Temperature Conversion and the Delta-T Rule

Why this matters

Everyone can convert a temperature. Fewer people notice that a temperature and a temperature difference convert by different rules, and the equipment does not care which one you meant. A manual written in Celsius that asks for a 10 degree rise across a heat exchanger is asking for 18 F, not 50 F. A sensor rated plus or minus 0.5 C is good to plus or minus 0.9 F, not 32.9 F. Put the wrong one into a heat formula and the answer is out by a factor of nearly three, with every digit looking plausible. This card is the two conversions, the rule that separates them, and the checks that make the difference visible before it reaches a setting or a report.

The two formulas for a reading

A reading is a point on a scale: the water is at 60 C, the room is at 68 F. The two scales have different zeros and different degree sizes, so converting a reading needs both a multiply and an offset.

• Celsius to Fahrenheit: F = 1.8 x C + 32
• Fahrenheit to Celsius: C = (F - 32) / 1.8

The 1.8 is the size ratio: water's freezing-to-boiling span is 100 Celsius degrees and 180 Fahrenheit degrees, and 180 / 100 = 1.8. The 32 is where the zeros sit: 0 C is 32 F. Order matters in the reverse formula. Subtract the 32 first, then divide. Divide first and 120 F becomes 120 / 1.8 - 32 = 34.7 C instead of the correct 48.9 C.

The delta-T rule

A difference between two readings has no zero to shift. The offset of 32 appears in both readings and cancels when you subtract one from the other, so only the size ratio is left:

• A Celsius difference to Fahrenheit: dF = 1.8 x dC
• A Fahrenheit difference to Celsius: dC = dF / 1.8

No 32 in either direction. A 10 C rise is an 18 F rise. A 20 F drop is an 11.1 C drop. Adding 32 to a difference is the single most common temperature error in the field, and it never looks wrong, because 50 is a perfectly ordinary number.

The same rule covers anything that is a difference in disguise: a tolerance, a differential, a hysteresis band, a superheat or subcooling target, a rate of change per hour. Each is a span, not a point, so each converts by 1.8 alone.

Reading and difference side by side

A HYDRONIC LOOP READ IN CELSIUS

                 C        F, as a READING
supply          12 C   1.8 x 12 + 32 = 53.6 F
return          24 C   1.8 x 24 + 32 = 75.2 F

difference      12 C   as a DIFFERENCE: 1.8 x 12 = 21.6 F
check                  75.2 - 53.6         = 21.6 F

WRONG dT        12 C   1.8 x 12 + 32 = 53.6 F

THE OTHER WAY
reading        120 F   (120 - 32) / 1.8 = 48.9 C
difference      20 F    20 / 1.8        = 11.1 C
WRONG dT        20 F   (20 - 32) / 1.8 = -6.7 C

The check line is the whole rule in one subtraction. Convert both readings first, subtract, and you get the same answer as converting the difference by 1.8. Do it once with real numbers and the missing 32 stops feeling arbitrary.

The wrong lines are worth staring at. A 12 C rise written as 53.6 F is almost two and a half times the real 21.6 F. A 20 F difference run through the reading formula comes out negative, which is the one case where the slip announces itself; everywhere else it hides.

Worked example: a metric manual on a US job

A tech commissions a water heater whose manual is written in Celsius. It gives a setpoint of 55 C, a switching differential of 5 K, and a sensor accuracy of plus or minus 0.5 K. (A kelvin is the same size as a Celsius degree, and manuals use K for differences precisely to flag that they are spans, not readings.)

Convert each by the rule that fits what it is:

• Setpoint, a reading: 1.8 x 55 + 32 = 99 + 32 = 131 F. Check backward: (131 - 32) / 1.8 = 99 / 1.8 = 55 C.
• Differential, a difference: 1.8 x 5 = 9 F. The heater lets the tank fall about 9 F below the setpoint before it fires again, so it cycles somewhere around 122 to 131 F, subject to where exactly the manufacturer places the band.
• Sensor accuracy, a difference: 1.8 x 0.5 = 0.9 F. A reading of 131 F could be a real 130.1 to 131.9 F.

Now the slip in action. A tech who converts the differential like a reading gets 1.8 x 5 + 32 = 41 F and expects the tank to drop to 90 F before reheating. When the customer reports the heater "keeps coming on," the tech goes hunting for a fault that is just the manual working as written. One wrong formula on one line, and an hour of diagnosis follows from it.

Worked example: a rise in a heat formula

A Celsius-spec boiler states a design temperature drop of 11 C across the loop at full output of 88,000 BTU/h. How much flow does that need?

• Convert the drop as a difference: 1.8 x 11 = 19.8 F.
• Use the water formula: GPM = 88,000 / (500 x 19.8) = 88,000 / 9,900 = 8.9 GPM.
• Check forward: 500 x 8.9 x 19.8 = 88,110 BTU/h, which closes to rounding.

With the 32 wrongly added, the drop becomes 51.8 F and the flow 88,000 / (500 x 51.8) = 3.4 GPM. That is about 38 percent of the real requirement (3.4 / 8.9 = 0.38). A pump selected on 3.4 GPM would leave the far end of the loop cold, and every calculation downstream of the wrong drop would agree with it. The formula itself is covered in The Sensible Heat Formulas for Air and Water.

Absolute scales, and when they matter

Kelvin and Rankine start at absolute zero, the lowest temperature possible.

• K = C + 273.15
• R = F + 459.67

A kelvin is the same size as a Celsius degree and a Rankine degree the same size as a Fahrenheit degree, so differences on the absolute scales are the same numbers: a 10 C rise is a 10 K rise, and an 18 F rise is an 18 R rise.

The absolute scales earn their place wherever a formula multiplies or divides by a temperature rather than subtracting one: gas laws, density corrections for air, radiant heat. There, a temperature has to be absolute or the ratio means nothing. "80 F is twice as hot as 40 F" is false; in Rankine those are 539.67 and 499.67, and the warmer air is about 8 percent hotter in the sense that matters for its volume (539.67 / 499.67 = 1.08). The same goes for percentages: a temperature that rose from 40 F to 80 F did not go up 100 percent in any physical sense, which is why the percent-change rules in Percent Change and Percentage Points Done Right are never applied to an ordinary Fahrenheit or Celsius reading.

Probe offsets, and why a pair of probes is a difference problem

A calibration offset is a difference too, and it behaves like one. Check two probes in a properly made ice bath (crushed ice packed in water and stirred, which sits at 0 C, 32 F). Probe A reads 0.6 C and probe B reads -0.3 C. Their offsets are +0.6 C and -0.3 C, and in Fahrenheit, by the difference rule, +1.1 F and -0.5 F. Not 33.1 F and 31.5 F; those are the readings the probes showed, and the offset is how far each reading sits from 32 F.

Now use the pair to read a temperature difference across a coil or a heat exchanger. If both probes carried the same offset, it would cancel in the subtraction and the delta-T would be exact even though both readings were wrong. These two do not match. Put A on the supply and B on the return, and the measured difference is off by the gap between their offsets: 0.6 - (-0.3) = 0.9 C, which is 1.6 F. On a 20 F design difference that is 1.6 / 20 = 8 percent of the quantity you are trying to read, enough to make a healthy system look marginal.

Two fixes, both cheap. Correct each reading by its own offset before subtracting, writing the raw reading, the offset and the corrected value as three separate lines so nobody applies the correction twice. Or swap the probes between supply and return, take the difference again, and average the two differences: the offset gap appears once with each sign and drops out. Either way the offset gets labeled "difference" and converted by 1.8 alone, like every other span on this card.

Estimating in your head

Two quick checks keep a converted number honest without a calculator:

• Double and add 30. 20 C becomes about 70 F (the exact answer is 68). 35 C becomes about 100 F (exact: 95). It runs a few degrees high in warm ranges, which is fine for catching a slip and not fine for a setpoint.
• For a difference, nearly double. A 10 C span is about 20 F (exact: 18). If a converted difference is not somewhere near double the Celsius figure, the 32 crept in.

And three fixed points are worth memorizing because they verify any formula you write: 0 C is 32 F, 100 C is 212 F, and the scales cross at -40, where -40 C and -40 F are the same temperature. Plug -40 into a formula you have just written; if it does not return -40, the formula is wrong.

Where the slip shows up on a ticket

• Metric equipment data. Superheat and subcooling targets in K, a heat pump's defrost differential, a chiller's design range, a heat exchanger's approach temperature. Every one is a difference, so every one converts by 1.8 alone.
• Instrument specs. A thermometer's accuracy or resolution in Celsius is a span, so plus or minus 1 C is plus or minus 1.8 F.
• Rates. A slab, a tank or a cold room losing 2 C per hour is losing 3.6 F per hour.
• Tolerances on a setpoint. "60 C plus or minus 2 C" is 140 F plus or minus 3.6 F: convert the setpoint as a reading and the tolerance as a difference, separately.

The habit that prevents all of it is to label each number before converting it: write "reading" or "difference" next to it on the page. The label decides the formula. A number you cannot label is a number you do not yet understand, and that is the moment to look back at where it came from.

References

• Equipment installation manuals for setpoints, differentials and sensor accuracy as stated by the manufacturer
• See related: The Sensible Heat Formulas for Air and Water; What a Delta-T Is Actually Telling You; The Delta-T That Somebody Added Thirty-Two To; Common Unit Conversions + Math Formulas