U-values, and what the number is measured across

Typical figures

• Single glazed, 1970s: around 5.0
• Early double glazed, no coating: around 3.0
• Modern double, low emissivity and argon: 1.2 to 1.4
• Triple glazed: around 0.8

The jump from an old unit to a modern one is dramatic. The jump from modern double to triple is much smaller.

The number that gets quoted

This is where comparisons go wrong.

Centre pane U-value measures the glass in the middle of the pane, away from the edges. It ignores the spacer bar and the frame, which are exactly where heat escapes. It is the flattering number.

Whole unit U-value includes the frame, the spacer and the glass together. It is what the building actually experiences and it is the number Building Regulations use.

Whole unit is always worse than centre pane, often by 0.4 or more. If one quotation shows 1.0 and another 1.4, check whether you are comparing the same measurement before concluding anything.

What Part L expects

For a replacement rooflight in an existing dwelling, no worse than 2.2 whole unit. Any decent modern unit clears that comfortably, which is why it is rarely discussed.

The bit nobody quotes

The kerb. A rooflight U-value describes the unit, not the upstand it sits on, and an uninsulated kerb frequently loses more heat than the glass.

There is no number on a quotation for that, which is why it is worth asking directly whether the kerb is insulated. More on thermal bridging at the perimeter.

What a U-value actually is

A U-value measures how readily heat passes through a building element. It is expressed in watts per square metre kelvin, and it describes how many watts cross each square metre for every degree of temperature difference between inside and out.

Lower is better. An insulated flat roof is around 0.15. A solid Victorian wall is around 2.0. A modern rooflight is 1.2 to 1.4.

The number is straightforward. What causes almost all the confusion in rooflight quotations is that two different U-values describe the same product, and only one of them is the one the building experiences.

Centre pane against whole unit

Centre pane measures the glass alone, away from the edges. It ignores the spacer bar and it ignores the frame entirely. It is a genuine measurement of the glazing, and it is not a measurement of the rooflight.

Whole unit includes glass, spacer and frame, weighted by the proportion of the unit each occupies. It is what Building Regulations use and it is what the room feels.

Whole unit is always worse, frequently by 0.4 or more. On a small rooflight, where the frame is a large proportion of the total area, the gap is wider still.

So a rooflight advertised at 1.0 and another at 1.4 may be identical products described two different ways. Comparing them without establishing which figure you have been given is comparing nothing.

The practical rule: ask for whole unit figures in writing, on the quotation, alongside the model and size code. A supplier quoting centre pane is not necessarily being dishonest, and you still cannot compare on it.

Why the edges matter so much

The difference between the two figures is not a technicality. It is a description of where the heat actually goes.

In the middle of a sealed unit, two panes with a low emissivity coating and an argon filled cavity perform well. Heat crossing that route has to conduct through glass, then radiate across a gas filled gap where the coating reflects most of the infrared back, then conduct through glass again. That is a genuinely good insulator for its thickness.

At the perimeter, the two panes are connected by the spacer bar, and whatever the spacer is made of provides a direct conductive path from the warm pane to the cold one. Around that sits the frame, and beyond the frame the junction with the structure.

On a typical rooflight the perimeter is where most of the avoidable loss is, and it is exactly the part the centre pane figure excludes.

The spacer bar

Traditionally aluminium: cheap, easy to form, dimensionally stable, and an excellent conductor of heat. That last property is the problem.

An aluminium spacer bridges the panes around the entire edge of the unit. Heat travels along it from the warm side to the cold side, and the glass edge runs several degrees colder than the centre.

The visible consequence is condensation forming as a band around the edge of the pane on cold mornings while the middle stays clear. It is extremely common on units fitted before the mid 2000s and it is regularly mistaken for a failed sealed unit, which it is not.

A warm edge spacer, made from thin walled stainless steel, a structural foam or a composite, conducts far less. It improves the whole unit figure by around 0.1, which is modest, and it improves the thing you actually notice a great deal, because the edge misting largely disappears.

On a rooflight it matters more than on a wall window, because a rooflight faces the cold night sky directly and its edges get colder.

The frame

Timber is a reasonable insulator and it is why most quality roof windows have laminated timber frames with an aluminium cladding outside: the cladding takes the weather and the timber does the insulating.

Aluminium on its own conducts extremely well, which is why every reputable aluminium framed product, and every lantern worth buying, has a thermal break: a section of low conductivity material within the profile that interrupts the path from inside to outside.

An aluminium framed lantern without a thermal break is a set of cold bridges holding glass, and it shows as condensation on the internal frame members long before the glass misses up.

Steel conservation rooflights conduct too. Good ones are thermally broken, and because the profiles are slim by design the frame is a smaller proportion of the unit, so the whole unit figure is better than you might expect.

The part that appears on no specification at all

The kerb, on a flat roof, and the insulation collar on a pitched one. Neither is part of the rooflight, so neither appears in any U-value, and on many installations the kerb is the largest single loss.

Consider a 1200 by 1800 rooflight. That is 2.16 square metres of glazing and six metres of perimeter. An uninsulated timber kerb 150mm high presents very nearly a square metre of additional envelope, punching straight through the roof insulation, at a U-value several times worse than the glass sitting on top of it.

In other words, a poorly detailed kerb can lose comparable heat to the entire rooflight above it, and no amount of glazing specification compensates for it.

Insulating the outer face of the kerb is a sheet of rigid board and an hour. It disappears under the covering, which is precisely why it gets skipped, and it is the single most useful question you can put to anyone quoting for a flat roof rooflight.

On a pitched roof the equivalent is the insulation collar, which closes the gap between the frame and the structure. Without it the reveal runs cold enough for room air to condense on it, which is both a heat loss and the commonest cause of water appearing on a plaster reveal months after the work.

Air leakage, which no U-value describes

A U-value measures conduction. It says nothing about air moving through gaps, and air leakage around an opening loses more heat than conduction through a small area of glass.

The defence is a continuous vapour control layer, sealed to the frame of the rooflight, so that warm moist air from the room cannot pass into the construction. It stops heat leaving and it stops moisture reaching a cold surface inside the build up, where it would condense out of sight.

That is a detail rather than a product, and it is not visible when the job is finished.

What the regulations ask

For a replacement rooflight in an existing dwelling, Part L expects a whole unit U-value no worse than 2.2. That is a compliance threshold rather than a target, and any current double glazed unit with a low emissivity coating, argon fill and a warm edge spacer lands at 1.2 to 1.4.

So compliance is rarely the question. It bites harder on an extension, where the rooflights form part of the thermal calculation for the whole structure and a poor unit forces better performance somewhere else in the build, and where there is a limit on total glazed area as a proportion of floor area.

Getting the glass right first is cheaper than compensating for it in the walls afterwards.

Where triple glazing sits

It takes a whole unit figure from around 1.2 to around 0.8. That is a real improvement and a modest one, and it comes with four costs.

It transmits noticeably less daylight, which works directly against the reason a rooflight exists. It is significantly heavier, which shortens actuator life on an opening unit and affects the structure on a large one. It costs substantially more. And it does very little about solar gain, which on a sunward facing rooflight is the complaint people actually have.

Spend the same money on solar control glass, an insulated kerb and a properly sealed vapour control layer and the room will feel better through the whole year.

The other number: the g value

A U-value describes heat leaving. The g value describes solar energy arriving, on a scale from 0 to 1.

A clear double glazed rooflight is around 0.6 to 0.7; a solar control unit 0.3 to 0.4. It is not regulated for a replacement, and on a south or west facing rooflight it matters more to how the room feels than the U-value does, because overheating is the most common regret about rooflights.

Ask for both figures, and read them together with the light transmission percentage, because a unit that stops heat by being dark is not a good rooflight.

What to ask for on a quotation

The manufacturer, model and size code. The whole unit U-value, stated as whole unit. The g value and the light transmission where the unit faces the sun. Confirmation of a warm edge spacer. And, separately from the glazing, confirmation that the kerb is insulated on its outer face or that an insulation collar is included.

Those six lines tell you more about how the installation will perform than any brand comparison will.

Asking for the figure at your size

The whole unit U value changes with the size of the unit, because the perimeter is a larger proportion of a small one.

A manufacturer quoting a single figure for a range is quoting it at a particular size. Ask for the number at the size you are actually buying, and compare like with like across the shortlist.