A thermocouple has no battery, no excitation supply, and no electronics at the sensing tip, yet it produces a voltage that tracks temperature. That voltage comes from the Seebeck effect, the thermoelectric phenomenon that turns a temperature difference directly into an electrical signal. This page is about the effect itself rather than the thermocouple as a device: why joining two dissimilar metals and heating one end generates a millivolt EMF, why that makes the sensor self-powered, and why the effect - not a quirk of manufacturing - is the reason cold-junction compensation is needed and why type J, K, and other curves are nonlinear. It is written for automation techs reading those millivolts into a SCADA input.
Seebeck effect in one line: The Seebeck effect is the generation of a voltage in a conductor when its two ends are at different temperatures: charge carriers diffuse from the hot end toward the cold end, creating a small electromotive force proportional to the temperature difference. In a thermocouple, two dissimilar metals with different Seebeck coefficients are joined, and the net millivolt EMF around the loop depends on the temperature difference between the measuring junction and the reference junction.
Take a single length of metal wire and heat one end while keeping the other cool. Inside the metal, the mobile charge carriers - electrons in most metals - have more thermal energy at the hot end than at the cold end. That energy difference makes them diffuse, on average, from the hot end toward the cold end, piling up slightly and leaving the hot end marginally positive. The result is a small voltage between the two ends of the wire, set up purely by the temperature gradient along it. This is the Seebeck effect in its simplest form: a temperature difference in a conductor produces an electromotive force.
How strongly a given metal does this is captured by its Seebeck coefficient, a property expressed in microvolts per degree that says how many microvolts of EMF appear per degree of temperature difference. Different metals and alloys have different Seebeck coefficients, and some are positive while others are negative depending on how their carriers behave. The coefficient is not perfectly constant - it changes with temperature - which is a fact worth remembering because it is the root of the nonlinearity that shows up later in thermocouple tables.
The effect is entirely passive. Nothing pumps the charge; the temperature gradient itself drives the diffusion, and the voltage exists as long as the gradient exists. That is why a thermocouple needs no power supply at its sensing point. The energy that produces the signal comes from the heat flowing through the wire, not from any external source, which makes the Seebeck effect a direct thermal-to-electrical conversion.
A single wire develops a Seebeck voltage along its length, but you cannot measure it usefully with that same wire, because connecting a voltmeter with leads of the same metal would set up equal and opposite Seebeck voltages that cancel. To capture a net voltage you have to build a loop out of two different metals with different Seebeck coefficients. That is exactly what a thermocouple is: two dissimilar conductors joined at one point, the measuring junction, with their other ends forming a second junction where the reading is taken.
When the two junctions are at different temperatures, each metal develops its own Seebeck voltage along its length, but because the two metals have different coefficients, those voltages do not cancel. The difference between them is the net EMF the loop produces, and it depends on the difference in temperature between the measuring junction and the reference junction. This is the essential point that trips people up: a thermocouple does not measure absolute temperature at its tip. It measures the temperature difference between its two junctions, and turning that into an absolute reading requires knowing the temperature of the reference junction.
That requirement is precisely why cold-junction compensation exists. The reference junction - where the thermocouple wires meet the copper of the instrument - is the cold junction, and its temperature is usually whatever the terminal block happens to be, not a fixed zero. To recover the true temperature at the measuring junction, the system has to measure the cold junction's temperature independently and add back the EMF that the temperature difference to that reference represents. Cold-junction compensation is not a workaround for a defect; it is a direct consequence of the Seebeck effect producing a voltage from a temperature difference rather than an absolute temperature.
For a technician wiring a thermocouple into a SCADA input module, the Seebeck effect explains the whole signal chain. The sensor delivers a tiny voltage - typically tens of microvolts per degree, so a full-scale reading may be only a few dozen millivolts - which is why thermocouple inputs need high-gain, low-noise conditioning and why lead resistance matters far less than for a resistance sensor. It also explains why the input module must have a cold-junction sensor at its terminals: without knowing the reference temperature, the millivolts alone cannot be converted to a temperature, because they encode a difference.
The nonlinearity built into every type J, K, T, or E table is another direct inheritance from the effect. Because each metal's Seebeck coefficient changes with temperature, the millivolts-per-degree relationship is not a straight line, and the same millivolt change means slightly different temperature changes at the cold and hot ends of the range. That is why a SCADA system does not simply scale the millivolts linearly; it applies the standardized polynomial or lookup table for the specific thermocouple type to convert the measured EMF, after cold-junction correction, into engineering units.
A cloud SCADA platform such as Merobix carries this conversion so the operator sees temperature, not raw millivolts, while the underlying signal chain does exactly what the Seebeck effect dictates: measure the loop EMF, measure the cold-junction temperature, add back the reference contribution, and map the total through the correct nonlinear curve. Understanding that the whole procedure follows from one thermoelectric effect makes troubleshooting far more direct - a reading that drifts with ambient temperature points to a cold-junction problem, and a reading that is off across the range points to a wrong thermocouple type in the conversion, each traceable to the physics rather than guesswork.
Because the Seebeck effect converts heat directly into a voltage. A temperature gradient along the thermocouple wires drives charge carriers to diffuse from hot to cold, producing an electromotive force with no external excitation. The energy comes from the heat flowing through the metal, not from a battery or supply, which is why a thermocouple is self-powered and delivers a signal the instant its junctions are at different temperatures.
Because the Seebeck effect produces a voltage from the difference in temperature between the two junctions, not from the absolute temperature at the tip. To convert the measured EMF into an actual temperature, the system must know the temperature of the reference, or cold, junction - usually the instrument's terminal block - and add back the EMF that difference represents. Cold-junction compensation is therefore a direct consequence of how the effect works, not a defect being corrected.
Because each metal's Seebeck coefficient - its microvolts of EMF per degree - changes with temperature rather than staying constant. That means the millivolts a thermocouple produces do not rise in a perfectly straight line with temperature, so the same millivolt change represents a slightly different temperature change across the range. SCADA systems handle this by converting the measured EMF through the standardized polynomial or lookup table for the specific thermocouple type.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.