Measuring temperature and pressure — thermowells, thermocouples and gauges

Temperature and pressure are the two measurements a plant takes most often, and they are the two where the sensor is least often the problem. What goes wrong is almost always what sits between the process and the sensor — a well, a leg, a length of the wrong wire.

Thermocouples: a junction of two dissimilar metals

Join two different metals and the junction produces a small voltage that varies with temperature. That is the entire principle, and its simplicity is the source of both its usefulness and its traps.

Sensor	How it works	Usable range	Accuracy and stability	What catches people out
RTD (Pt100)	Resistance of platinum rises predictably with temperature	-200 to 600 °C	High, and stable over years	Lead resistance adds straight into the reading — use three or four wires, never two
Thermocouple type K	Voltage generated across a junction of two dissimilar metals	-200 to 1260 °C	Moderate; drifts with age and cycling	Needs matching extension wire and cold junction compensation
Thermocouple type J	Iron and constantan junction	-40 to 750 °C	Moderate	The iron leg rusts. Poor choice in a damp or outdoor location
Thermocouple type T	Copper and constantan junction	-200 to 350 °C	Good at low temperatures	Limited top end; often chosen for cryogenic and refrigeration duty
Thermocouple type N	Nicrosil and nisil junction	-200 to 1300 °C	Better long-term stability than K	Chosen where a K would drift too fast; less commonly stocked
Bimetallic dial gauge	Two bonded metals expand differently and curl	-50 to 500 °C	Low, but entirely self-contained	No power, no signal, no wiring to go wrong — which is why local gauges survive

Almost every process measurement between -50 and 500 °C is better served by an RTD, and almost everything above it has to be a thermocouple. The choice is usually made by the range first and everything else second.

The first trap follows directly from the principle.

Cold junction compensation

A thermocouple never tells you a temperature. It tells you the difference between its two ends.

Historically the reference end was held in an ice bath at a known 0 °C — hence “cold junction”. Modern instruments measure the temperature of their own terminals with a separate sensor and add it back arithmetically.

Get this wrong, or lose it, and the reading is out by roughly the ambient temperature — and it drifts as the room warms. A thermocouple reading that is wrong by about twenty degrees and moves with the weather is a compensation problem, not a sensor problem.

RTDs: resistance, and the wires that corrupt it

A platinum resistance thermometer changes resistance predictably with temperature. A Pt100 is 100 ohms at 0 °C, roughly 138 ohms at 100 °C.

It is more accurate and far more stable than a thermocouple over normal process ranges, and it is the right default below about 500 °C. Its trap is the opposite of the thermocouple’s.

Thermowells

Almost no sensor touches the process directly. It sits inside a thermowell — a closed tube that penetrates the pipe or vessel wall.

The benefit is maintainability: the sensor can be pulled out and replaced with the plant running, because the well keeps containment.

The costs are worth knowing:

• Response time. Heat must cross the well wall and the gap to reach the sensor, so the reading always lags. A spring-loaded element held hard against the bottom of the well makes that lag as small as it can be, and an excessive air gap makes it worse — a sensor that reads slow and low.
• Insertion depth. The tip has to reach flowing fluid. A well too short reads the pipe wall rather than the process, which on an insulated line means it reads something closer to ambient.
• Mechanical survival. This one is a safety matter.

Pressure: protecting the element

A pressure gauge or transmitter has a thin element that flexes. Most pressure instrument failures are that element being given something it was not designed to see.

Steam — fit a siphon, the pigtail loop below the gauge. Condensate collects in the loop and stands between live steam and the element. It must be filled to work, and a gauge commissioned on an empty siphon is exposed until it fills itself.

Pulsation — fit a snubber. On a reciprocating pump discharge the element sees several hundred cycles a minute, and fatigue failure of a bourdon tube is a release on a pressure boundary. This is a containment item, not a gauge-life item.

Corrosive or solidifying fluid — fit a diaphragm seal, which keeps the process out of the instrument entirely and transmits pressure through a fill fluid. It solves one problem and introduces two smaller ones: the fill fluid brings its own temperature effects, and the capillary adds lag.

Choose the range before you choose the gauge

What to take away

• A thermocouple is a junction of dissimilar metals, so terminating onto copper creates a second one. Use matching extension wire.
• A thermocouple measures a difference, which is why cold junction compensation exists. Lose it and you are out by roughly the ambient.
• An RTD must never be two-wire in the field — lead resistance adds straight into the reading. Use three or four wires.
• A thermowell trades response time for maintainability, needs enough insertion depth, and is subject to a wake frequency calculation because it can vibrate off.
• Siphon for steam, snubber for pulsation, diaphragm seal for corrosive fluid. Each protects the element from a different thing.
• Gauge accuracy is a percentage of full scale. Put the working pressure near mid-scale or the error swamps the reading.