Equivalent Temperature Difference), emissivity correction accuracy, atmospheric transmission compensation, and the accuracy of the blackbody reference used during factory radiometric calibration. NETD characterizes the minimum temperature difference the camera can resolve, with values of 20–80 mK being typical for uncooled microbolometer detectors used in industrial applications.

Emissivity — the ratio of radiation emitted by a surface to radiation emitted by an ideal blackbody at the same temperature — is the largest source of temperature measurement error in practice. Industrial surfaces exhibit emissivities ranging from 0.05 for polished aluminum to 0.98 for oxidized steel. Incorrect emissivity settings in the camera’s compensation model introduce systematic temperature errors. AT Sensors infrared cameras provide adjustable emissivity settings and multi-spectral measurement options to minimize emissivity-related accuracy errors across the material types encountered in industrial thermographic inspection.

Industrial operating environments introduce 4 accuracy-degrading conditions not present in laboratory calibration settings: elevated ambient temperature, airborne contamination, electromagnetic interference, and thermal radiation from neighboring hot objects. Each condition introduces accuracy deviations that compound with the sensor’s base accuracy specification.

Ambient temperatures above the calibration reference temperature of 20 °C activate thermal drift mechanisms in both 3D sensors and infrared cameras. Airborne contamination — including oil mist, metal particles, and condensation — deposits on protective windows and optical surfaces, attenuating optical signals. Signal attenuation increases measurement noise and introduces apparent distance and temperature offsets. Regular cleaning of optical surfaces in contaminated environments maintains accuracy within specification.

Electromagnetic interference from servo drives, induction heaters, and welding equipment couples into sensor electronics and output signal lines, adding noise to measurement data. EMV/ESD-compliant installation practices — shielded cables, proper grounding, and physical separation from interference sources — maintain sensor accuracy in electromagnetically dense production environments. AT Sensors specifies the EMV/ESD immunity levels of its sensors per IEC 61000-4 test standards.