# Thermocouples: Simple but misunderstood

-October 28, 2013

The lowly thermocouple is probably the most widely used type of sensor, yet so many people don't understand how it works.

That's because many descriptions out there get it wrong and that misinformation keeps circulating. I once took that misinformation as correct, but no more. Here's the misleading statement, paraphrased:

A thermocouple is made of two dissimilar metals joined to form a junction. A voltage occurs across that junction that changes with temperature.

There is no voltage generated at the junction where the two metals meet. What actually happens is a result of the Seebeck Effect, as described in Encyclopedia Brittanica:

Seebeck effect, production of an electromotive force (emf) and consequently an electric current in a loop of material consisting of at least two dissimilar conductors when two junctions are maintained at different temperatures. The conductors are commonly metals, though they need not even be solids. The German physicist Thomas Johann Seebeck discovered (1821) the effect. The Seebeck effect is used to measure temperature with great sensitivity and accuracy (see thermocouple) and to generate electric power for special applications.

If you have a wire of a given metal composition and there's a temperature gradient along its length, a voltage will occur along the wire that's characteristic of the metal. A wire made of a different metal will, under the same conditions, produce a different voltage across it. Thus, the voltage is across the wires, not across the junction of the two metals. Figure 1a is wrong.

Figure 1. The voltage difference between two wires is what makes a thermocouple work.

Figure 1b shows what's really going on. In this case, assume that the ends of the wires at the thermal block are both at 0°C and the junction of the wires is at temperature T1. Therefore, the temperature gradient across the wires is the same. Because the wires are made of different materials, each will develop a different voltage across it, shown as V1 and V2. The difference between them is the voltage that's proportional to the temperature. That temperature/voltage relationship VT=V1-V2 is, of course, nonlinear. Fortunately, it's predictable and well documented for numerous thermocouples.

For simplicity, I assumed that T0 was 0°C. But, you can't keep your wires in an ice bath just to get a known temperature.

Thermal blocks used in digital thermometers have another temperature sensor, preferably embedded in the thermal block. Often, that's an integrated IC sensor whose output is linearly proportional to temperature or it may even be a smart sensor with a digital output. With one known temperature, a digital thermometer can calculate the unknown voltage through software based on VT and the type of thermocouple being used.

There are other issues, too. At some point, the connections to the circuits will be copper PCB traces. Thus, you will have essentially more thermocouples in the system that form where the thermocouple wires connect to the system (Figure 2). Software can compensate for that, too.

Figure 2. The connections of the thermocouple wires to a junction block produce more thermocouples.

If you're designing an embedded system that will use a thermocouple, you'll also need some signal-conditioning circuits before sampling the thermocouple voltage. For one thing, the output of the thermocouple is in microvolts/°C. It's also best to use a differential input to avoid ground loops that can lead to errors.