Aircon Tonnage: An Exact Unit, and Who Actually Uses It

A ton of cooling measures a rate, not a mass. It says how fast a machine pulls heat out, and nothing in it describes how heavy that machine is. A three ton system does not weigh three tons.

Delivered ice is where the unit began. Before mechanical cooling, buildings were chilled by ice brought in by cart, and capacity was quoted as tons of ice per day.

The arithmetic behind that is short. A short ton is 2,000 pounds, which NIST publishes as 907.1847 kg. Melting ice at its normal melting point takes about 333.43 kJ/kg, a value IAPWS now carries in its reference equation of state. Spread over a full day, the answer is close to 3.50 kW.

A whole day sits hidden inside the unit. BTU/h wears its time base in the name, and a watt is a rate by construction. The ton buries one day of melting in its definition and then divides that day back out, which is how a figure built to describe a day's work reads as an instant rate.

What got published is rounder than the ice that produced it. A ton of refrigeration is set at a flat 12,000 BTU/h, where the ice gives about 11,950. The definition therefore sits some 0.45% above its own origin. ASHRAE puts this plainly in the terminology entry, calling the ice figure an approximation of the unit and not its source.

That small gap is the useful part. A measured unit would carry an uncertainty with it. A defined unit carries none, because the number is the agreement. The ton stopped describing melting ice the moment 12,000 was written down, and everything downstream of it now converts cleanly.

A ton of refrigeration is 12,000 BTU/h because that is where the unit was fixed. Nothing was measured to arrive at that number.

The British thermal unit underneath it is defined too. NIST puts the international table figure at 1,055.06 J, so the chain from tons through to watts closes with no slack in it. Run that chain and one ton lands on 3,516.85 W, which is the figure NIST publishes.

Going the other way uses the same constant. One kilowatt is 3,412.14 BTU/h, so a spec sheet quoting both figures is quoting one measurement twice. They are the same quantity written in two alphabets.

Drift starts at the rounding. The trade writes 3.5 kW where the ton would give 3.5169, which makes a unit sold as 3.5 kW about 0.995 of a ton. Half a per cent vanishes in a bedroom but not across a plant of forty machines.

A capacity stated in kilowatts, rounded into tons for the client, then back into kilowatts for the supplier, leaves the round trip a different number from the one it entered with. Convert once, work from the source figure, and keep it visible next to the answer.

Household paperwork here does not carry the ton at all. NEA states cooling capacity in kilowatts, on the label and in the register behind it. The 2023 circular that pulled portable units into the scheme sets its thresholds the same way, with neither tons nor horsepower appearing.

The international test standard for this class of machine agrees. ISO 5151 covers non-ducted air conditioners, and its definition of total cooling capacity closes with a note saying the figure is expressed in watts.

One tier up, in commercial plant, the ton is written into the rules. BCA's Green Mark standard for existing non-residential buildings sets its chiller prerequisites in kilowatts per refrigeration ton, and splits the thresholds on a building cooling load stated in RT. A water-cooled plant under 500 RT has to reach 0.85 kW/RT. At 500 RT and above it has to reach 0.75.

Air-cooled and unitary systems get their own line at 1.1 and 1.0 kW/RT. Even the metering is specified: plant instrumentation has to compute kW/RT to within 5% of true value, following ASHRAE Guide 22 and AHRI 550/590.

The Green Mark technical guide in force from mid 2024 still runs on kW/RT and points at SS 591 for the long-term measurement.

Ownership decides who ever sees the number. In a mall or office tower the chilled-water plant belongs to the building, and its efficiency is audited against a threshold. A household on a split system meets the figure nowhere, because there is no plant to audit.

Notice what the ton is doing in all of that. It sits below the line, as the denominator of an efficiency ratio, and never as a claim about how big anything is. Inside the same regulatory family, equipment capacity stays in kilowatts: the floor-mounting rule for air handlers is pitched at units above 35 kW.

The ton is exact, and almost nothing sold to a flat here is built to it. NEA's air-conditioner register held 340 models at the July 2026 snapshot. Eleven of them sit within 1% of a whole number of tons. Tighten that to half a per cent and five survive.

Capacities cluster somewhere else entirely. The most repeated values on the register are 4.81 kW, 6.64 kW, 3.43 kW and 9.32 kW, which work out to 1.37, 1.89, 0.98 and 2.65 tons. Only two models land on 3.52 kW, the two-decimal rounding of a single ton, out of all 340.

Those 340 models carry 218 distinct capacity values between them. That is a continuum, not a ladder. The ton describes a neat grid, and the register was never laid out on it, because manufacturers build to the watts their test standard reports and to their own model families.

A residential quote naming tonnage has named a bracket somebody chose. Read it the way a horsepower rating gets read: ask for the model number, then look up the registered capacity in kilowatts, since that one resolves to a published record and the tonnage figure does not.

The gap between bracket and record is where money moves. Two systems can wear the same tonnage label while their registered capacities sit hundreds of watts apart. The label was never precise enough to carry the difference.

Two decimal places is where the register does its work. A model at 3.43 kW and one at 3.52 kW both answer to 1 ton in loose talk, and the register holds them 0.09 kW apart. A tonnage label has nowhere to put that gap; rounding to a whole ton discards the digits that tell the two machines apart.

Check where a tonnage figure came from. A supplier converting from a model's registered kilowatts is doing arithmetic, and that number runs backwards on request; a supplier reaching for a familiar size class is doing something else, and that number does not run backwards. Both look identical on paper, which is why the question has to be asked out loud.

Put every figure into one unit before comparing anything, and make that unit the kilowatt. The label uses it, the register uses it, and the test standard behind both uses it.

The costliest misreading is not a conversion error at all. Efficiency in kW/RT and capacity in kW look alike on a page and measure opposite things. One counts the electricity a plant spends per unit of cooling. The other counts the cooling. A commercial quote showing 0.68 kW/RT describes an efficient plant, never a small one, and lower is better in that column.

A household version runs on the same confusion. A spec sheet states a cooling kW and a power kW, and treating the smaller as the machine's size makes a healthy unit look feeble. Check which of the pair a quoted kilowatt belongs to before comparing.

Ask for different things depending on the tier the job sits in. For a flat, ask for the model number and the rated cooling capacity. For a plant, ask for the design cooling load in RT and the operating efficiency in kW/RT. Name the hours as well. A reading taken at full load on a cool morning is not the number the building lives with.

Write the unit beside every figure that gets copied down. Most mixed-unit errors begin as a bare number in a message thread, where 3.5 could be the cooling, the power draw or somebody's rounded ton. A figure carrying its unit survives forwarding; a bare one gets reinterpreted.

A quote that will not commit to a unit has not committed to a number. Tonnage, horsepower and BTU all narrow a conversation, and none resolves to a record the way a kilowatt figure does. Convert once, keep the source figure beside the converted one, and compare like against like.