Temperature sensor selection is driven by five engineering gates: required accuracy, operating temperature range, response time, environmental exposure, and the signal the downstream temperature controller or temperature transmitter can accept [S3][S4][S6].
The market spans four dominant families: platinum resistance temperature detectors (Pt100/Pt1000 RTDs), thermocouples (Types K, J, T, N, S, R, B), semiconductor IC sensors (LM35, DHT11, on-chip CMOS), and non-contact infrared units such as the Campbell Scientific IRTS-P [S3][S4][S9]. Australian custom manufacturer D&N Engineering advertises parallel build capability for thermocouples, RTDs, thermowells, and extension cables, signalling that buyers can source mixed-principle inventories from a single fabricator [S9].
Selection criteria ranked by engineering weight
Accuracy drives the shortlist first: the LM35 IC sensor delivers ±0.5°C guaranteed at 25°C and is rated across 55–150°C, with a linear +10 mV/°C scale factor that removes the offset math required by older Kelvin-calibrated devices [S4]. The IRTS-P infrared unit specifies ±0.3°C from -10°C to 55°C and tightens to ±0.1°C when sensor body and target are at the same temperature, with 0.05°C repeatability [S3]. Industrial Pt100 Class A RTDs typically deliver ±0.15°C at 0°C; this is the benchmark most process temperature measurement loops are designed around [S3][S4].
Range and environment come second. Thermocouples extend coverage to 1800°C (Type K up to ~1260°C in oxidising atmospheres, Type S/R to 1480°C, Type B to 1700°C) and tolerate vibration that would fatigue an RTD's helix; RTDs cap reliably near 600°C because of platinum wire drift and sheath contamination. The IRTS-P is constrained to -10–55°C outdoor operation with a 6–14 µm wavelength silicon lens and a 3:1 field of view, limiting it to canopy, leaf, and surface-average work rather than furnace duty [S3]. For submerged or wet-service lines, the Global Water WQ101 ships in stainless steel or PVC housing on marine-grade cable up to 500 ft with a 4–20 mA two-wire output, illustrating how industrial packaging constrains sensor choice as much as the sensing element [S6].
Side-by-side comparison: four main sensor families
Comparison of the four main sensor families on the criteria that drive a buy decision:
1. Platinum RTD (Pt100/Pt1000): accuracy ±0.1–0.3°C typical (Class A ±0.15°C @ 0°C); range -200 to ~600°C; output resistance change requiring 3- or 4-wire lead compensation; best for temperature monitor loops in chemical, pharmaceutical, and HVAC plants; weak point is vibration-induced lead fracture and self-heating at high excitation currents.
2. Thermocouple (K, J, T, N, S, R, B): accuracy ±1–2°C for base-metal types, ±0.5–1°C for noble-metal types; range -200 to 1800°C depending on type; output low-voltage mV requiring cold-junction compensation; best for furnace, exhaust, and high-temperature process work; weak point is drift at elevated temperature and sensitivity to lead-wire uniformity.
3. Semiconductor IC (LM35, DHT11, on-chip CMOS): accuracy ±0.5–2°C; range -55 to 150°C typical; output analogue voltage (LM35, 10 mV/°C) or digital single-wire (DHT11); best for PCB, battery, IoT, and HVAC air-handling; weak point is narrow range and self-heating in SMD packages [S4].
4. Infrared non-contact (IRTS-P class): accuracy ±0.3°C (-10 to 55°C, ±0.1°C matched-body); range limited by detector wavelength (6–14 µm longwave for ambient work); output mV signal emulating a Type K thermocouple; best for moving targets, surfaces that must not be contacted, and spatial averaging; weak point is emissivity dependence and operating-temperature window of the detector housing [S3].
Output signal, wiring, and integration with controllers
Output signal is the gate between sensor and temperature controller. The LM35 outputs a linear +10 mV/°C analogue voltage between +V_s (4–30 V) and GND, read on an Arduino analogue pin with 10-bit resolution at a 5 V reference (1 LSB ≈ 4.88 mV ≈ 0.488°C) [S4]. The IRTS-P outputs two Type K thermocouple mV signals (target + sensor body) on 15 ft twisted shielded pairs, with the body signal feeding compensation for the target reading [S3]. Industrial loop-powered transmitters, such as the WQ101's 4–20 mA two-wire configuration, allow long cable runs to a temperature recorder or DCS without analogue signal loss [S6].
Steam and HVAC control loops that integrate electric control valves typically pair a Pt100 or thermocouple input to a PID-capable controller with PI or PID control modes, remote setpoint, temperature compensation, over-temperature alarm, and heat-metering functions; most valves remain manually operable on power loss, which is a safety advantage in process heating [S2]. For sites that also depend on flow regulation through the same control valves, the electric temperature-control valve functions as a flow regulator at the heat-exchanger primary side, modulating opening against a temperature feedback signal [S1].
Calibration, hazardous-area, and environmental constraints
Calibration is non-negotiable for safety-class loops. The Chinese military standard GJB 4044-2000 governs calibration methods for platinum resistance temperature sensors used in missiles, while national metrological technical specifications cover industrial and laboratory practice [S8]. Field calibration typically uses a dry-block or stirred-liquid bath with a reference standard traceable to national metrology institutes; the comparison test for an RTD involves three wire configurations (2-, 3-, 4-wire) at 0°C ice bath and at upper-range points.
Hazardous-area selection adds another filter. For refineries and chemical plants, intrinsic-safety (IS) or explosion-proof (Ex d) certified heads are mandatory; the buyer must verify ATEX, IECEx, or regional equivalents against the zone classification, gas group, and temperature class. Buyers in such service are well-served by the parallel selection logic used for thermocouple selection for hazardous-area temperature points, where sheath material, hot-junction construction, and approval markings are inseparable from the sensing element.
Environmental exposure drives housing and sheath decisions. Submerged applications need welded 316L stainless steel or titanium thermowells; chemical service may require Hastelloy or PTFE-coated probes; outdoor meteorological use, as on the IRTS-P, demands a water-resistant housing rated for continuous outdoor exposure across -10 to 55°C [S3][S6]. Buyers who already follow a temperature measurement loop-by-loop methodology will recognise the same five-gate filter (accuracy, range, response, environment, output) applied to other instruments in the spirit level vs displacement sensor vs process temperature selection logic.
Who should NOT pick the default Pt100 RTD
The default Pt100 RTD is the wrong choice when any of three conditions appear. First, target surface temperature must be measured without physical contact, for example a moving web, rotating drum, or hazardous surface; an infrared sensor is the only practical fit, and the buyer should evaluate a longwave 8–14 µm detector with adjustable emissivity setting [S3]. Second, the process exceeds 600°C or drops below -200°C, in which case a Type K, Type N, or Type S thermocouple extends range at the cost of accuracy and cold-junction compensation complexity. Third, the application is a battery-powered IoT node, a microcontroller thermal-flag input, or a PCB-mounted monitor; here an LM35, TMP36, DHT11 (which also delivers relative humidity on a single digital line), or an on-chip CMOS thermal sensor reduces BOM and firmware load [S4][S5][S7].
On-chip CMOS sensors, as described in the RF power-dissipation measurement literature, integrate directly into the die and can observe circuit performance through DC-output thermal coupling, a method used for built-in self-test of RF front-ends [S5]. That capability is irrelevant to a process plant but central to a telecom or RF power-amplifier buyer, illustrating how the same question (which sensor?) lands on different families depending on the duty.
Sourcing, cost bands, and shortlist logic
Industrial Pt100 probes with stainless-steel sheath and 4–20 mA transmitter head commonly list in the 200–800 USD range; rugged water-quality probes such as the Global Water WQ101 list at 685 USD with marine-grade cable up to 500 ft and a 4–20 mA two-wire output [S6]. Thermocouple assemblies for general process use sit lower, with consumer-grade Type K probes and pilot-generator thermopiles on general-merchandise channels in the 20–30 USD band, but these are not equivalent to a mineral-insulated Type K with a calibrated transmitter [S10]. Custom shops such as D&N Engineering in Australia build to spec with parallel lines for RTDs, thermocouples, thermowells, and extension cables, an option worth requesting for non-standard sheath lengths, fittings, or connector terminations [S9].
Shortlist logic for a process buyer: define the accuracy budget first (e.g. ±0.5°C → Pt100 or IC; ±2°C → base-metal thermocouple acceptable); fix the temperature range (this eliminates three of four families); confirm the output type the existing temperature controller accepts (analogue V, mV, 4–20 mA, or digital bus); verify the environmental rating (IP, IS/Ex, sheath material); and only then compare supplier lead times and calibration certificates. Where the duty is broad and the buyer lacks a standard, the temperature sensor selection matrix above is the most defensible starting point.