Temperature sensors are the most widely deployed measurement elements in industry, converting thermal energy into an electrical signal that a controller can read. The four dominant families are resistance temperature detectors (RTDs, typically platinum Pt100 or Pt1000), thermocouples, thermistors, and semiconductor IC sensors, each governed by its own international standard and each best suited to a specific window of temperature, accuracy, and cost.
This guide treats the bare sensing element and the field assembly together. Where signal conditioning, 4-20 mA output, and HART communication are integrated into the head, the assembly becomes a temperature transmitter, the form most common on a factory purchase order. The physics and selection logic below apply to both.
This guide is aimed at industrial purchasing engineers and design engineers. It covers 6 chapters from sensing principles, sensor families, accuracy classes and spec-sheet decoding to selection decisions, with 7 selection FAQs and manufacturer comparisons. All parameters reference the IEC 60751 (RTD), IEC 60584 (thermocouple), ASTM E1112 and ISO 80601-2-56 public standards, cross-checked against manufacturer datasheets from WIKA, Endress+Hauser, Texas Instruments, and Optris.
Chapter 1 / 06
What is a Temperature Sensor
A temperature sensor is a device that converts the temperature of a body, fluid, or surface into a measurable electrical signal. Temperature is one of the four fundamental process variables in automation, alongside pressure, flow, and level, and it is by a wide margin the most frequently measured: nearly every furnace, reactor, motor bearing, transformer, food line, and HVAC plant carries one or more temperature points. A qualified industrial temperature sensor must hold a repeatable and traceable relationship between temperature and output over its full service life, despite vibration, thermal cycling, and chemical attack on its sheath.
Functionally, a temperature sensor has three layers: (1) the sensing element, where a physical property changes with temperature, by electrical resistance in an RTD or thermistor, by thermoelectric voltage in a thermocouple, or by a semiconductor junction's behaviour in an IC sensor; (2) the mechanical assembly, comprising a protection sheath or thermowell, insulation, and a connection head that isolates the element from the process medium; and (3) optional signal conditioning, which linearizes, compensates, and converts the raw signal into a standardized 4-20 mA, HART, or digital bus output. When the third layer is integrated, the assembly is called a temperature transmitter.
The history of electrical temperature measurement begins in 1821, when Thomas Johann Seebeck observed that a circuit of two dissimilar metals produces a voltage when its junctions are at different temperatures, the thermoelectric effect that underpins every thermocouple. In 1871 William Siemens proposed using the resistance of platinum to measure temperature, the basis of the modern RTD; Hugh Longbourne Callendar's careful platinum-resistance work in the 1880s made it a practical instrument. Samuel Ruben patented the metal-oxide thermistor in 1930. Semiconductor band-gap temperature sensors arrived in integrated-circuit form in the 1970s and now sit inside almost every CPU, battery pack, and smartphone.
In terms of measurement span, temperature sensing covers an enormous range: cryogenic platinum and special thermocouples reach below minus 250 degrees Celsius near the boiling point of liquid helium, while type B platinum-rhodium thermocouples and radiation pyrometers measure metal melts above 1,700 degrees Celsius. No single technology spans this whole range with good accuracy. The essence of selection is matching the temperature window, the required accuracy, the response speed, and the chemical and mechanical environment to a specific sensing principle and a specific mechanical build.
Four engineering attributes determine the quality of an industrial temperature sensor: accuracy and its tolerance class, long-term stability or drift, response time, and the environmental ruggedness of the sheath and junction. As with pressure instruments, the cheapest sensor is rarely the lowest total cost of ownership: a low-grade thermocouple that drifts several degrees per year, or an RTD whose 2-wire lead resistance is never compensated, can quietly bias a control loop and waste energy or product long before it visibly fails.
Chapter 2 / 06
Sensor Families and Classification
Industrial temperature sensors split into four electrical families plus the non-contact radiation category. Each family occupies a distinct band of temperature range, accuracy, linearity, and cost. The single most common selection error is forcing one family outside its natural window, for example pushing an RTD above 600 degrees Celsius where it loses stability, or using a thermocouple for a tightly toleranced sub-100 degree loop where an RTD would be far more accurate. The table below summarizes the trade-offs that drive the family decision.
Family
Typical Range
Typical Accuracy
Linearity
Relative Cost
RTD (Pt100 / Pt1000)
-200 to +850 °C
±0.15 °C (Class A)
Near-linear
Medium
Thermocouple
-270 to +1,820 °C
±1 to 2.5 °C
Nonlinear
Low to medium
NTC thermistor
-55 to +200 °C
±0.1 °C (in band)
Highly nonlinear
Low
IC / semiconductor
-55 to +155 °C
±0.1 to 0.5 °C
Linear (digital)
Low
Infrared / radiation
-50 to +3,000 °C
±1% of reading
Computed
Medium to high
RTDs exploit the predictable rise in platinum resistance with temperature. The reference type is the Pt100, defined by IEC 60751 as 100.00 ohms at 0 degrees Celsius with a temperature coefficient of 0.003851 per degree (the alpha value) between 0 and 100 degrees. Pt1000 uses the same curve scaled to 1,000 ohms, which makes lead resistance a smaller fraction of the signal and suits long cable runs. RTDs are the accuracy and stability leaders below about 400 to 600 degrees Celsius and are the default for custody-grade and tightly controlled processes.
Thermocouples generate a self-powered voltage from two dissimilar metal legs joined at the measuring junction. They are simple, rugged, fast, inexpensive, and reach far higher temperatures than any RTD, which makes them the workhorse of furnaces, kilns, gas turbines, and heat treatment. The trade-off is lower accuracy, a nonlinear voltage curve, and the need for cold junction compensation and matched extension wire. IEC 60584 designates the letter types (K, J, T, E, N, R, S, B) by their alloy pairs.
Thermistors are sintered metal-oxide ceramics whose resistance changes steeply and nonlinearly with temperature. NTC (negative temperature coefficient) types dominate measurement; their high sensitivity gives excellent resolution within a narrow band, and their tiny mass gives fast response, but the curve must be linearized in software with the Beta or Steinhart-Hart model. IC sensors integrate the sensing band-gap junction with an analog-to-digital converter and a digital interface (I2C, SPI, 1-Wire) on one chip, delivering a calibrated, linear, directly readable value, ideal for electronics, batteries, and consumer products. Infrared radiation sensors are covered in Chapter 3 as the non-contact case.
Chapter 3 / 06
Sensing Technologies and Principles
Beneath the family labels sit distinct physical principles, each with its own error sources and best-fit envelope. The thermocouple type, in particular, is itself a major selection axis because alloy pair sets both the usable temperature range and the chemical atmosphere the sensor will survive. The table below lists the IEC 60584-1 letter types with their overall range, approximate sensitivity, and Class 1 and Class 2 tolerances; the permissible error at any point is the greater of the fixed degree value or the percentage of measured temperature.
Type
Alloy Pair
Range
Sensitivity
Class 1 Tol.
Class 2 Tol.
K
NiCr-NiAl
-270 to 1,370 °C
~41 µV/°C
±1.5 °C / 0.4%
±2.5 °C / 0.75%
J
Fe-CuNi
-210 to 1,200 °C
~50 µV/°C
±1.5 °C / 0.4%
±2.5 °C / 0.75%
T
Cu-CuNi
-270 to 400 °C
~43 µV/°C
±0.5 °C / 0.4%
±1.0 °C / 0.75%
E
NiCr-CuNi
-270 to 910 °C
~68 µV/°C
±1.5 °C / 0.4%
±2.5 °C / 0.75%
N
NiCrSi-NiSi
-270 to 1,300 °C
~39 µV/°C
±1.5 °C / 0.4%
±2.5 °C / 0.75%
R / S
PtRh-Pt
-50 to 1,760 °C
~10 µV/°C
±1.0 °C
±1.5 °C / 0.25%
B
PtRh30-PtRh6
+20 to 1,820 °C
~10 µV/°C
—
±1.5 °C / 0.25%
Resistance temperature detection uses a platinum wire winding or a thin-film platinum element on a ceramic substrate. The element is excited with a small constant current, and the voltage across it is read as a resistance and converted to temperature through the IEC 60751 Callendar-Van Dusen polynomial. Platinum is chosen because it is chemically stable, available in very pure form, and has a highly repeatable resistance-temperature relationship. The main error sources are lead resistance (solved with 3-wire or 4-wire connection), self-heating from excitation current, and slow long-term drift, which is far smaller than a thermocouple's.
The thermoelectric effect drives thermocouples. Because the legs only produce a voltage proportional to the difference between the measuring junction and the reference junction, the instrument must add back the reference (cold junction) temperature, a step called cold junction compensation. Type K is the general-purpose default; type J must stay below about 760 degrees because an iron magnetic transformation permanently decalibrates it; type N resists the high-temperature drift and green-rot that afflict bare type K; and the noble-metal types R, S, and B serve furnaces above 1,200 degrees. The measuring junction can be grounded (fast response, shares sheath potential), ungrounded or isolated (slower, immune to ground loops), or exposed (fastest, no chemical protection).
Thermistor sensing relies on the steep, exponential resistance change of metal-oxide ceramics. The relationship is captured approximately by the Beta parameter and more precisely by the three-coefficient Steinhart-Hart equation. NTC thermistors deliver plus-or-minus 0.1 degree resolution across roughly minus 55 to 200 degrees Celsius and are cheap and small, which is why they appear in battery packs, appliances, and HVAC. The penalties are strong nonlinearity and a narrow span. Semiconductor IC sensors derive temperature from the temperature dependence of a silicon band-gap voltage. The classic analog LM35 outputs 10 mV per degree Celsius with about plus-or-minus 0.5 degree accuracy across minus 55 to 150 degrees; modern digital parts such as the Texas Instruments TMP117 reach plus-or-minus 0.1 degree across minus 20 to 50 degrees and plus-or-minus 0.3 degree over the full minus 55 to 150 degree span, meeting ASTM E1112 and ISO 80601-2-56 for electronic patient thermometers.
Infrared radiation thermometry is the principal non-contact method. Every object above absolute zero emits thermal radiation whose intensity grows with temperature, per the Planck and Stefan-Boltzmann laws. A radiation thermometer focuses this radiation onto a thermopile or pyroelectric detector and computes temperature. Two settings govern accuracy: emissivity, the surface's radiating efficiency (near 0.95 for most painted or oxidized surfaces, very low for polished metal), which must be entered correctly; and the distance-to-spot (D:S) ratio, which fixes the measured spot size at a given range. Industrial pyrometers usually work in the 8 to 14 micrometre atmospheric window to avoid water-vapour and CO2 absorption, and reach well above 1,000 degrees Celsius without touching the target.
Chapter 4 / 06
Standards, Sheaths and Junctions
Two standards families define almost every industrial temperature sensor on the market. IEC 60751 specifies the Pt100 and Pt1000 resistance curve and the RTD tolerance classes (AA, A, B, C), including the rule that any element finer than Class B must use 3-wire or 4-wire connection. IEC 60584-1 specifies the thermocouple reference functions and Class 1, Class 2, and (for some types) Class 3 tolerances, all expressed on the International Temperature Scale of 1990 (ITS-90); IEC 60584-3 fixes the extension and compensating-cable colour codes. In North America, ANSI/ASTM E230 plays the thermocouple role with Special and Standard Limits of Error that roughly correspond to IEC Class 1 and Class 2.
The RTD tolerance classes deserve a closer look because their error grows linearly with temperature. The table below states the IEC 60751 Pt100 tolerance formula for each class and the resulting error at three reference temperatures. Class AA (formerly 1/3 DIN) is the tightest mass-produced grade; Class B is the industrial default; Class C is used where economy matters more than precision.
RTD Class
Tolerance Formula
Error at 0 °C
Error at 100 °C
Error at 450 °C
Class AA
±(0.10 + 0.0017·t)
±0.10 °C
±0.27 °C
±0.87 °C
Class A
±(0.15 + 0.002·t)
±0.15 °C
±0.35 °C
±1.05 °C
Class B
±(0.30 + 0.005·t)
±0.30 °C
±0.80 °C
±2.55 °C
Class C
±(0.60 + 0.01·t)
±0.60 °C
±1.60 °C
±5.10 °C
The protection sheath and thermowell isolate the fragile element from the process medium and carry the bulk of the mechanical and chemical duty. The most common sheath material is 316 or 316L stainless steel, suitable for water, steam, air, and mild process media up to roughly 800 degrees Celsius. Inconel 600 extends service to about 1,150 degrees and resists oxidizing and reducing atmospheres at high temperature, which is why it is standard for furnace type K and type N assemblies. For corrosive or molten-metal duty, ceramic protection tubes (alumina or silicon carbide) protect noble-metal type R, S, and B couples above 1,300 degrees, while tantalum or Hastelloy sheaths handle aggressive acids. A separately mounted thermowell lets the sensor be replaced without breaching the process, the standard arrangement for pressurized pipelines.
Insertion length and time response are governed by how the element couples to the medium. As a practical rule, the sensitive tip should be immersed to at least ten times the sheath diameter so that heat conducted away along the stem (stem loss) does not bias the reading low. Response time, usually quoted as the time to reach 50 percent or 90 percent of a step change (t50 or t90), depends on junction type and sheath mass: a bare exposed thermocouple junction responds in well under a second, a grounded mineral-insulated junction in a few seconds, and a thick thermowell assembly in tens of seconds. Adding thermal grease or a snug thermowell fit improves coupling and shortens t90.
The measuring junction configuration is a defining choice for sheathed thermocouples. A grounded junction is welded to the sheath tip, giving the fastest sheathed response but tying the signal to sheath (and possibly earth) potential, which can introduce ground loops. An ungrounded or isolated junction floats inside the sheath, immune to ground loops and preferred where the sensor connects to single-ended electronics, at the cost of slower response. An exposed junction protrudes from the sheath for the fastest possible response but offers no chemical or mechanical protection and suits only clean, dry, non-pressurized gas.
Chapter 5 / 06
Key Specification Parameters
Reading a temperature-sensor datasheet is a core purchasing skill. A given assembly may list dozens of parameters, but only a handful truly drive the selection decision: sensor type and curve, temperature range, accuracy or tolerance class, response time, long-term stability, connection method, output signal, and the sheath and process connection. Each is explained below.
Sensor type and curve come first because they fix the readout electronics. A datasheet must state, for example, Pt100 to IEC 60751 with alpha 0.003851, or thermocouple type K to IEC 60584-1; a transmitter or DCS input card must be configured for the exact curve or the reading will be systematically wrong. Pt1000 is increasingly preferred over Pt100 for long cable runs because its higher base resistance reduces the relative effect of lead resistance.
Accuracy and tolerance class are quoted per the relevant standard, and the engineer must remember that the tolerance grows with temperature. An RTD Class A is plus-or-minus 0.15 degrees at 0 degrees but plus-or-minus 1.05 degrees at 450 degrees; a Class 2 type K is plus-or-minus 2.5 degrees up to 333 degrees and then 0.75 percent of reading, reaching about plus-or-minus 7.5 degrees at 1,000 degrees. Always compute the tolerance at the actual operating temperature, not at room temperature, and add the transmitter and cold-junction-compensation errors to get the true loop accuracy.
Response time (t50 / t90) matters in fast control loops, safety trips, and surface measurement. Long-term stability, expressed as drift in degrees per year, separates families sharply: a quality Pt100 drifts well under 0.1 degree per year, while a base-metal thermocouple in a hot oxidizing furnace can drift several degrees per year and needs periodic verification. Self-heating applies to RTDs and thermistors: the excitation current dissipates power in the element, raising its own temperature, so the datasheet's dissipation constant (milliwatts per degree) must be respected in low-flow gas or vacuum.
Output signal is the interface to the control system. The mainstream options are:
Raw element (mV or ohms): the bare thermocouple voltage or RTD resistance, wired to a DCS or PLC temperature input card with on-card cold junction compensation and linearization.
4-20 mA / 4-20 mA + HART: a head-mounted or rail transmitter converts the element to a robust current loop that tolerates long cable runs; HART overlays digital configuration and diagnostics.
Digital bus: Foundation Fieldbus, PROFIBUS PA, or Ethernet-APL for large DCS projects with many points on a shared pair.
Digital IC interface: I2C, SPI, or 1-Wire from an integrated sensor such as the TMP117, used in electronics and OEM products for a directly readable, calibrated value.
Connection method is a critical RTD specification: 2-wire (lead resistance becomes error, only acceptable for short runs or Pt1000), 3-wire (one lead compensated, the industrial standard), or 4-wire Kelvin (lead resistance fully cancelled, reference grade). Process connection, sheath material, and ingress protection (thread, flange, or sanitary clamp; 316L, Inconel, or ceramic; IP65/IP67/IP68 head) round out the mechanical specification and must match the medium and mounting described in Chapter 4.
Chapter 6 / 06
Selection Decision Factors
To turn the preceding chapters into a specific model, follow the ordered decision sequence below. Most selection mistakes come not from a single wrong answer but from deciding accuracy or output before the temperature range and atmosphere have been pinned down. These eight steps work as a fixed RFQ template.
Temperature range and atmosphere: establish the minimum, normal, and peak temperatures, plus the surrounding gas (oxidizing, reducing, inert, vacuum). This alone narrows the family: below 400 degrees Celsius favour an RTD or thermistor; above 600 degrees a thermocouple; for furnaces above 1,200 degrees a noble-metal type R, S, or B or an infrared pyrometer.
Required accuracy and tolerance class: compute the allowable error at the actual operating temperature, not at room temperature. Choose RTD Class AA, A, B, or C, or thermocouple Class 1 or Class 2 accordingly, and confirm the transmitter and cold-junction errors still leave margin.
Response time and dynamics: for fast loops, safety trips, or surface work, specify junction type (grounded, ungrounded, exposed) and sheath diameter to meet a t90 target; for slow bulk processes, prefer a protected thermowell assembly.
Sheath, thermowell and process connection: select sheath material per the medium (316L, Inconel 600, ceramic, tantalum) and the connection (thread G1/2 or NPT1/2, flange DN, or Tri-Clamp for sanitary), with insertion length at least ten sheath diameters.
Wiring and output signal: for RTDs choose 3-wire or 4-wire; decide between raw element, head-mounted 4-20 mA / HART transmitter, digital bus, or an integrated IC interface to match the control system.
Certifications and ingress protection: hazardous areas need ATEX, IECEx, or NEPSI; safety loops need SIL2 or SIL3 reports; sanitary lines need 3-A or EHEDG; outdoor or washdown heads need IP67 or above.
Environmental and mechanical stress: account for vibration (gas turbines, motors), thermal cycling, and stem-loss conduction; specify a vibration-rated assembly per IEC 60068-2-6 where needed.
Total cost of ownership (TCO): purchase price plus installation, periodic verification, spare elements, and downtime. A drifting thermocouple that biases a furnace setpoint a few degrees can waste far more in energy or scrap over its life than the price gap to a stable, properly compensated assembly.
One often-overlooked dimension is serviceability and interchangeability: whether a failed element can be replaced through a thermowell without a process shutdown, whether the chosen RTD class guarantees interchangeability without per-unit calibration, and whether the supplier maintains local calibration laboratories and spare-parts inventory. WIKA, Endress+Hauser, Emerson, Yokogawa, ABB, and Honeywell offer broad process portfolios with calibration support; Texas Instruments, Analog Devices, and Sensirion supply digital IC sensors; and Optris, Fluke Process Instruments, and KEYENCE cover infrared pyrometry. These service factors decide repair response time long after the purchase order is closed.
FAQ
What is the difference between an RTD and a thermocouple?
An RTD (resistance temperature detector) measures temperature through the predictable resistance change of a platinum element, while a thermocouple generates a small voltage from two dissimilar metals joined at a measuring junction (the Seebeck effect). An RTD is more accurate and stable: a Class A Pt100 under IEC 60751 is about plus-or-minus 0.15 degrees Celsius at 0 degrees, versus plus-or-minus 1.5 degrees for a Class 1 type K thermocouple. A thermocouple covers a far wider range (type K reaches roughly 1,370 degrees Celsius, type B near 1,820 degrees) and responds faster with low mass. RTDs are limited to about 600 to 850 degrees Celsius. Choose an RTD below 400 degrees for accuracy, a thermocouple above 600 degrees for range.
What does the Pt100 accuracy class actually mean?
Pt100 accuracy is defined by IEC 60751 tolerance classes whose error grows with temperature. Class AA is plus-or-minus (0.10 + 0.0017 times t) degrees Celsius, Class A is plus-or-minus (0.15 + 0.002 times t), Class B is plus-or-minus (0.30 + 0.005 times t), and Class C is plus-or-minus (0.60 + 0.01 times t), where t is the absolute temperature in degrees Celsius. So a Class A element is about plus-or-minus 0.15 degrees at 0 degrees but plus-or-minus 1.05 degrees at 450 degrees. IEC 60751 also requires that any element better than Class B be wired in 3-wire or 4-wire configuration so lead resistance does not corrupt the reading.
Why does a 3-wire or 4-wire RTD connection matter?
An RTD reads temperature as a resistance, so the resistance of the connecting leads adds directly to the measurement. A 2-wire connection includes the full lead resistance as error: even a few ohms of copper can equal several degrees. A 3-wire connection lets the transmitter subtract one lead's resistance, removing most of the error and the dominant choice in industrial loops. A 4-wire (Kelvin) connection forces a known current through two leads and senses voltage through the other two, cancelling lead resistance entirely, and is used for laboratory and reference-grade measurement. IEC 60751 mandates 3-wire or 4-wire for any element finer than Class B.
Which thermocouple type should I choose?
Type K (NiCr-NiAl) is the general-purpose default, roughly minus 270 to 1,370 degrees Celsius with about 41 microvolts per degree sensitivity, good in oxidizing atmospheres. Type J (Fe-CuNi) suits reducing atmospheres but cannot exceed about 760 degrees because the iron leg undergoes a magnetic transformation that permanently decalibrates it. Type T (Cu-CuNi) is excellent for cryogenic and sub-zero work down to minus 270 degrees. Type N (NiCrSi-NiSi) resists drift and green-rot better than K at high temperature. Types R, S and B are noble-metal platinum-rhodium couples for furnaces above 1,200 degrees, with type B usable to about 1,820 degrees. Match the type to atmosphere and peak temperature first, then accuracy.
What is cold junction compensation and why is it required?
A thermocouple only generates voltage from the temperature difference between its measuring (hot) junction and its reference (cold) junction at the instrument terminals. To recover an absolute temperature, the instrument must know the cold-junction temperature and add it back. This is cold junction compensation (CJC). Historically the reference junction was held in an ice bath at exactly 0 degrees Celsius; modern transmitters measure the terminal-block temperature with a built-in RTD or thermistor and compensate electronically. CJC error is a major real-world error source: a 1 degree error in measuring the terminal block becomes a 1 degree error in the reading. Use proper thermocouple-grade extension wire all the way to the CJC point, never ordinary copper.
How do non-contact infrared sensors measure temperature?
Every object above absolute zero emits infrared radiation whose intensity rises with temperature. An infrared (radiation) thermometer focuses that radiation with optics onto a thermopile or pyroelectric detector, converts it to a signal, and computes temperature from the Planck and Stefan-Boltzmann relations. Two parameters dominate accuracy: emissivity, the surface's ability to radiate (0 for a mirror, near 0.95 for most painted or oxidized surfaces), which must be set correctly or the reading is wrong; and the distance-to-spot (D:S) ratio, which fixes how large the measured area is at a given distance, for example 20:1 gives a 25 mm spot at 500 mm. Industrial pyrometers commonly work in the 8 to 14 micrometre band to minimize water-vapour and CO2 interference, and reach well above 1,000 degrees Celsius without contact.
Which manufacturers and standards apply to industrial temperature sensors?
RTD elements follow IEC 60751 (Pt100/Pt1000 tolerance classes and resistance tables), thermocouples follow IEC 60584-1 (reference functions and Class 1/2/3 tolerances) and IEC 60584-3 (extension-wire colour codes), and patient-grade electronic thermometers follow ASTM E1112 and ISO 80601-2-56. Established suppliers include WIKA, Endress+Hauser, Emerson Rosemount, Yokogawa, ABB and Honeywell for process assemblies and transmitters; Texas Instruments, Analog Devices and Sensirion for digital IC sensors such as the TMP117; and Optris, Fluke Process Instruments and KEYENCE for infrared pyrometers. For hazardous areas confirm ATEX, IECEx and NEPSI certification, and SIL2 or SIL3 functional-safety reports for safety loops.