Pt100 or Thermocouple: Which to Use in a Process

Pt100 or thermocouple: the selection rule

Below 600 °C use a Pt100 resistance thermometer; above 600 °C use a thermocouple. That boundary is not ours, it comes from WIKA technical documentation: platinum resistance sensors are intended for the −200…+600 °C range, while thermocouples win at high temperatures and reach 1700 °C. Everything else here refines that single line, and the most interesting refinement is this: on datasheet tolerance alone, a type K class 1 thermocouple beats a class B Pt100 from 240 °C upward, long before anything close to 600.

A word on the limits of our range. There are no thermocouples in our sensor catalogue; the temperature line is WIKA resistive instrumentation and Omron controllers. So the thermocouple appears here as the comparison arm, not as something to order from us. If your conditions point to a thermocouple, we will say so plainly rather than push a Pt100 because we happen to carry one.

• Different physics. A Pt100 is a passive resistor that has to be fed a measuring current. A thermocouple generates its own voltage from the temperature difference between hot and cold junction.
• Different error sources. For a Pt100 the enemy is lead resistance and self-heating. For a thermocouple it is the cold junction, the compensating cable and alloy ageing.
• Different ageing. Platinum barely drifts. Base-metal thermocouples (K, J, N, E, T) change their characteristic with service hours, and that is a documented effect, not a defect.
• Different response. A thermocouple is faster, but inside a massive thermowell that advantage nearly disappears.

What the standards guarantee, and what they do not

Pt100 per IEC 60751

A tolerance class is not "sensor accuracy". It is a formula in temperature plus a limited range over which that formula is recognised at all. The tighter the class, the shorter the range. A Pt100 is 100 Ω at 0 °C, a Pt1000 is 1000 Ω.

Thin-film is the standard build, and notice how much shorter its range is. Class AA on thin film exists only between 0 and 150 °C. The resistance-temperature relation follows the Callendar-Van Dusen polynomial with A = 3.9083 · 10⁻³ °C⁻¹, B = −5.7750 · 10⁻⁷ °C⁻², C = −4.1830 · 10⁻¹² °C⁻⁴. Put 100 °C into the 0…600 °C expression and you get 138.51 Ω, so the average Pt100 sensitivity is about 0.385 Ω per degree. We use that figure below.

Thermocouples per IEC 60584-1

Here there are only two working classes, 1 and 2; class 3 is specified only for types B and T, and type B has no class 1 at all. The tolerance is stated as "either an absolute value or a percentage of temperature, whichever is greater". Because of that floor the thermocouple looks poor at low temperatures and suddenly good at high ones.

Two practical consequences. First, type K and type N share identical tolerance rows, so choosing between them is decided by alloy durability, not by accuracy. Second, types R, S and B are not available in class 1 as mineral-insulated cable, so "the most accurate thermocouple in a thin cable" is a combination the standard does not offer.

Where the crossover actually sits: tolerance at 100, 250, 450 and 600 °C

Put the temperatures into both sets of formulas. The arithmetic is open, check it yourself.

The crossover is one line of algebra: 0.30 + 0.0050 · t = 1.5 gives exactly t = 240 °C. Below that point a class B Pt100 is tighter than a type K class 1; above it, no longer, because up to 375 °C the thermocouple sits on its ±1.5 °C floor and does not grow. A class A Pt100, on the other hand, wins everywhere it is defined at all: at the 450 °C edge of its range it gives ±1.05 K against ±1.8 °C for type K. Those 450 °C apply to the wire-wound element (W) only: on thin film (F), which is WIKA's standard build, class A ends at +300 °C, so for an off-the-shelf sensor the real class A limit is 300, not 450. Beyond its own limit class A simply does not exist, and the sensor becomes class B automatically.

Accuracy in the real circuit, not on the datasheet

Pt100: lead resistance and 2/3/4-wire connection

One two-wire cable can wipe out a class A rating. Lead resistance adds to sensor resistance, and a Pt100 has 0.385 Ω/°C sensitivity, so every extra ohm in the loop reads as plus 2.6 °C. Three metres of cheap cable there and back, and class A turns into a systematic error larger than its whole tolerance.

• 2-wire. Lead resistance enters the result in full. WIKA documentation explicitly advises against it for Pt100 classes A and AA and limits the use to cables up to 250 mm. For Pt1000, two-wire is the normal arrangement.
• 3-wire. The standard build. Lead influence is compensated as far as possible; practical length is roughly up to 30 m, beyond which deviations start.
• 4-wire. Cable influence is eliminated completely, including asymmetry between the conductors. Laboratory work, calibration, classes A and AA, lines up to 1000 m. A four-wire sensor can be used as three-wire or two-wire simply by leaving the spare conductors disconnected.

This is why a Pt1000 is often the better buy in dirty conditions: at 3.85 Ω/°C the same ohm of lead resistance gives 0.26 °C instead of 2.6 °C. An order of magnitude, over the same two conductors.

Thermocouple: cold junction and compensating cable

A thermocouple measures a temperature difference, not a temperature. Every tolerance value in IEC 60584-1 assumes a 0 °C cold junction, and that zero does not exist in your cabinet. Electronic cold junction compensation is built into modern instruments with a thermocouple input, but it measures terminal-block temperature, not the temperature of your junction, and every degree of compensation error goes straight into the reading.

The second contribution is the cable. You cannot extend a thermocouple with ordinary copper wire: the splice creates a new thermoelectric junction. You need thermocouple or compensating cable per IEC 60584-3, and the documentation warns outright that it adds its own deviation on top of the tolerance class. None of this applies to a Pt100: plain copper, and the only question is how many conductors. If your run is long and noisy, moving to a current loop is easier anyway. We covered shielding and the grounding point in our article on shielding a 4-20 mA sensor.

Pt100 self-heating: run the numbers

Self-heating is real, but within the specified current band it is tiny. For the TR33 WIKA specifies a measuring current of 0.1…1.0 mA for Pt100 and 0.1…0.3 mA for Pt1000; in the transmitter version the current is below 0.3 mA and the datasheet states plainly that self-heating can be ignored. The power is elementary: 1 mA through 100 Ω is 0.1 mW, 0.3 mA through 1000 Ω is 0.09 mW.

The practical conclusion is not the obvious one. The measuring current is set by your electronics, not by the sensor: the controller, the transmitter, the PLC module. If the input drives current outside the specified band, the datasheet class guarantees nothing any more, and in still air or a viscous liquid poor heat transfer piles on top of self-heating. A thermocouple is free of this whole class of problem: it needs no excitation, so it has no self-heating by definition.

Drift and ageing: what a thermocouple pays for its range

Green rot in type K

Between roughly 800 and 1050 °C, with a shortage of oxygen or in the presence of steam, type K loses chromium from its positive NiCr leg. The thermoelectric voltage falls, the reading drifts downward, and the process is irreversible: the thermocouple is destroyed. The name comes from the greenish tint at the fracture of the wire. Thanks to its silicon content, type N forms a protective oxide layer under the same conditions, so for long high-temperature service it outlasts K at identical tolerance.

The K effect between 400 and 600 °C

Less famous, and far nastier. The NiCr leg of a type K has an ordered crystal lattice below about 400 °C and above 600 °C, and a disordered one in between. If the thermocouple cools faster than roughly 100 °C per hour, the disordered state is frozen in, and the characteristic departs from the table: WIKA quotes a deviation of up to ≈5 °C. The effect is reversible: annealing above 700 °C followed by slow cooling removes it. Thin mineral-insulated thermocouples are especially sensitive, where cooling in still air alone already produces deviations around 1 °C.

A class A Pt100 can be spoiled too

The symmetric caveat nobody mentions. If a standard resistance thermometer has been operated above or below the temperature range of its class, class A can no longer be confirmed. Dwell time is irrelevant, and returning to the working range fixes nothing: the tolerance class of that resistor simply stops being defined. Platinum does not rot, but the paperwork behind the class is already gone.

Response time, vibration, mechanics

A thermocouple is faster, and the reason is mechanical: mineral-insulated cable is made from 0.25 mm in diameter, and a resistive element physically does not come that small. But the moment you put the sensor into a massive thermowell, the response times of both groups converge, and WIKA documentation says outright that in such an assembly it can run into double-digit minutes. The thermocouple advantage stays measurable and mostly irrelevant.

• Vibration, the requirement. IEC 60751 expects a resistance thermometer to withstand up to 3 g (30 m/s²) across 10…500 Hz, quoted peak-to-peak.
• Vibration, the reality. Standard WIKA resistance thermometers reach 6 g, the thin-film vibration-resistant build 20 g, and a special construction 50 g. The miniature TR33 is rated 20 g over 10…2000 Hz and 50 g shock per IEC 60068-2-27.
• Flexibility. Mineral-insulated thermocouple cable bends to five sheath diameters, so it reaches places a rigid stem never will.
• The thermowell ceiling. The thermowell limit is often lower than the thermocouple limit. Rate the assembly, not the sensor.

One more type K figure that often gets skipped: the recommended upper limit depends on sheath diameter. At 0.5 and 1.0 mm it is only 700 °C, at 3.0 mm it is 1070 °C, at 6.0 mm it is 1100 °C. A thin thermocouple is fast, but that does not make it a high-temperature one.

What we actually fit

In short, from the process to a catalogue item.

• Process up to 150 °C, signal into a PLC. A TR33 with the integrated 4…20 mA transmitter: 37114012 at L = 100 mm, 37114013 at L = 50 mm on a thin wall, 14081647 at L = 150 mm on a large diameter.
• Process in the 10…110 °C band, local reading without power. A TG54: 83137991 with G½B and an axial connection, 46955940 for the same build radial.
• A thermowell is already in place. A plain-stem TG54: 83138081 for 0…+120 °C or 37114010 for −30…+50 °C.
• Control needed, not just measurement. An E5CC controller: E5CC-RX2ASM-800 for a valve or contactor, E5CC-QX2ASM-800 for an SSR, E5CC-QX3D5M-000 for a 24 V cabinet.

A 4-20 mA loop instead of resistance: Wika TR33

For most industrial processes up to 150 °C we fit Wika resistance thermometers, model TR33, in the integrated-transmitter build. The logic is simple: the output is a current loop rather than a resistance, so the whole question of 2, 3 or 4 conductors and cable length disappears along with the lead resistance itself.

• The element. In the transmitter version the sensing element is a class A Pt1000 per IEC 60751, not a Pt100. If you specifically need a direct Pt100 output in 2, 3 or 4-wire, that is a different TR33 build and has to be ordered separately. Write to us through contacts.
• Supply and load. UB 10…30 V DC, permissible load RA ≤ (UB − 10 V) / 23 mA. At 24 V that is ≈609 Ω, comfortably enough for a controller input plus the run resistance.
• Error as a chain. The datasheet works an example for a 0…150 °C range under stated conditions: 100 °C process temperature, 200 Ω load, 16 V supply, 33 °C ambient. That gives 0.52 K typical and 1.2 K maximum. The class A element itself contributes only ±0.35 K of that, the rest is the transmitter, the load, the supply and the ambient.
• Break diagnostics. Fault signalling per NAMUR NE 043: ≤ 3.6 mA downscale or ≥ 21.0 mA upscale. The controller sees the fault itself instead of "minus two hundred degrees".
• Conditions. IP67/IP69 per IEC/EN 60529, 85 °C ceiling at the connector, wetted parts in 1.4571 stainless, pressure up to 270 bar with the 6 mm tube.
• Paperwork. Which calibration document or factory test report is available for a given build is something to confirm with the manager: the set depends on the model and on the build ordered.

If none of the three builds in the table fits, a TR33 is ordered to the application. To let us pick the build code, send four parameters: working temperature range, insertion length L, process connection (thread or plain stem) and the output you need, 4…20 mA, Pt100 or Pt1000.

For a specific TR33 build, ask for price and availability.

Wiring a current loop into discrete and analogue inputs is covered in our piece on how to wire a sensor to a PLC. Sizing the loop supply and its current headroom has its own article on the 24 V power supply for PLCs and sensors.

When there is no power at all: Wika TG54 bimetal

Where a local indication without electricity is needed, we fit TG54 bimetal thermometers to EN 13190, class 1. This is not a Pt100 competitor on accuracy, it is a different instrument: the operator reads the temperature on the spot, the case is hermetically sealed to IP65, and there is no electronics to wear out because there is no electronics.

Read the third column carefully, because that is where people get caught. On a TG54 the scale range and the measuring range are different things: class 1 per EN 13190 is guaranteed only between the two triangular marks on the dial, not across the whole scale. On a 0…120 °C instrument that means a 10…110 °C band. If you need 5 °C with the stated error limit, pick a different scale. The zero, incidentally, is trimmed with a screwdriver on the back of the case.

We do not carry thermowells in the catalogue: for the plain-stem items we select a thermowell on request, by stem diameter, insertion length and process thread. The thermowell temperature ceiling is often lower than the instrument's own, so rate the assembly rather than the stem.

If you need a different scale, a different stem length or a thermowell for the stem, ask for price and availability.

A controller that does not force the choice: Omron E5CC

Here is the pleasant part: in Omron controllers of the E5CC family the input type is a parameter, not a hardware variant. The same 48 × 48 mm unit takes a Pt100, a thermocouple and a current loop, so the "Pt100 or thermocouple" decision is not cast into metal at purchase time.

• Inputs. Pt100 and JPt100; thermocouples K, J, T, E, L, U, N, R, S, B, W, PL II; linear inputs 4…20 mA, 0…20 mA, 1…5 V, 0…5 V, 0…10 V. The type is set by the in-t parameter in the initial setting level.
• Indication accuracy at 23 °C. For a resistance thermometer, ±0.2 % of the indicated value or ±0.8 °C, whichever is greater, plus one digit. For a thermocouple, ±0.3 % or ±1.0 °C, whichever is greater, plus one digit. The difference does not favour the thermocouple, and it stacks with the difference in sensor tolerance.
• Builds in the catalogue. E5CC-RX2ASM-800, relay output, 100-240 V. E5CC-QX2ASM-800, pulse output for an SSR, 100-240 V. E5CC-QX3D5M-000, pulse output, three auxiliary outputs, 24 V AC/DC supply.

Work the error budget to the end, not just to the sensor. At 300 °C the controller adds ±0.8 °C on the resistive input against ±1.0 °C on the thermocouple input, while the sensor at that temperature gives ±0.75 K (Pt100 class A) against ±1.5 °C (type K class 1). Taken together the resistive channel comes out about 1.6 times tighter: 1.55 K against 2.5 K if you add those contributions directly. It costs nothing extra, just a different input type.

If you need a different output, a different case size or a different supply, write to us: the E5_C family has other builds as well. Ask for price and availability.

Frequently asked questions

Can I fit a thermocouple instead of a Pt100 on the same controller?

On an E5CC, yes: the input type is switched by parameter. Physically it is a different job, though. A thermocouple has two conductors instead of three, it needs thermocouple or compensating cable per IEC 60584-3, and cold junction compensation is in play. Ordinary copper will not do as an extension, it creates a parasitic thermoelectric junction.

What makes a Pt1000 better than a Pt100?

Sensitivity: 3.85 Ω/°C against 0.385 Ω/°C. The same parasitic lead resistance produces ten times less error, which is why two-wire is the normal arrangement for a Pt1000 rather than a compromise. The IEC 60751 tolerance classes are identical for both.

How many degrees does 1 Ω of lead resistance cost on a two-wire Pt100?

About 2.6 °C, positive, since 1 Ω divided by the 0.385 Ω/°C sensitivity. That is more than the entire class A tolerance at 100 °C, which is ±0.35 K. This arithmetic is exactly why WIKA documentation limits two-wire Pt100 connection to cables up to 250 mm.

Why does a type K thermocouple drift after service at 900 °C?

Most likely green rot: between 800 and 1050 °C with a shortage of oxygen, chromium is depleted from the positive NiCr leg, the thermoelectric voltage falls and the reading drifts down. The process is irreversible. For sustained service in that band the usual answer is type N, whose silicon forms a protective oxide layer.

I need a 4-20 mA signal, not a resistance. What should I take?

A TR33 in the integrated-transmitter build: a class A Pt1000 inside, a two-wire 4…20 mA loop out, 10…30 V DC supply. The load works out as RA ≤ (UB − 10 V) / 23 mA, so ≈609 Ω at 24 V. An E5CC accepts that input directly, with no separate barrier or converter.

The short version

Open your process on one number: maximum working temperature. Up to 300 °C any Pt100 build holds class A; up to 450 °C take class A only in the wire-wound version (W), because on the standard thin-film (F) element class A ends at +300 °C. Wire it three-wire, or go straight to a current loop if the run is long. Between 450 and 600 °C put it on a calculator: your Pt100 is class B up there, and a type K class 1 may well be tighter. Above 600 °C IEC 60751 no longer defines tolerance classes for a Pt100.

If your range sits in the resistive zone, look through the temperature sensors in the catalogue and the Wika equipment selection; the general logic of matching a sensor to a task is laid out in sensors for automation. If the calculation pushes you past 600 °C, write to us: we do not sell thermocouples, but we will tell you which type and class to look for and what thermowell to rate for it. A suspect sensor is worth testing before replacement, and the procedure is in our article on checking a sensor with a multimeter.