U-values and glazing specification

What the number is counting

The unit is watts per square metre per kelvin, written W/m²K. A kelvin and a degree Celsius are the same size, so a unit with a U-value of 1.2 loses 1.2 watts through every square metre for every degree of difference between the room and the outside air.

Put real numbers on it. A three square metre flat rooflight at 1.2 W/m²K, with the kitchen at 20 degrees and a Chelmsford January night at 2 degrees, is losing about 65 watts continuously. Swap that unit for one at 0.8 and the loss drops to about 43 watts. That is the whole of what the number does. It is a steady state figure, measured under laboratory conditions with no wind, no sun and no draught, which is why it describes the product rather than the building.

Ug, Uf and Uw, and the letters people mix up

Three subscripts matter. Ug is the centre pane figure for the glass alone, measured away from the edges. Uf is the frame. Uw is the whole product, the number that describes what actually gets fitted.

They are combined by area weighting, plus a correction for the perimeter of the glass:

Uw = (Ag × Ug + Af × Uf + Lg × Ψg) ÷ (Ag + Af)

Where Ag is the glazed area, Af is the frame area seen from outside, Lg is the total length of the glass perimeter and Ψg is the linear thermal transmittance at the edge of the pane in W/mK. The consequence is simple and it catches people out constantly: Ug is always the best looking of the three, and it is the one that appears in advertising. Uw is always higher. A rooflight with a Ug of 1.0 will typically land somewhere near 1.3 to 1.5 as a whole product.

Why the same unit scores differently at different sizes

A small rooflight performs worse than a large one made from identical materials, and the reason falls straight out of the formula above. Frame and edge are the weak parts, and a small unit has proportionally more of both.

Take a 550 by 780 millimetre roof window, the size that fits between trusses on the estates at Great Notley and South Woodham Ferrers. Its glass perimeter is roughly 2.4 metres against about 0.3 square metres of glass. A 1340 by 1400 unit has around 5.1 metres of perimeter against 1.6 square metres of glass, so the edge effect is spread over five times the area. The same product can differ by 0.3 W/m²K between those two sizes. This is why any declared figure should be accompanied by the size it was declared at, and why the standard reference size for windows exists at all.

Why glass performs worse lying down than standing up

A sealed unit in a wall and the same unit in a flat roof do not behave the same way, and the difference is convection inside the cavity.

In a vertical unit, gas warmed by the inner pane rises up the inside of the cavity, cools against the outer pane and falls, forming a slow loop that carries some heat across. In a horizontal unit the warm pane is underneath and the cold pane is on top, which is the classic condition for buoyancy driven overturning. The gas rises directly and continuously against the cold surface. Heat moves across the gap more readily as a result, and the same build-up typically declares a Ug something like 0.2 to 0.4 W/m²K worse in the horizontal plane than in the vertical.

Which standard the figure was tested to

Ask what the number was produced by, because there are two routes and they are not interchangeable.

The first is measurement. A hot box test to BS EN ISO 12567-2 puts the complete rooflight in a rig and measures the heat crossing it, with roof windows tested at an inclination rather than upright. That is the honest number. The second is calculation to BS EN ISO 10077-1 and 10077-2, which computes the whole product from the glass, the frame profile and the spacer using modelled thermal conductivities. Calculation is legitimate and it is what most of the market uses, but it depends on the modeller’s assumptions. Where a data sheet declares a figure under BS EN 14351-1 or, for plastic rooflights, BS EN 1873, the route used should be stated with it.

Cavity width, and the point where more gap gives nothing back

The gas gap between the panes carries most of the insulation, and its width has an optimum rather than a direction.

Below about 12 millimetres the panes are close enough that gas conducts heat across efficiently and the unit underperforms. Between 14 and 16 millimetres, argon reaches its best figure. Push past about 18 millimetres and convection sets in inside the cavity, which starts giving the benefit back. Sixteen millimetres is the practical answer for a rooflight and it is what a good unit uses. Where a slim frame or a conservation profile limits the overall thickness, a narrower cavity filled with krypton recovers most of the loss at a real cost. This is one of the few places where a thicker unit is genuinely a better unit up to a point, and no better beyond it.

What the coating and the gas contribute

Two invisible things do most of the work. A low emissivity coating on the cavity face of the inner pane reflects long wave infrared back into the room, and it is worth roughly 1.0 W/m²K on its own. The gas fill is worth another 0.2 or so against dry air.

Neither is exotic and both should be assumed present on any current unit. The detail that separates products is which surface the coating occupies and how many coatings the build-up carries, which is set out on the page covering low emissivity coatings and argon fills. If a quote does not mention a coating at all, the unit is either very old stock or the person writing the quote has not read the data sheet.

What the edge of the pane takes away

The spacer bar around the perimeter of the sealed unit is a conductive path from the outer pane to the inner one, and it is the reason Uw is always worse than Ug.

An aluminium spacer carries a Ψg around 0.08 W/mK. A stainless steel or composite warm edge spacer runs nearer 0.03 to 0.05. On a unit with five metres of glass perimeter, that difference alone shifts the whole product figure by something in the order of 0.1 W/m²K, and it changes the surface temperature at the edge of the glass by rather more than that suggests. Why the edge is where trouble shows up first is covered separately under warm edge spacers.

Frame section, and the depth it needs to work

The frame is between a tenth and a third of the visible area of a rooflight, and its Uf ranges from about 1.0 for an insulated timber or polyurethane section to well over 3.0 for an aluminium profile with a token thermal break.

What makes a frame perform is the depth of the break and whether there is anything insulating behind it. A polyamide break of 24 millimetres or more, with the cavity behind it filled rather than hollow, is doing real work. A 14 millimetre break in an otherwise solid aluminium section is a formality. Timber and polyurethane frames start ahead because the material itself insulates, which is why VELUX, Fakro, Keylite and Roto all offer them in the same sizes as their aluminium clad ranges.

Declared performance against installed performance

The declared figure describes the product in a rig. What the building gets depends on what the product was set into.

A rooflight at 1.2 W/m²K sitting on an uninsulated timber kerb is surrounded by a band of construction performing far worse than the glass, and the heat takes the easy route. The kerb is a separate subject in its own right and it is covered under thermal bridging at the rooflight perimeter. The point to carry from here is that specifying an excellent unit and building it into a poor upstand throws away most of what you paid for.

The g-value sitting alongside it

Every glazing specification has a second number pulling the other way. The g-value is the fraction of solar energy that passes through, and lowering it is how a room stays usable in July.

The two numbers are set together or not at all. A unit optimised purely for U-value, with clear glass and a soft coat low-E on the room side of the cavity, will run a g-value around 0.6 and deliver a great deal of unwanted heat into a south facing kitchen in summer. The summer half of the calculation is worked through on the solar gain and overheating page. A specification that names only a U-value is half a specification.

Reading the data sheet a manufacturer sends

Manufacturer literature is generally accurate and occasionally arranged to flatter. Three habits keep you straight.

Check whether the headline figure is Ug or Uw, because they will be within a few lines of each other and only one is the product. Check the size the figure applies to, since a table of U-values by size is a sign of an honest data sheet and a single number for a whole range is a sign of a marketing one. And check the plane, because a company that also makes vertical windows may have quoted a vertical test. The general case for why any of this pays back is set out on the energy efficient rooflights hub.

Checking the unit that arrives is the unit that was ordered

Sealed units carry their identity in the edge seal, and it is worth ten seconds on site before anything is lifted onto a roof.

The spacer is usually stamped or printed along its length with the maker, the build-up and often the manufacturing week. Warm edge spacers are visually distinct from aluminium: a composite or stainless spacer is dark or matt where mill finish aluminium is bright. The laminated pane can be identified by the fine line of the interlayer at the edge of the glass. Sunspire checks the glazing against the order before the unit leaves the ground, because a wrong unit identified on the driveway is an inconvenience and a wrong unit identified on the roof is a day.

Where the U-value stops being the number that matters

Past a certain point, further improvement in the glass changes very little in the room, and other things change a great deal.

Air leakage around the opening moves more heat than the difference between a 1.2 and a 1.0 unit. So does an uninsulated reveal lining. So does comfort itself, which is about the radiant temperature of the surfaces around you rather than the air temperature: a cold pane overhead makes a room feel cold at the same thermostat setting, which is one of the strongest arguments for going further on the glass than the regulation requires. Once the unit is at 1.0 or better, the money is better spent on the perimeter, the reveal and the ventilation strategy than on a fourth decimal place.