What a BTU Actually Is

Why this matters

Four people on the same job say "a hundred thousand BTU" and mean four different things: what the appliance burns, what it delivers, what the building needs on the coldest day, and what the customer used all winter. Three of those are rates and one is a quantity. Two of them are input and two are output. Nobody notices, because the unit is the same word each time, and the estimate that comes out the other end is oversized, mispriced against the fuel bill, or both.

None of this is advanced. It is bookkeeping. But it is bookkeeping that quietly decides equipment size, and equipment size decides comfort, cycling, and moisture removal for the next fifteen years.

What the unit is, precisely

A British thermal unit is a quantity of energy: roughly the heat needed to raise one pound of water one degree Fahrenheit in the range where water's specific heat is close to 1.0 Btu per pound per degree. That is the field definition and it is good enough for every calculation you will do. The formal definition no longer references water at all; the International Table Btu is fixed against the joule at about 1,055 J, which is why conversions between Btu and metric units are exact rather than approximate.

Two consequences you use constantly. Water's specific heat near 1.0 is why 500 x gpm x delta-T gives Btu/h on a water circuit, the 500 covering about 8.33 pounds per gallon times 60 minutes. Air's much lower specific heat and density are why the equivalent air constant is only 1.08 x cfm x delta-T at sea-level densities, and why moving heat with air takes so much more volume than with water.

Rate is not quantity, and the "per hour" is load-bearing

A Btu is an amount. A Btu per hour is a speed. Mixing them is the single most common unit error in the trades, and it is invisible in speech because people drop the "per hour."

The test that catches every one of these: write the units out beside the numbers and cancel them. A therm figure divided by an hours figure gives Btu/h and is an average rate over that whole span, which is not the same thing as the peak rate the equipment has to meet. Any sizing number derived from a bill without that distinction being made out loud is wrong.

Natural gas is usually spoken of as about 1,000 Btu per cubic foot, and that is a workable field approximation, but the actual heating value varies by supply and by month. The utility bills in therms precisely because it applies a measured heating value to your metered volume. When a calculation has to be tight, get the posted heating value for the billing period rather than assuming 1,000.

Input, output, and three efficiency numbers that are not interchangeable

A fuel-fired appliance nameplate typically lists both an input rate and an output rate. Input is the fuel energy going in. Output is the heat leaving into the water or the air. The gap is what went up the flue and off the jacket.

Three different efficiency figures describe that gap and they are not substitutes for each other.

Combustion efficiency is what a flue-gas analyzer reports at the moment of the test. It accounts for the sensible and latent losses in the flue gas at that firing condition and nothing else.

Steady-state or thermal efficiency is output divided by input with the appliance running continuously at rated fire. It is the number that connects the two nameplate figures.

AFUE is a seasonal measure produced under a standard test procedure that includes cycling losses and off-cycle losses across a heating season. It is not a divisor you apply to a nameplate input to get output on a design day, and treating it as one understates capacity. Use the nameplate output for sizing and AFUE for talking about seasonal fuel use.

Nominal, rated, and delivered

Three more words for what looks like one number.

Nominal is a label, usually rounded to a marketing increment. A "three ton" unit is a size class.

Rated is a measured capacity at a stated set of conditions defined by a rating standard. It is a real number and it is only true at those conditions.

Delivered is what the equipment produces on this building, on this day, through this duct or pipe, at these conditions. It is the only one the customer experiences, and the gap between rated and delivered is its own subject with its own card.

Before you clock a gas meter

• If you smell gas at any point, everyone leaves the building immediately. No switches are touched on the way out, no lights, no phone used inside, and the call goes to the gas utility's emergency number from outside. Do not go back in to shut anything off.
• Never use a flame to look for a leak. Use a listed leak-detection solution or a calibrated combustible-gas detector.
• To clock a meter, isolate the other gas appliances at their own controls, not at the meter, so a pilot or a control does not restart against an isolated supply. Restore each one and confirm it lights and proves before you leave.
• Carbon monoxide is odorless and the exposure route is inhalation, so wear a personal CO monitor while you work on any combustion appliance and do not adjust firing or air without a calibrated analyzer. If a building CO alarm sounds, treat it the same as the gas response above: everyone out, nothing switched, call from outside.
• Metering at the appliance control board is electrical work under 29 CFR 1910.333(b)(2), separate from mechanical lockout. Prove your meter live, prove the circuit dead, prove it live again, the sequence in NFPA 70E-2021, 120.5.

Worked audit: four unit errors in one estimate

A replacement boiler quote for a small commercial building. Every number on it was real and four of them were the wrong kind of number.

Measured firing rate. The 2 ft3 test dial took 48 seconds for one revolution. Rate = 2 x 3,600 / 48 = 150 ft3/h. The utility's posted heating value for that month was 1,030 Btu per ft3, so input = 150 x 1,030 = 154,500 Btu/h.

Error one, the assumed heating value. The quote had used 1,000 Btu per ft3 and reported 150,000 Btu/h. That is 3 percent low, which is minor on its own. It is worth correcting anyway because it costs nothing and because the habit of pulling the real value is what saves you on a supply where the deviation is larger.

Error two, input read as output. The nameplate showed 150,000 Btu/h input and 123,000 Btu/h output. The quote sized the replacement on 150,000 of "existing capacity." Against the measured input of 154,500, the delivered output of 123,000 gives a thermal efficiency of about 80 percent, which is consistent with the appliance type. Sizing the replacement to deliver 150,000 rather than 123,000 oversizes it by a factor of 1.22, or 22 percent, before any other error.

Error three, a season treated as a rate. The quote noted 850 therms used last winter. That is 85,000,000 Btu of fuel over the season, a quantity. Across a 210 day heating season, 5,040 hours, the average input rate is 85,000,000 / 5,040 = 16,865 Btu/h. Against the measured 154,500 Btu/h firing rate, the burner ran about 11 percent of the season's hours.

An 11 percent seasonal run fraction is a flag, not a size, because a season average is dragged down by every mild day in it. So pull a shorter, colder window. The coldest seven days in January consumed 96 therms, which is 9,600,000 Btu over 168 hours, an average input of 57,143 Btu/h and a run fraction of 37 percent. Even through the coldest week on that building's record, the boiler was idle roughly two thirds of the time, so its output exceeds what that week required by a factor of about 2.7.

Two conditions gate that figure and belong in the same breath as it. The coldest week is not necessarily the design condition, so the true design-day ratio is smaller than 2.7; and if the same appliance also serves domestic hot water, part of that gas was not space heating and the space-heating oversizing is larger than 2.7 suggests. Treat it as grounds for a heat loss calculation, never as the new size.

Error four, total read as sensible. The same quote listed a cooling replacement at 24,000 Btu/h and set it against a room load worked out in sensible terms. A cooling rating is total capacity, sensible plus latent, at the conditions of the rating standard. The sensible share at site conditions is less, and how much less depends on the entering air and the airflow. Comparing a total rating against a sensible load is a comparison between two different quantities that happen to share a unit.

What the audit changed. Three of the four errors pushed the same direction, toward oversizing, and they compound: a 22 percent input-for-output error on top of a size chosen without a load calculation, on a building whose own fuel record says the existing appliance was already well oversized. The corrected recommendation was a heat loss calculation before any equipment was selected, with the fuel record used as the sanity check on the result rather than as the input to it.

Verifying: the four fields every energy number carries

Before an energy number goes on a quote, on a report, or into a sizing calculation, write these four beside it. If any is blank, the number is not yet usable.

1. Rate or quantity. Btu/h or Btu. Written out, not implied.
2. Input or output. For anything that burns fuel or draws power.
3. Total or sensible. For anything on the cooling side.
4. At what condition. The rating point, the measured conditions, or the design condition, named.

Then run the arithmetic with the units attached and cancel them. Therms divided by hours must resolve to Btu/h and be labeled as an average over that span. Cubic feet times Btu per cubic foot must resolve to Btu. A calculation whose units do not cancel to what you claimed is wrong even when every number in it is right.

Where the unit itself changes meaning

Four contexts where a Btu is not simply a Btu, each of which shows up on real jobs.

Higher versus lower heating value. Fuel energy content can be counted with the water vapor in the flue gas condensed or not. North American practice prices gas and rates appliance efficiency on the higher figure, which includes that latent heat. That convention is why a condensing appliance recovering part of that heat still rates below 100 percent rather than above it, and why an efficiency figure quoted on the other basis is not comparable.

Site energy versus source energy. A kilowatt-hour at the meter is 3,412 Btu delivered. Producing and delivering it consumed considerably more primary energy. Comparing a fuel appliance and an electric one on site Btu answers a question about the building; comparing them on source energy answers a different one about total consumption. Say which you are making.

Weather-normalized comparison. Two winters are not the same winter. Comparing last season's therms against this season's without normalizing for heating degree days measures the weather at least as much as the equipment. Divide by degree days for the same period from the same weather station before claiming a change in performance.

Delivered versus produced. A Btu made at the appliance and lost from a duct or pipe in an unconditioned space never reaches the load. It is real energy and it does not count toward capacity at the room, so keep it as its own line rather than folding it into either the load or the rating.

References

• ASHRAE Handbook, Fundamentals volume, for unit definitions, fluid properties, and the sensible heat relationships behind the 500 and 1.08 constants
• Appliance nameplate data for input and output ratings, and the manufacturer's installation instructions for the rating basis
• Local gas utility posted heating value for the billing period, used in place of any assumed value per cubic foot
• 29 CFR 1910.333(b)(2) and NFPA 70E-2021, 120.5 for electrical work at appliance controls
• See related: How to Think About Capacity Versus Output; How Heat Gets Into a Space You Are Trying to Condition