Convert K to °C

Kelvin starts at absolute zero, the point where there is no thermal energy left to remove, and its degrees are the same size as Celsius ones. It takes no degree sign — it is 300 K, not 300 °K.

Celsius puts zero at the freezing point of water and 100 at its boiling point, at sea level. Anders Celsius originally had the scale the other way up, with zero as boiling.

Kelvin and Celsius do not start counting from the same place, so no single number converts between them. Multiplying is the mistake this page exists to prevent: it is right at exactly one temperature and wrong everywhere else, and the error grows the further you get from that point.

The two ends of this page's own scale show it. 0 K is -273.15 °C and 100 K is -173.15 °C — a hundred steps on one side, 100 on the other, and the two zeros in different places. That is why the formula above has two parts, something to multiply by and something to add; drop the addition and the answer stops being wrong slowly and starts being wrong badly.

Kelvin starts at absolute zero, the point where there is no thermal energy left to remove, and nothing below it exists to measure. That is what makes it a scale you can multiply on: 200 K really is twice 100 K, in a way that 200 °C is not twice 100 °C.

Neither Celsius nor Fahrenheit can say that. Their zeros were chosen, not found, which is why doubling a Celsius reading means nothing at all.

Take 273.15 off and the conversion is finished. A kelvin and a degree Celsius are the same size, so nothing is scaled: 300 K is 26.85 °C, 273.15 K is 0 °C, and 0 K is −273.15 °C. There is no multiplication anywhere and no order of operations to get wrong, which makes this the least error-prone conversion in the category.

The check worth running is against 300. Almost every kelvin figure describing something a person could stand beside sits between 250 and 320 K, so an answer far from that band should read as far from ordinary — 200 K is −73 °C and belongs to a laboratory freezer, 400 K is 127 °C and is past the boiling point of water. A kelvin figure in the low thousands is almost certainly a colour temperature rather than heat, which is a different problem entirely.

A lamp marked 2700 K is not at 2,427 °C, and converting it says nothing useful about the lamp. Colour temperature names the temperature to which an idealised black body would have to be heated for its light to match the colour of this one. For an old incandescent the coincidence is nearly literal, because the filament really did run near 2,700 K, but an LED marked 2700 K sits at perhaps 60 °C and is matching a colour rather than a temperature.

The shelf is a short list: 2700 K is the warm yellow of an incandescent, 3000 K a whiter warm, 4000 K a neutral office white, 5000 to 5600 K daylight, and 6500 K the bluish standard used for screens and colour matching. Converting any of them produces a Celsius figure between about 2,400 and 6,200 that is correct about a hypothetical radiator and irrelevant to the object in your hand.

The vocabulary runs backwards from the physics. A 2700 K lamp is sold as warm and a 6500 K one as cool, while 6500 K is by far the hotter black body. The words describe how the light feels — yellow and orange read as warm, blue reads as cold — and they settled long before the numbers were printed on packaging, so the two systems now point in opposite directions on the same box.

This is the source of most disappointing lighting purchases. Somebody wanting a cosier room reaches for the larger number because it sounds like more warmth, and buys 5000 K, which is the light of an overcast noon. The rule that survives contact with a shop is that the number describes the colour and runs orange to blue as it rises, while the word beside it describes the feeling and runs the other way.

Above about a thousand kelvin the 273.15 is noise. The Sun’s photosphere is roughly 5,772 K, which is 5,499 °C, and quoting it as 5,500 °C loses nothing, because the kelvin figure carried more uncertainty than the offset does. The same holds for a tungsten filament near 3,000 K, a candle flame near 1,700 K or a stellar interior in the millions.

Astronomy uses kelvin throughout partly for that reason and partly because most of the interesting range is nowhere near anything a Celsius zero is relevant to. Venus’s surface is 737 K, or 464 °C. The cosmic microwave background is 2.725 K, or −270.4 °C — and there the offset is essentially the whole answer, the Celsius figure being made almost entirely of the constant, which is a neat illustration of why the field does not bother converting.

Below about 100 K the Celsius figure becomes the unhelpful one. Liquid nitrogen boils at 77 K, which is −196 °C; liquid helium at 4.2 K, or −269 °C. Working in Celsius here means carrying a large negative number whose informative digits sit at the far end, and halving the temperature turns −196 into −235, a change that reads as small and is not.

In kelvin the same step is 77 to 38.5, which is visibly a halving, and that is the practical reason cryogenics is written in kelvin rather than a matter of taste. Superconducting transition temperatures follow the same convention: the high-temperature superconductors that caused so much excitement transition somewhere around 90 to 135 K, which is −183 to −138 °C and does not sound high at all until you know the earlier ones needed 4 K.

A datasheet quoting a coefficient in ppm/K, a thermal resistance in K/W or a drift in µV/K is describing a change, and a change is the same number in Celsius. Nothing is subtracted: 50 ppm/K is 50 ppm per degree Celsius. Manufacturers write K because the kelvin is the SI unit of temperature, not because the figure belongs to an absolute scale.

The slash is the tell. A reading is written 300 K; a rate is written per K, and only the reading takes the offset. Some datasheets mix both on one page — an operating range in °C, a coefficient in ppm/K, a junction-to-ambient figure in °C/W — and the mixture is not an inconsistency. The genuine exception is a thermistor’s B value, quoted in kelvin because it sits inside an exponential of 1/T and needs the absolute temperature.

Kelvin figures arrive with wildly different precision, and the conversion cannot add any. 2.725 K is a measurement with four significant figures and converts to −270.425 °C with all of them intact. 5,772 K is a nominal value, and 5,498.85 °C overstates it. The 2700 K on a bulb is a marketing bin, and lamps sold under it vary by a hundred kelvin or more between manufacturers.

The habit that keeps a report honest is to keep the digits the kelvin figure had and no more, then decide whether the .15 survives. Below about 10 K it always does, because it dominates the answer. Between 250 and 400 K it usually does, because a tenth of a degree is a real distinction there. Above 1,000 K it never does, and carrying it is a sign the number came out of a converter rather than a decision.

A camera’s white balance in kelvin is set to the colour of the light falling on the scene, not to the colour you want out of it. Under tungsten at about 3,200 K, setting the camera to 3,200 K yields neutral whites; setting it to 5,600 K tells the camera the light is bluer than it really is, and it compensates by warming everything toward orange.

The direction confuses people because the dial behaves like a look: raising the number warms the image and lowering it cools it, which is the opposite of what the physics of the number suggests. It is the same inversion as the lighting aisle, seen from the other side. Converting any of these settings to Celsius has no use whatever — 3,200 K would be 2,927 °C, a claim about a hypothetical filament, printed on a page about a camera menu.