U-values and glazing specification

What the figure counts, and what sits outside its boundary

The U-value is a steady state measurement. It assumes a fixed temperature either side, no sun on the glass, no wind beyond a defined convention, and no air passing through the assembly. Under those conditions it counts every watt crossing the unit from the room air to the outside air, divided by the area and by the temperature difference.

Two things sit outside that boundary and both matter on a roof. Solar gain is excluded entirely, which is the g-value’s job. And air leakage is excluded, because the specimen is sealed into the test rig. A unit rated at 1.0 with a draughty perimeter loses more heat in practice than a unit rated at 1.3 sealed properly, and nothing in the U-value will tell you that.

Conduction, convection and radiation across a sealed cavity

Heat crosses the gap between two panes by three mechanisms, and knowing their relative sizes explains why every improvement in glazing has targeted the same two things for thirty years.

Conduction through the gas is set by the gas itself. Convection is the gas circulating in a loop, set by the cavity width, the temperature difference and the angle of the unit. Radiation is the warm inner pane radiating long wave infrared straight across the gap to the cold outer pane, and in an uncoated air-filled cavity radiation carries roughly two thirds of the total. That is why a coating that suppresses radiation delivers a bigger improvement than anything done to the gas, and why the two are always specified together. The mechanics of both are set out on low-emissivity coatings and argon fills.

Working the arithmetic through on one roof window

Take a unit 780 by 1180mm overall, which is a common size in a Gloucester rear addition. The frame takes 90mm all round, so the visible glass is 600 by 1000mm.

Overall area is 0.92 square metres. Glass area is 0.60, frame area is 0.32, and the perimeter of the glass is 3.2 metres. Put in a glass value of 1.0, a timber frame value of 1.4 and a warm-edge psi of 0.05, and the sum runs: 0.60 multiplied by 1.0 is 0.60. Then 0.32 multiplied by 1.4 is 0.448. Then 3.2 multiplied by 0.05 is 0.16. Add those to get 1.208 watts per kelvin, divide by 0.92 square metres, and the whole-unit figure is 1.31 W/m2K.

The glass was 1.0. The unit is 1.31. Nothing has gone wrong: the frame and the glass edge have simply been counted, and together they add nearly a third to the answer.

Surface resistances, and heat that is travelling upwards

Every U-value calculation includes two thin layers of still air, one clinging to the inside surface and one to the outside, expressed as surface resistances Rsi and Rse. They are part of the number, and they are not the same for a roof as for a wall.

For heat flowing horizontally through a vertical window the internal surface resistance is taken as 0.13 square metre kelvin per watt. For heat flowing upwards through a horizontal surface it drops to about 0.10, because warm air rising off the underside of the glass strips heat away faster than it does off a vertical pane. The external resistance is around 0.04 in both cases. Losing 0.03 of internal resistance sounds trivial and it is worth roughly 0.05 to 0.1 on the finished U-value of a good unit. Combine that with harder cavity convection at a shallow angle, described on the hub page, and a rooflight declared at 1.2 in the vertical plane is realistically nearer 1.4 to 1.5 lying at ten degrees on a flat roof.

Declared size against the size you are buying

Manufacturers declare performance at a fixed reference size so that products can be compared. For roof windows that reference is typically somewhere around 1.1 by 1.2 metres, and for made-to-size flat roof units the fabricator usually states the size the calculation was run at.

Buy a unit smaller than the reference and the real figure is worse. Buy a larger one and it is better. Neither is dishonest, and both are invisible unless the size is asked for. On a made-to-size rooflight the fabricator can generally run the calculation at your actual dimensions, which is a reasonable thing to request on a large order and takes them very little time.

Safety glass, laminates and what they do to the figure

Overhead glazing needs a toughened outer pane and a laminated inner pane wherever there is habitable space, a stairwell, a bath or a shower underneath. Both are non-negotiable in a roof, so the useful question is what they cost thermally.

Very little. Glass itself conducts at about 1.0 W/mK, so going from a 4mm pane to a 6.8mm laminate changes the resistance of the assembly by a fraction of a per cent. The polyvinyl butyral interlayer is a similar conductor. A laminated inner pane costs perhaps 0.01 on the whole-unit figure and buys a large reduction in rain noise, which is why it appears on almost every shallow-pitch specification regardless of the safety requirement. Rain noise itself is dealt with on acoustic glazing and rain noise.

Where laminating does change the arithmetic is weight and rebate depth. A thicker build-up needs a deeper frame rebate, and on a slim conservation profile that can rule out the triple unit you wanted, which pushes the specification back towards a better double.

Specifying to a requirement rather than to a product

The order of decisions matters more than the product chosen. Start from the room, not the catalogue.

Establish what the space is for and how it is heated. Establish the orientation and pitch of the slope, because those set the summer load. Establish the size the roof will take. Then set a target whole-unit U-value, a target g-value and a minimum light transmittance, in that order, and ask suppliers to meet them. A specification written as three numbers can be quoted against by anyone. A specification written as a product code can only be quoted by whoever sells that product.

Where the declared figure and the installed figure part company

A unit performs as declared only if the hole it sits in is built to match. Three things routinely undo it.

The first is insulation stopping short of the frame, leaving an air gap behind the plasterboard reveal that no inspection will ever find. The second is the vapour control layer trimmed to the board edge rather than taped to the frame, which lets room air move into the construction. The third is the kerb on a flat roof unit, insulated on the inside face only, so the timber runs cold from top to bottom and becomes a conductive ring around the glass. The linear heat loss at that junction is a subject in its own right and is set out on thermal bridging at the rooflight perimeter. The short version is that a poorly detailed perimeter can cost more than the entire difference between double and triple glazing.

Small openings on Cotswold stone, and the arithmetic that follows

On the stone slate roofs through Painswick, Bibury, Northleach and Chipping Campden, a conservation officer will usually accept a slim steel conservation rooflight and nothing else. Steel conducts at roughly 50 W/mK, the section is too slim to carry a thermal break, and there is often a central glazing bar adding a second conductive line across the pane.

Run the area weighting on a 600 by 900mm unit with a frame value near 2.5 and the answer lands somewhere between 1.6 and 2.0 rather than the 1.2 a timber roof window would give. The honest framing is proportion. That opening is half a square metre in a roof that is often uninsulated above a lime mortar torching, so it is not the weakest thermal element by a wide margin, and where a manufacturer offers a thermally improved conservation profile or a slim krypton-filled panel in the same rebate, that is worth pushing for at order. The consent side is on conservation rooflight installation.

Replacement units across the county’s older stock

A great deal of the work in Gloucester’s Victorian terraces, the Tewkesbury vale villages and the interwar estates at Longlevens and Hucclecote involves taking out a unit fitted between 1985 and 2005. Those sit at 2.8 to 3.0 W/m2K, sometimes worse where the cavity has lost its gas.

Dropping to a current unit at 1.2 halves the heat loss through that opening and roughly doubles the inner surface temperature margin above dew point, which is usually the change the householder actually notices. The limiting standard a replacement has to meet, and what gets recorded with the notification, is covered on rooflights and Part L for a replacement. The practical point at survey is that the old size code stamped on the sash tells you what is coming out, and the new specification is written against the room rather than against the old unit.

Where the specification gets fixed, and what happens next

The glazing specification is settled at survey, once the orientation is off a compass, the pitch is off a digital level, the rafter or truss depth is known and the room’s use and heating are recorded. From that comes a target U-value, a target g-value, a build-up and a reveal detail, all written down.

An indicative range comes back by email from photographs and an address, and a fixed price follows the survey, itemised so the glass upgrade is visible as its own line. Email info@severnskylightsgloucester.co.uk with the orientation of the slope and roughly what size you have in mind. Further reading sits in the written guides, and the sequence from enquiry to fixed price is on the costs page.