Sizing Radiators, at Boiler and at Heat-Pump Temperatures

Rooms disagree about what warm means

Before any surface is measured, fix the internal design temperature for each room, because it sets the temperature difference every subsequent calculation is multiplied by. UK dwelling practice, as set out in the CIBSE Domestic Heating Design Guide and carried through into MCS heat-loss work, assigns a temperature per room: the bathroom is designed warmer than the bedroom next to it, and the hall is designed cooler than the living room it opens off. US residential practice under ACCA Manual J takes a different route and applies one indoor design temperature — commonly 70 °F (21.1 °C) — across the whole dwelling, doing its room-by-room work through the surfaces rather than through the setpoints. Canada's National Building Code takes a third: a dwelling's heating is sized to CSA F280 and must hold at least 22 °C (71.6 °F) in every living space (NBC 2020, Articles 9.33.3.1 and 9.33.5.1).

Neither convention is wrong, but they are not interchangeable and a survey that mixes them produces radiators that are too big upstairs and too small in the bathroom. Pick the convention that matches the standard you are working to and write it at the top of the sheet.

The per-room convention has a consequence people miss: internal partitions become loss paths. A bathroom at 22 °C shares a stud wall with a bedroom at 18 °C, and four kelvin across eight square metres of uninsulated partition is real heat that has to come out of the bathroom emitter. Whole-house rules of thumb erase it, which is one reason the bathroom is so often the room the owner complains about.

Loss, one element at a time

Fabric loss is the area of each element multiplied by its thermal transmittance and by the temperature difference across it. The reason it is done element by element rather than as a single averaged envelope is that the transmittances present in one ordinary room differ by roughly a factor of ten: a filled cavity wall, a suspended timber floor over a ventilated void, and a twenty-year-old sealed unit are three completely different rates of loss occupying the same room. BS EN 12831-1 formalises this build-up for European work; ACCA Manual J does the equivalent job in the US, and Canada's National Building Code sizes a dwelling's heating to CSA F280.

Run every element that faces something colder: external walls and their openings, the ceiling where the room is on the top floor, the ground floor or any floor over an unheated void, and partitions to a garage, porch or store. That includes the party wall where the neighbour is genuinely unheated, which is worth asking about rather than assuming.

Ventilation loss is calculated separately and added: in metric work, the volumetric heat capacity of air at about 0.33 watt-hours per cubic metre per kelvin, times the room volume, times the air change rate, times the same temperature difference. The air change rate is where surveys go soft. An assumed figure for a room with a working chimney, trickle vents and a forty-year-old sash is a guess that can carry a third of the load, so measure it where the answer matters and record the assumption where you cannot.

The external figure driving all of it is a statistical design condition for the site, published as climatic design data — not a monthly mean, and not the coldest night anyone can remember. Apply one such figure consistently to every room in the building.

Where the R in that box comes from

Thermal resistances add in series through an assembly, so a wall build-up is the sum of every layer plus the surface air films on both faces. BS EN ISO 6946 sets out the calculation and the standard surface resistances; the interior and exterior films together are not a rounding error, and omitting them makes a modest assembly look worse than it performs.

On a retrofit survey the hard part is not the arithmetic but the input. You cannot see into a cavity from a bedroom, and the difference between an unfilled 1935 cavity and one blown in 1998 is most of the wall load. Where no boroscope inspection is possible, use the published default assumptions for the age band, mark them as assumptions on the sheet, and tell the client the radiator schedule inherits that uncertainty.

Two corrections belong on the number before it goes anywhere. Framed assemblies perform below their insulation layer because studs, plates and rim beams conduct around the batts, so apply a framing-fraction correction and use the effective figure. And where board has been added inboard of a solid wall, what counts is the finished build-up rather than the thickness on the invoice: a service void, a dot-and-dab air gap or an unclosed perimeter all bypass it.

The specification is in U and the field wants R

Building regulations, window certificates and manufacturers' assembly data are almost always expressed as transmittance, because transmittance is what multiplies straight through a heat-loss calculation. Insulation is marketed as resistance, because a bigger number reads better on a bag. The two are exact reciprocals, and the only rule that matters is the order of operations: total the resistances of every layer first, then invert once at the end. Adding transmittances layer by layer is not slightly wrong, it is wrong by a wide margin, and it is the single most common arithmetic error in retrofit paperwork.

There is a second trap in the same place, and it costs a factor of 5.678. A U-value written in W/m²K and a U-factor written in BTU/hr·ft²·°F are different units of the same quantity, and a regulation-style figure of 1.4 W/m²K is about 0.25 in imperial terms — which is a perfectly plausible-looking number in the wrong system. Windows make this worse because both conventions are alive in the same market: an NFRC 100 label states a whole-window U-factor in imperial units, while a unit certified under BS EN ISO 10077 states it in metric. Check which one is printed on the paper in front of you before typing anything.

The glass is where the room leaks

In most rooms the glazing is a small fraction of the envelope area and a large fraction of the loss, which is why a floor-area rule of thumb reads low precisely in the rooms with the bay, the patio doors or the rooflight — the rooms the client notices first. Use the whole-window figure there, never the centre-pane figure: a sealed unit and the frame around it perform differently, and the certified whole-window value takes in the frame, the spacer and the edge effects a glass-only number leaves out. Across older stock the spread between products all sold as double glazed is wide enough on its own to move a radiator by a size.

Sequencing matters more here than anywhere else in the survey. If glazing is being replaced in the same programme as the heating, size the emitters against the house that will exist when the job is finished, not the one you are standing in. Get that the wrong way round and you have specified radiators for a fabric that is about to improve, which is how a house ends up with panels that overshoot on every mild day and a system that cannot hold a low flow temperature because the emitters short-cycle the room past setpoint.

What a catalogue watt is a measurement of

A European radiator output is a test result declared under BS EN 442-2, taken in a standardised chamber at 75 °C flow, 65 °C return and 20 °C room — a mean water temperature of 70 °C, and therefore an excess over the room of 50 K. That is the ΔT50 figure on the price list, and it is the reason two panels of the same physical size from different makers can be compared at all.

Output does not scale linearly with that excess. The characteristic relationship is a power law: output equals a constant for the emitter multiplied by the excess temperature raised to an exponent, and for pressed steel panels that exponent typically lands between about 1.24 and 1.35. The EN 442 certificate for a specific product carries the actual constant and exponent, and the manufacturer publishes correction factors derived from them. The table below uses 1.30, which is representative of steel panels and close enough to plan with, but for a schedule that is going to be built, take the maker's own factors — a cast iron column radiator, a fan-assisted convector and a towel rail do not share an exponent.

Two historical rating bases still turn up on site and both mislead. UK catalogues predating EN 442 quoted at ΔT60, from the old 90/70/20 regime, and a ΔT60 figure restated at ΔT50 is only about 79 percent of what it said — so an old data sheet flatters a radiator by more than a quarter. In North America the older convention is square feet of EDR, in which cast iron was rated at 240 BTU/hr per square foot with steam at 215 °F and around 150 BTU/hr per square foot with water at a 170 °F mean; a cast iron radiator does not have a watt figure stamped anywhere on it, and the EDR route is how you get one. Finned-tube baseboard is different again: the Hydronics Institute I=B=R rating practice publishes output per foot at stated average water temperatures with 65 °F entering air, and because the fin geometry does not follow the panel exponent, use those published tables directly rather than correcting a single rating.

Then there is the unit split itself. A UK schedule is in watts at ΔT50, a US catalogue in BTU/hr at a stated average water temperature, a heat pump datasheet in kilowatts at a stated flow temperature — and all three land in one folder the moment imported equipment is involved. Convert once, at a defined point, and record the direction.

The same radiator at forty-five degrees

Now put the two halves of the survey together, because this is the moment the job is decided. A heat pump running 45 °C flow and 40 °C return into a room at 20 °C is working at an excess of 22.5 K, which is a correction factor of roughly 0.35. Every existing radiator in the house is about to deliver a third of the figure it was chosen for. A 1,600 W panel that comfortably covered a 1,200 W bedroom on a boiler becomes a 566 W panel, and the bedroom is 634 W short.

That arithmetic, applied room by room, is the survey. Measure each emitter, identify its type, look up the declared output, apply the factor for the design flow temperature, set it against the calculated room load. Some rooms pass, usually bedrooms with a modest setpoint and a generous panel. Some are one type deeper away from passing. And one or two — the bathroom with no free wall, the through-lounge with the bay — cannot be met at that temperature by anything that fits.

The trap is what happens next. Raising the design flow temperature to rescue those two rooms raises it for the whole house, and the machine pays for it everywhere: manufacturers' capacity tables generally show something in the order of two to three percent of coefficient of performance lost per kelvin of flow temperature, so a five-degree concession to one bathroom is a permanent tax on every kilowatt-hour the system ever delivers. Take the number from the specific machine's own table rather than from a rule, and price the alternative — a different emitter in the two failing rooms — against it explicitly.

Three documents police this. MCS MIS 3005 requires a calculated room-by-room heat loss and emitters assessed at the design flow temperature, not inherited from the boiler being replaced. ACCA Manual S requires selection against the Manual J load using the manufacturer's expanded performance data at the design condition. And Approved Document L, Volume 1 sets a maximum design flow temperature of 55 °C for wet systems in new English dwellings, pushing replacements in existing ones toward the lowest practicable temperature — check the edition in force, because that requirement has moved more than once.

Four times the water for the same heat

Bigger emitters at lower temperatures change the hydraulics, and this is the consequence most radiator schedules ignore. The flow a circuit carries is its heat output divided by the specific heat capacity of water and by the temperature drop across the emitter, so a boiler system designed on a 20 K drop and a heat pump system designed on a 5 K drop stand four times apart in flow rate for exactly the same kilowatt into the room. Friction does not respond politely to that. Pressure loss climbs with flow to a power near 1.85, so quadrupling the flow multiplies the friction loss rate by roughly thirteen. Tails, valve bodies and lockshields that were unremarkable on the old system become the restriction, and the effect concentrates in the fittings rather than the straight runs. Fifteen-millimetre tails feeding a heavily upsized emitter are the classic finding, and no amount of pump is a cure for a bore.

Work the loop that resists most — furthest out, smallest bore, most fittings — and see what duty it now demands. If the pump has to work substantially harder than before, revisit the temperature drop before revisiting the pump: widening the design drop reduces the flow, but it also lowers the mean water temperature in the emitter and therefore its output, which sends you straight back to the correction table. That loop between emitter size, temperature drop and flow is the real design work in a low-temperature retrofit, and it is why a radiator schedule and a pipework layout cannot be produced by two people who never speak.

Rooms where no radiator fits

Every survey turns up at least one room where the arithmetic is satisfiable and the geometry is not. A bathroom with a bath along one wall, a door, a window and a basin has perhaps 700 mm of free wall, and 700 mm of anything will not make the output a 22 °C setpoint demands at heat-pump temperatures. A galley kitchen is often worse, because the only free wall is the one the units are on. The honest responses are changes of emitter type, not changes of size. A fan-assisted convector gives several times the output of a passive panel off the same wall length, specified from its own data at the actual water temperature with fan speed and noise level as real selection criteria. Underfloor heating removes the wall problem outright and suits low water temperatures better than any panel. A towel rail is the weakest option and should never be counted at its catalogue figure in a room it is meant to heat, because that figure assumes no towels on it. In care settings, HSE guidance on accessible hot surfaces caps what a resident can touch at 43 °C, which rules out several product families before sizing begins.

There is also the room that is simply not on the wet system at all: a loft conversion, a garden office, a study stranded at the end of a microbore leg. Electric panel or baseboard heating answers those, and it is sized far more crudely, on watts against floor area. Know what that basis assumes — average insulation, a standard ceiling height, nothing at all about glazing, exposure or setpoint — and treat a wide gap between it and a proper room loss as a prompt to re-check inputs, with the element-by-element figure carrying the authority.