a thermocouple
/ THUR-mo-kup-ul /
A thermocouple is the strangest temperature sensor of the bunch because it does not change resistance at all — it generates its own tiny voltage out of a temperature difference, with no power supply needed. Join two different metals at one end, heat that junction, and a small voltage appears across the free ends. The hotter the junction relative to the other end, the bigger the voltage. It is a temperature sensor that, in effect, runs on heat alone.
This is the Seebeck effect, and the voltages are small — typically tens of microvolts per degree. A type-K thermocouple (chromel-alumel) gives about 41 microvolts per °C, so 1000 °C produces only about 41 mV. Here is the crucial twist beginners must understand: a thermocouple measures the difference between its hot junction and its cold (reference) junction where it connects to your circuit. To get an absolute temperature you must know the cold-junction temperature and add it back in — a step called cold-junction compensation, usually done by measuring the terminal block with a little thermistor or RTD and correcting in software. Dedicated chips do this whole job for you.
Why this matters: thermocouples win where others cannot go. They are rugged, cheap, and span enormous ranges — type K reads from below -200 °C to above 1200 °C — so they live in furnaces, kilns, engines, and ovens. The honest caveats: the output is tiny and needs careful low-offset amplification; it is fairly nonlinear so it needs a lookup table or polynomial; and if you forget cold-junction compensation your reading is wrong by whatever the room temperature happens to be. They are about wide range and toughness, not the last fraction of a degree of accuracy.
A type-K thermocouple in a kiln sits with its hot junction at 800 °C and its terminals at 25 °C. The wire produces about 41 microvolts per degree of difference, so it outputs roughly (800 - 25) times 41 uV = about 31.8 mV. Software adds back the 25 °C measured at the terminals to report the true 800 °C.
A self-powered voltage from a temperature difference — but you must know the cold end.
Forgetting cold-junction compensation is the number-one thermocouple mistake: the sensor reports a difference, not an absolute temperature, so your reading is off by the terminal-block temperature unless you measure and add it back.