Illustration of a common configuration for the use of
a thermocouple. The thermocouple junction is placed
where the temperature needs to be measured, and
the other ends of the wires are joined to copper wires
in a reference bath held at a known, stable temperature
(typically that of melting ice). The thermocouple voltage
(typically 40 µV for each degree Celsius difference
between the junction temperature and the reference
temperature) is measured by a high-resolution
voltmeter or analogue to digital converter.

Thermocouples are the most common sensors in industrial use due to the fact that they are small, simple, rugged and of low cost. They consist of two dissimilar conductors (wires) joined at a junction and contained in an insulator.

If the voltage across the unconnected ends of these wires is measured then it is found to vary according to the temperature at the junction between the two wires. This is the simplest form of thermocouple circuit. In practice, several techniques and tricks are nearly universally used:

• The wires are insulated. This prevents the wires touching except at the tip. The ability of the insulation to withstand temperature is often the limiting factor in working out the highest temperature at which a thermocouple can be used.
  
• The wires are usually welded together at the tip. Any kind of contact at the tip will create a thermocouple. Soldering is possible, as is mechanical crimping, but welding produces the smallest and most robust tips.
  
• Standard combinations of wires are used. Many different materials can be used in the manufacture of thermocouples. These include platinum and platinum-rhodium alloys for use up to 1600 °C; cheaper, mainly nickel-based alloys for industrial use up to 1200 °C; and refractory alloys up to and beyond 2000 °C. Standardised letter-designated thermocouple types (eg Types K, N, T, B, R, S) are manufactured to meet the specifications of IEC 584 (BS EN 60584). This standard gives tables and equations for the EMF (or voltage) generated at every degree in the range, and the tolerances with which they should be manufactured. Thus, for example, a K-type thermocouple has wires made from carefully selected alloys of Nickel-Chromium and Nickel-Aluminium, for which the EMF at 1000 °C should be 41.276 mV, within about 0.16 mV. Summaries of the tables can be found in many reference books, such as (Kaye and Laby, Tables of physical and chemical constants).
  
• · Cold junction compensation. Because the voltage is generated along the wires where there are temperature gradients (see below), it follows that the temperature of the 'free ends' of the wires is important and must be controlled. The 'reference' or 'cold' junctions are commonly held at 0 °C, using melting ice for best accuracy, but in most digital instruments the cold junction effect is compensated by sensing the reference temperature and adding the appropriate voltage to the signal, thereby allowing digital multi-meters to display the temperature directly.

The voltage generated by thermocouples is not usually very large, typically between 10 µV and 40 µV per degree Celsius.

How thermocouples work:
Although the voltage measured by a thermocouple reflects the temperature at the junction between the two dissimilar metals, surprisingly, none of the thermovoltage is actually generated there. Instead it arises all along the wires and depends on the product of the so-called Seebeck coefficient S(x) for a material and the temperature gradient dT/dx at each point on the wire.

In the figure above, points A and C are at the same reference temperature T_R and point B is the measurement junction with temperature T_J. A is connected to B by wire with Seebeck coefficient S₁ and C is connected to by wire with Seebeck coefficient S₂. The voltage generated in the circuit, V_AC , is built up along the two wires, and mathematically this can be written as an integration between points A and C:

If the Seebeck coefficients are constant along each type of wire, then it can be shown that the measured thermovoltage is proportional to the temperature difference between T_R and T_J and given by:

This mathematics is placed here to emphasise the importance of temperature gradients: the function of the junction is connection for detection.

The role of calibration
Buying standard thermocouple wire from a reputable company should allow you to construct thermocouples which are accurate within the specified tolerances. If measured correctly, the tip temperature inferred is unlikely to be in error by more than 5 ºC over the range from -100 ºC to over 1000 ºC. However, if you want to be sure that this is so you must make or buy a standard thermocouple, and have it calibrated at a laboratory whose thermometers can be traced back to the ITS-90. NPL can supply and calibrate various types of thermocouple.