How is the accuracy of datalogger devices ensured?

How to build confidence in your recorded measurements

Reliable readings from a data logger require a suitable sensor, correct wiring, appropriate settings and traceable calibration. A calibration sticker cannot establish that the complete installation measures correctly. Follow the measurement chain from the probe to the report, and identify where errors can enter. Begin by defining the measurement error your application can accept.

Do not confuse resolution with accuracy

Resolution concerns the smallest change an instrument can distinguish or display. Testo's 174 T USB-C version (0572 1740 01), for example, specifies 0.1 °C resolution and ±0.5 °C accuracy over −30 to +70 °C. A display that advances in tenths of a degree does not establish ±0.1 °C accuracy. Nor does a steady reading prove correctness: a measurement can be highly repeatable while retaining a systematic offset.

Start with the probe and its installation

The sensor type must suit both the input and the quantity being measured. For Pt100 measurements, lead resistance contributes error in a two-wire connection; a suitable three- or four-wire measurement arrangement can reduce that contribution. Thermocouples need the correct type, appropriate extension wiring, correct polarity and cold-junction compensation. Pt100 probes and thermocouples require different input configurations, while an NTC sensor needs its correct resistance–temperature characteristic.

Placement matters too. A humidity probe next to warm electronics may not represent the surrounding air, and a pressure sensor measuring the wrong pressure reference will not answer the application's question. Check thermal contact, immersion depth, response time and environmental stability where relevant. Calibration cannot fix an installation that measures a different quantity from the one you need.

Understand what calibration does

Calibration establishes the relationship between reference values and instrument indications, with associated uncertainties, under specified conditions. Adjustment changes the instrument's response when necessary. Verification checks whether defined requirements are met. A calibration does not automatically include adjustment or establish a pass result. If the response is adjusted, assess the resulting performance through recalibration.

Use a reference with a documented calibration chain and uncertainty suitable for the task. A higher purchase price does not make a device an adequate reference. Check the certificate's instrument and probe identities, channels, test points, results and uncertainty. Injecting an electrical signal into the logger checks its input path; it does not test the installed sensor. Where needed, evaluate the sensor and recorder together.

Compare against a reference at relevant points

Use an appropriate comparison method with the probe and reference exposed to the same stable conditions. Allow the specified warm-up time and time for the sensor to settle. Choose several points covering the working range and important decision thresholds. Calculate error = recorded value − reference value. A reference value of 20.00 °C and a recorded value of 20.40 °C give an error of +0.40 °C at that point. This does not establish a valid correction across the entire range.

Uncertainty is different from error. The reference, repeatability, resolution and test conditions can contribute to comparison uncertainty. Do not decide acceptability from the observed difference alone. Suppose the error is +0.40 °C, expanded comparison uncertainty is 0.20 °C and the acceptance limit is ±0.50 °C. A conservative rule requiring |error| + uncertainty ≤ limit would not accept that result: 0.60 exceeds 0.50. This illustrates one decision rule; establish the rule appropriate to your application before testing.

Check scaling before trying to correct the instrument

Consider a 4–20 mA pressure transmitter configured for 0–10 bar. Its ideal linear mapping is 4 mA = 0 bar, 12 mA = 5 bar and 20 mA = 10 bar. If the recorder is mistakenly configured for 0–20 mA, it may interpret 4 mA as 2 bar. During configuration, verify input type, range, units and scaling coefficients. With Modbus/RS485 data, also check the register, data type, byte/word order and scaling.

A smooth trend can conceal a poor measurement

Route signal cables appropriately away from power wiring, and follow the manufacturer's shielding and grounding instructions. Where supported, suitable mains-frequency rejection can reduce interference. But heavy filtering or a long recording interval can hide short excursions. For live monitoring, data analysis and storage planning, distinguish measurement frequency from recording frequency and understand how averages, minima and maxima are generated. Compare the values shown in SCADA with exported records as well.

Use performance history to set recalibration intervals

There is no single interval suitable for every logger. Consider manufacturer guidance, required accuracy, workload, environmental exposure and previous drift, alongside any requirements applicable to the task. During maintenance, check the probe, connections, battery and power supply. Review the interval if intermediate checks show increasing deviation. An impact, repair, probe replacement or excessive exposure may also justify checking performance again.

Technical references: JCGM: calibration and adjustment · NIST: traceability and uncertainty · ILAC/OIML: recalibration intervals · Testo 174 T USB-C specifications · Tektronix: RTD measurements · Campbell Scientific: thermocouple measurements