Thermocouple vs RTD vs Thermistor: Which Temperature Sensor Is Right for Your Application?

Why the sensor type matters beyond the temperature range

When engineers specify a temperature sensor, the first question is usually range. Can this sensor measure at the temperatures my process operates? That question is necessary, but it is far from sufficient. A sensor that survives your operating range but drifts unpredictably, requires a measurement circuit your instrument cannot provide, or generates a signal your data logger cannot interpret is not a workable choice regardless of its rated range.

The three main sensor families (thermocouples, RTDs, and thermistors), differ fundamentally in their operating principle, output signal, accuracy, long-term stability, and calibration complexity. Each of those differences has direct implications for how you specify, install, maintain, and audit the measurement system.

In Singapore's industrial landscape, the stakes are particularly concrete. For manufacturers operating under ISO 9001 quality management systems, the calibration certificates for critical temperature instruments must carry stated measurement uncertainty to satisfy clause 7.1.5 of the 2015 standard. For pharmaceutical manufacturers regulated by HSA GMP, temperature monitoring instruments at critical process steps must be calibrated with evidence traceable to a national standard, and the calibration uncertainty must be appropriate for the tolerance of the process being monitored. For food businesses operating HACCP plans, temperature instruments at critical control points must be calibrated regularly, with records available for SFA and third-party food safety audits.

Choosing the wrong sensor type does not just affect measurement quality. It creates calibration complexity, introduces systematic errors that drift unpredictably, and can create a situation where the instrument cannot realistically achieve the measurement uncertainty your compliance framework demands. Understanding the differences before you specify saves rework, failed qualifications, and audit findings downstream.

How thermocouples work, and why they drift

A thermocouple works on the Seebeck effect: when two dissimilar metals are joined at one end and exposed to a temperature gradient, a small voltage is generated proportional to that temperature difference. The measurement end (the hot junction), is placed in the process. The other end (the cold junction), connects to the measuring instrument. The instrument measures the voltage and, knowing the cold junction temperature through a separate sensor, calculates the absolute temperature at the hot junction.

The Seebeck effect produces a very small signal. Typically in the range of 40 to 60 microvolts per degree Celsius for common thermocouple types. This small signal is one of the reasons thermocouples are inherently less accurate than RTDs: any noise, contact resistance, or cold junction error represents a significant fraction of the measurement signal. The non-linear relationship between voltage and temperature also requires compensation tables (defined in standards like IEC 60584) and well-characterised cold junction compensation in the measuring instrument.

Despite these limitations, thermocouples dominate high-temperature industrial applications because they are robust, inexpensive, available in a wide range of materials and constructions, and self-powered, they require no excitation current. The standard thermocouple types and their typical ranges are:

• Type K (Chromel-Alumel): −200°C to +1,260°C. The most widely used thermocouple type globally. Good general-purpose performance across a wide range. At temperatures above 800°C, undergoes a reversible magnetic transformation (the "Green Rot" phenomenon) that can cause drift in oxidising atmospheres. Most suitable for continuous measurements up to about 1,000°C in clean environments.
• Type T (Copper-Constantan): −200°C to +350°C. Excellent performance at cryogenic and low temperatures. The copper leg has excellent thermal conductivity, making Type T particularly well-suited for food processing, pharmaceutical cold chains, and laboratory freezer monitoring. Commonly used in HACCP applications because it performs reliably in the −40°C to +120°C range typical of food cold chain operations.
• Type J (Iron-Constantan): 0°C to +750°C. Common in older industrial equipment, particularly in the US market. The iron wire is susceptible to rust below 0°C and at high temperatures in moist atmospheres. Not recommended for new installations. Type K is preferred where the range overlaps.
• Type N (Nicrosil-Nisil): −200°C to +1,300°C. Developed specifically to address the drift mechanisms of Type K at high temperatures. More stable than Type K above 800°C because its alloy composition is chosen to reduce grain boundary effects and eliminate the magnetic transformation. An increasingly preferred choice for high-temperature industrial applications requiring long-term stability.
• Types S, R, and B (Platinum-Rhodium alloys): These noble-metal thermocouples are used for very high temperature applications. Type S and R to approximately +1,600°C, Type B to +1,820°C. They are used in glass melting, precious-metal processing, and high-temperature furnace qualification. High cost and fragility limit them to applications where base-metal types are insufficient.

Thermocouple drift is a real and important limitation for precision applications. Three mechanisms drive it. First, grain growth: at elevated temperatures, the crystal structure of the thermoelectric alloys coarsens over time. This changes the Seebeck coefficient and shifts the calibration. The rate of grain growth (and therefore the rate of drift), increases sharply with temperature. A Type K thermocouple used continuously at 1,000°C may drift several degrees Celsius per hundred hours of use. The same sensor at 200°C may be stable for years. Second, contamination: foreign elements from the protection sheath, insulation, or process atmosphere can diffuse into the thermocouple wire and alter its thermoelectric properties. Silicon contamination of Type K sensors in certain atmosphere furnaces is a well-documented example. Third, mechanical stress from vibration, thermal cycling, or physical distortion introduces heterogeneity in the wire that causes the Seebeck coefficient to vary along its length. Generating spurious voltages in the measurement signal. For these reasons, thermocouples used in high-temperature or vibration-intensive environments should be calibrated at intervals appropriate to the drift rate, not just on a fixed annual schedule.

How RTDs work, and why they are the accuracy standard

Resistance Temperature Detectors (RTDs) operate on a different and more stable physical principle: the electrical resistance of pure metals increases predictably and almost linearly with temperature. The relationship between resistance and temperature for a platinum RTD is defined by the Callendar-Van Dusen equation and standardised in IEC 60751. This standardisation means that a PT100 sensor from any compliant manufacturer has a known, predictable characteristic. The resistance at 0°C is 100Ω, and the sensitivity is approximately 0.385Ω per degree Celsius. This predictability and the relative linearity of the response are the reasons RTDs achieve accuracy that thermocouples cannot match in the common industrial range.

Platinum is the preferred metal for precision RTDs because of its chemical stability, reproducible and stable resistance-temperature relationship, and resistance to contamination. The standard variants are:

• PT100: 100Ω nominal resistance at 0°C. The dominant industrial standard. The relatively low resistance means lead wire resistance must be managed carefully , 2-wire connections introduce lead resistance error, making 3-wire and 4-wire configurations the standard for precision measurement.
• PT1000: 1000Ω nominal resistance at 0°C. The higher resistance means the lead resistance error is proportionally smaller as a fraction of the total signal, making PT1000 better suited to long cable runs or 2-wire configurations where 3- or 4-wire is impractical. Signal-to-noise is better, but compatible instruments are less common than for PT100.

The connection method used to measure an RTD has a substantial impact on accuracy. In a 2-wire configuration, the measured resistance includes the resistance of both lead wires, which introduces a positive offset error, approximately 0.1Ω per metre of copper lead wire, equivalent to roughly 0.26°C per metre round trip. For a sensor 10 metres from its instrument, this is a 2.6°C systematic error that is fixed and cannot be corrected without knowing the lead resistance precisely. The 3-wire configuration partially compensates for lead resistance by using one extra wire to measure one lead and subtracting it from the total. The 4-wire (Kelvin) configuration forces a known current through the sensor via two wires and measures the voltage drop across the sensor via two separate wires (completely eliminating lead resistance error. For any precision measurement application), pharmaceutical temperature mapping, laboratory instruments, calibration reference , 4-wire connection is the standard. For process control where ±0.5°C is sufficient, 3-wire PT100 is commonly used.

RTDs also have meaningful limitations. Above 600°C, platinum becomes fragile and susceptible to contamination from metallic vapours and reducing atmospheres, which is why thermocouples take over at high temperatures. RTDs are also more susceptible to vibration failure than thermocouples because the fine platinum resistance element is relatively delicate mechanically. Self-heating (the small amount of heat generated by the excitation current passing through the resistive element), must be managed by using appropriately low excitation currents, particularly in still-air environments where the generated heat cannot dissipate quickly. For a well-designed measurement system with appropriate excitation current, self-heating error is typically below 0.05°C and often negligible.

The long-term stability of platinum RTDs is exceptional. A well-constructed, undamaged PT100 installed in a benign environment may drift less than 0.05°C per year. This stability is why RTDs are the sensor of choice in any application where long-term measurement reliability is critical. Including GMP temperature mapping, where sensors must remain calibrated and within specification across multi-year validation life cycles.

How thermistors work, and where they belong

Thermistors are semiconductor ceramic devices whose resistance changes with temperature. Unlike the near-linear characteristic of platinum RTDs, thermistors exhibit a highly non-linear, exponential relationship between resistance and temperature. The most common type (NTC (Negative Temperature Coefficient)), shows a sharp decrease in resistance as temperature rises. This steep response curve is the thermistor's key advantage: a small change in temperature produces a large change in resistance, giving extremely high sensitivity compared to either thermocouples or RTDs. A typical NTC thermistor has a temperature coefficient of around −4% per degree Celsius, compared to +0.385% per degree Celsius for a PT100.

This sensitivity advantage is real and useful in specific applications. In HVAC control loops, where the sensor is controlling a system to maintain a temperature within ±0.5°C of a set point, the thermistor's large signal swing makes it easy for simple analogue control circuits to respond quickly and accurately to small deviations. In medical devices (patient body temperature monitors, incubators, blood analysers), the high sensitivity in the 20°C to 45°C range relevant to human physiology is precisely what is needed. In consumer electronics, thermistors protect batteries and power components by providing a cost-effective thermal protection signal that triggers protection circuits before damage occurs.

The non-linearity is also the thermistor's greatest limitation for precision measurement applications. Converting the measured resistance to a temperature requires a non-linear mathematical model. Most commonly the Steinhart-Hart equation, which uses three empirically determined coefficients characterised for each individual sensor or sensor lot. Using a single sensitivity coefficient (as would be adequate for a near-linear RTD) introduces significant errors across anything but a very narrow temperature span. This mathematical complexity is manageable in microcontroller-based systems, but it means calibration requires multiple calibration points (at least three, and ideally five or more for good accuracy across the operating range), rather than the two-point calibration that is often sufficient for a near-linear RTD.

Self-heating is a more significant concern for thermistors than for RTDs because the high sensitivity cuts both ways: the high resistance change per degree means that even a small amount of resistive heating from the excitation current can shift the reading. This requires either very low excitation currents (which reduces signal-to-noise) or careful characterisation and correction of self-heating effects. In still air or low-flow environments, self-heating errors of 0.1–0.5°C are not uncommon if the excitation current is not carefully controlled.

Thermistors are not well-suited to harsh industrial environments. Their glass or epoxy encapsulation is less rugged than the stainless-steel sheathing common on industrial RTDs and thermocouples. They are not the right choice for elevated temperatures (above approximately 150°C for standard devices) or for any environment involving mechanical shock, vibration, or chemical exposure. Their role in industrial settings is narrowly but usefully defined: temperature monitoring in controlled environments where their sensitivity advantage is genuinely needed and their limitations are managed.

Full specification comparison

The table below summarises the key properties of the three sensor types side by side. The rows most relevant to application selection are operating range, accuracy, linearity, and GMP/HACCP suitability; the rows most relevant to measurement system design are connection type, self-heating, and calibration complexity.

The operating range row determines which sensor types are even candidates for a given application. Once range is satisfied, accuracy and linearity determine whether the sensor can achieve the measurement uncertainty your process requires. The GMP/HACCP suitability row reflects the regulatory reality in Singapore: PT100 RTDs are what auditors expect to see at critical measurement points in regulated industries, and using a different sensor type requires explicit justification in the qualification documentation.

When to use a thermocouple

Thermocouples are the right choice when the application places demands that RTDs and thermistors cannot satisfy. Primarily high temperature, harsh environments, or applications where the lower cost or self-powered nature of the thermocouple outweighs its accuracy limitations.

The clearest cases are high-temperature industrial processes. Industrial furnaces, kilns, glass melting tanks, and heat treatment equipment regularly operate above 600°C, and often above 1,000°C. Ranges where RTDs are not viable. For furnace temperature profiling (mapping the spatial temperature distribution within a furnace during a qualification run), thermocouples are the only practical choice. Type K is the workhorse for most industrial furnace applications up to about 1,000°C continuous use. Type N is increasingly preferred where greater stability above 800°C is needed. For very high temperature furnaces or precious metal processing, Types S, R, or B are appropriate.

Thermocouples also excel in mechanically demanding environments. Exhaust temperature monitoring on engines and turbines, temperature measurement inside rotating equipment, and sensors in high-vibration machinery all benefit from the thermocouple's simple, robust construction. A welded metal junction that has no fragile winding or ceramic substrate to fracture. A mineral-insulated metal-sheathed (MIMS) thermocouple is one of the most mechanically durable temperature sensors available.

Fast transient temperature measurement is another thermocouple strength. A fine-wire thermocouple with a bare exposed junction can respond to temperature changes in milliseconds. Useful for monitoring fast thermal events like ignition, quenching, or short-cycle process equipment where the thermal time constant of an RTD would miss the peak. The response time advantage disappears in sheathed and immersion-well configurations, where the thermal mass of the protection equipment dominates response time regardless of sensor type.

Cost-sensitive large-scale deployments (monitoring dozens or hundreds of points in a manufacturing plant where ±2°C accuracy is sufficient for process control), may appropriately use thermocouples where the per-sensor cost difference compounds across the installation. Where the accuracy of a thermocouple is within the process tolerance requirement, the economics frequently favour it.

The key constraint to remember when specifying thermocouples: accuracy of ±1–2°C is the realistic expectation after calibration, not ±0.3°C. If your process requires better than ±1°C, or if you are qualifying a validated process under GMP, the thermocouple is unlikely to provide sufficient accuracy to carry the measurement uncertainty budget.

When to use an RTD (PT100)

RTDs are the standard choice for any application where long-term accuracy, stability, and calibration reliability are the primary requirements, and where the temperature range stays below 600°C. That description covers the majority of precision industrial, laboratory, and regulated-industry temperature measurement in Singapore.

In pharmaceutical manufacturing, the starting point for any critical temperature measurement is PT100 Class A in 4-wire configuration. This is not arbitrary conservatism. It is the sensor type that can realistically achieve the measurement uncertainty needed to demonstrate that a temperature-controlled process step (sterilisation, cold storage, stability chamber monitoring) is operating within its validated limits. GMP Annex 15 qualification and validation studies require temperature mapping with sensors that have known, stated uncertainty, and PT100 RTDs calibrated by an accredited laboratory are the standard mechanism for producing that evidence. When an HSA inspector reviews temperature monitoring for a GMP-regulated process in Singapore, PT100 calibration certificates are what they expect to see. Using a thermocouple for a critical GMP temperature point requires explicit justification, typically reserved for cases where the process temperature genuinely exceeds the RTD's range.

Food manufacturing operating HACCP plans faces similar expectations. At critical control points. Cooking temperatures that must exceed a minimum to achieve log reductions in pathogens, cold storage temperatures that must remain below a maximum to prevent microbial growth. The temperature measurement must be accurate, calibrated, and documentable. SFA food safety audits and third-party certification audits (ISO 22000, FSSC 22000, BRC) review calibration records for CCP instruments. PT100 RTDs, calibrated annually by a SAC-SINGLAS accredited laboratory with stated uncertainty on the certificate, satisfy these requirements cleanly.

Cold chain monitoring (refrigerated warehouses, pharmaceutical cold rooms, vaccine storage), benefits from PT100 accuracy and stability for the same reasons. The consequences of an undetected temperature excursion in a pharmaceutical cold chain are severe; the monitoring system must be reliable enough that when it reports no excursion occurred, that claim is credible. An RTD-based data logging system with annual accredited calibration provides that credibility in a way that a thermocouple-based system cannot.

Environmental chambers (stability chambers, temperature-humidity test chambers, environmental test equipment), are typically specified with PT100 sensors and calibrated at multiple temperature and humidity set points. The uniformity mapping of the chamber (verifying that temperature is consistent throughout the usable volume) requires multiple PT100 probes, all calibrated to the same reference, so that differences between probe readings can be attributed to chamber non-uniformity rather than sensor-to-sensor differences. 21 CFR Part 11 compliance for electronic records in pharmaceutical settings applies to data logging systems for validated processes; the sensor calibration evidence is part of the audit trail.

Precision laboratory instruments (temperature-controlled water baths, ovens, incubators for non-medical use), almost universally use PT100 or PT1000 sensors where accuracy is specified. The combination of near-linear response, high stability, and well-established calibration procedures makes them the natural choice wherever measurement accuracy is a design requirement rather than an afterthought.

When to use a thermistor

Thermistors occupy a specific and genuine niche where their sensitivity advantage in a narrow temperature range is the dominant requirement, and the limitations (non-linearity, narrow range, self-heating susceptibility), are manageable within the application constraints.

HVAC control loops are the most common industrial application. A thermistor placed in an air duct feeding a control system for an air handling unit does not need to measure accurately across a wide range or survive high temperatures. It needs to respond quickly and sensitively to small changes around a set point. Typically somewhere in the 15°C to 30°C comfort zone. The thermistor's steep response curve means the control system can resolve sub-degree deviations easily, enabling tight temperature regulation. The non-linearity is handled by the controller's firmware, which already contains the linearisation curve for the specific thermistor type used.

Medical devices depend on thermistors in applications where compact size, fast response, and high sensitivity in the near-body-temperature range are all critical. Oral and ear thermometers use NTC thermistors to detect the small resistance change corresponding to a body temperature difference of tenths of a degree. The clinical significance of a fever requires resolving 0.1°C reliably, which thermistors can do very cost-effectively in the 35°C to 42°C range. Neonatal incubators use thermistors to maintain the skin temperature of premature infants within a very tight band. Blood analysers use them to confirm reagent temperatures before running assays. In all these cases, the thermistor is being used within its optimal range, by systems designed around its characteristics.

Consumer electronics thermal protection is another dominant thermistor application. Lithium-ion battery management systems use NTC thermistors to monitor cell temperature and cut power when temperatures rise above safe limits. Power supply thermal protection circuits, motor drive temperature monitoring, and laptop CPU thermal throttling all use thermistors because they are cheap, small, and provide a large, easily read signal change that simple analogue comparator circuits can act on without complex signal conditioning. This is an application where the thermistor's calibration accuracy and linearity across a wide range are far less important than its reliable signal output at the protection threshold temperature.

In industrial settings, thermistors occasionally appear in HVAC monitoring for building management systems, in battery room temperature monitoring for UPS installations, and in certain food equipment where they were installed as original equipment. They are not the recommended choice for new industrial measurement installations where the full capabilities of PT100 RTDs are available. But they remain the correct choice for the narrow applications where their sensitivity profile makes them genuinely superior, and understanding when that is the case prevents unnecessary complexity from specifying a more expensive, harder-to-integrate sensor type when a simple thermistor does the job better.

Calibration requirements for each sensor type

All three temperature sensor types drift over time and require periodic calibration against traceable reference standards. The calibration approach, the critical factors that affect calibration quality, and the expected calibration interval differ meaningfully between sensor types.

Calibrating thermocouples

Thermocouples are calibrated by the comparison method: the sensor under calibration and a reference thermometer (typically a calibrated PT100 or a standard platinum resistance thermometer) are placed together in a temperature-stable medium (a dry block calibrator or a liquid bath), at a series of calibration temperatures spanning the operating range of the sensor. At each point, both the reference and the sensor under calibration are read, and the deviation of the sensor from the reference is recorded.

The critical factor in thermocouple calibration is cold junction management. The thermocouple generates a voltage that represents the temperature difference between the hot junction (in the calibration medium) and the cold junction (at the measurement instrument or the calibration system terminal). If the cold junction temperature is not accurately measured and compensated, the calibration result will carry a systematic error equal to the cold junction error. A well-equipped calibration laboratory measures the cold junction temperature with a calibrated reference sensor and applies the appropriate correction. Certificates issued without cold junction temperature information should be treated with caution.

A good thermocouple calibration certificate states: the calibration points, the thermocouple type, the reference instrument used (with its own traceability documented), the cold junction temperature and compensation method, the correction values at each calibration point, and the expanded measurement uncertainty at each point. For thermocouples used in GMP or HACCP-regulated applications, the number of calibration points should span the full range used in the process, not just a single-point check at one representative temperature.

Calibrating RTDs

RTD calibration uses the comparison method in an oil bath or dry block calibrator capable of the required stability and uniformity. The RTD under calibration and the reference instrument are placed in the bath at a series of temperatures, and the resistance of the RTD is measured at each point using a 4-wire connection to eliminate lead resistance effects. The measured resistance at each point is compared to the IEC 60751 standard curve to derive the deviation, or correction, for the sensor at that temperature.

For precision RTD calibration, the measurement instrument used to read the RTD resistance must have appropriate accuracy. A precision resistance bridge or a calibrated precision multimeter with 4-wire measurement capability. The excitation current must be low enough to avoid significant self-heating error in the bath. Bath uniformity (the temperature difference between different positions in the bath where the reference and sensor under calibration are placed) must be small relative to the calibration uncertainty required.

A well-executed RTD calibration at multiple temperature points allows fitting of the Callendar-Van Dusen coefficients for the specific sensor, providing a correction curve rather than a single-point offset. For Class A PT100 sensors used in precision measurement, calibration at a minimum of three points (and ideally five or more), provides a more accurate correction than a single-point offset. The certificate should state the 4-wire measurement resistance at each calibration point, the deviation from the nominal IEC 60751 curve, and the expanded uncertainty including contributions from the reference standard, the bath uniformity, the repeatability of readings, and self-heating.

Calibrating thermistors

Thermistor calibration uses the comparison method at multiple temperature points, as for RTDs. However, because the resistance-temperature curve is highly non-linear, a minimum of three to five calibration points is needed to characterise the Steinhart-Hart equation coefficients for the specific sensor. A single-point or two-point calibration is insufficient to characterise a thermistor across its operating range. The resistance is measured with low excitation current to minimise self-heating, and the reference instrument must be appropriate for the expected accuracy requirement.

The number of calibration points and the mathematical model used for linearisation should be stated in the certificate, along with the uncertainty at each point. Because the sensitivity (resistance change per degree) varies substantially across the temperature range, the measurement uncertainty in degrees Celsius will not be uniform. It will be lowest where the sensitivity is highest (typically near the bottom of the operating range for NTC types) and highest at the extremes.

Calibration intervals and regulatory expectations in Singapore

A 12-month calibration interval is the default for industrial temperature sensors, and is typically sufficient for RTDs used in benign environments and thermocouples used below 500°C. Shorter intervals are appropriate where drift rate is higher: thermocouples used continuously above 800°C, sensors in vibration-intensive environments, or any sensor used in a validated GMP process where the calibration interval is specified in the validation protocol. Longer intervals (up to 24 months), may be defensible for RTDs in stable, clean environments where calibration history shows minimal drift, but any extension should be supported by calibration history data and justified in writing.

Unitest's SAC-SINGLAS accredited scope covers thermocouple calibration (Types K, T, J) and RTD calibration (PT100, PT1000) against NMC-traceable fixed-point references. Certificates issued under the accreditation scope carry stated measurement uncertainty and are accepted by GMP and HACCP auditors in Singapore without supplementary justification of the traceability chain.

Common mistakes when specifying temperature sensors

Experience with industrial temperature measurement in Singapore surfaces five specification errors that consistently cause problems. Either measurement errors that are invisible until an audit, or calibration complexity that was entirely avoidable.

Using Type K in food and pharmaceutical cold chain applications

Type K is the most common thermocouple type, and its availability and familiarity leads engineers to reach for it by default. But in food and pharmaceutical cold chain applications (where the operating range is typically −40°C to +25°C), Type T (Copper-Constantan) is the correct choice. Type T is specifically optimised for low-temperature performance, with better accuracy and stability in the subzero range. More importantly, using a thermocouple at all for pharmaceutical cold chain monitoring, when PT100 RTDs are available and expected by regulators, introduces unnecessary calibration complexity and accuracy limitations. The frequency with which Type K sensors appear in pharmaceutical cold rooms is not evidence that they are appropriate. It is evidence of a specification that was inherited from industrial practice without considering the regulatory context.

Specifying 2-wire PT100 for long cable runs

A 2-wire PT100 with 20 metres of copper cable has a lead resistance of approximately 0.4Ω per conductor, meaning the total lead resistance in the circuit is approximately 0.8Ω. Equivalent to a systematic temperature reading error of approximately 2°C that cannot be corrected without measuring the cable resistance separately. Engineers who specify 2-wire PT100 to save cable costs and then question why their readings are consistently high have made this mistake. For any run longer than a few metres, specify 3-wire or 4-wire PT100. If the instrument cannot accept 4-wire input, change the instrument. The cost of the extra conductors is trivial compared to the cost of the measurement error.

Placing a thermocouple where it will exceed its recommended range

The maximum temperature ratings on thermocouple data sheets represent limits for continuous use, not for brief excursions. Operating a Type K thermocouple continuously at 1,100°C (close to its rated maximum), will cause accelerated grain growth and rapid drift. The calibration interval for a sensor operating near its limit may need to be as short as a few hundred hours of use. Specifying a sensor type appropriate for the actual maximum temperature (Type N rather than Type K above 1,000°C, or Types S/R/B for applications approaching 1,400°C), prevents this accelerated drift.

Ignoring calibration after replacing a sensor in a validated system

A replacement sensor is not interchangeable with the original without recalibration of the complete measurement loop. In a validated GMP system, replacing a PT100 sensor requires calibrating the new sensor and verifying that the complete measurement loop (sensor + cable + transmitter + logger) still reads within the validated acceptance criteria at the relevant calibration points. Skipping this step (on the assumption that a replacement Class A PT100 will automatically perform identically to the one it replaced), is a change control failure that creates a gap in the validation life cycle and is likely to result in an observation during the next GMP inspection.

Using thermistor sensitivity data from the datasheet without calibrating over the actual operating range

Thermistor datasheets provide nominal Steinhart-Hart coefficients for the sensor family. Individual sensors within the same part number vary in their actual coefficients, and this variation may be significant relative to the accuracy required by the application. For HVAC control where ±0.5°C is sufficient, datasheet coefficients are generally adequate. For any application where accuracy matters more (medical devices, pharmaceutical incubators, analytical instruments), each individual sensor (or at minimum each batch) should be calibrated at the temperatures relevant to the application and the actual coefficients determined from the calibration data. Using only the sensitivity specification from the datasheet and assuming it applies to every individual sensor is the thermistor equivalent of assuming every PT100 reads exactly at its nominal curve. An assumption that holds well for Class A RTDs but less well for thermistors.