Heat, sun, movement and holding it down (Roof Lights by the Numbers)

A roof light lives in a harsher climate than the room below it. In summer its glass and frame bake in direct sun; on winter nights they radiate heat to a clear sky. Heat flows in and out through it far faster than through the insulated roof it replaces, and its aluminium frame grows and shrinks by millimetres every day. And the wind that flows over the roof tries to lift it off its kerb.

This chapter puts numbers on all three. It compares RL-24’s summer gain with its winter loss, works out how much its frame and glass move and how the fixings allow for it, and sizes the fixings that hold the frame down against uplift. The failure case is a frame screwed tight to its kerb, whose corners tore open in its first summer.

Summer in, winter out

Chapter 1 estimated the summer peak solar gain through RL-24’s 2.53 m² clear opening: about 2,024 W with clear glass and 1,012 W with a performance laminate whose solar heat gain coefficient is 0.40. The owner chose the performance laminate, so RL-24 admits about 1 kW at a clear summer peak.

Winter heat loss is the glass’s U-value times its area times the temperature difference. Chapter 6 used teaching U-values of 5.8 W/m²·K for single laminated glass and 2.8 for a double-glazed unit. On a 2 °C night with the room at 21 °C, the difference is 19 K.

Figure 1RL-24's heat flows, summer peak and winter night

	Summer peak gain	Winter night loss, 19 K
Single clear laminate	about 2,024 W	5.8 × 3.178 × 19 ≈ 350 W
Single performance laminate, as built	about 1,012 W	about 350 W
Double-glazed performance unit	about 1,000 W, depending on the unit	2.8 × 3.178 × 19 ≈ 169 W
Insulated roof, same plan area	negligible	0.3 × 2.88 × 19 ≈ 16 W

This is an illustration with this book's teaching values. The insulated roof it replaces, about 2.88 m² of roof with a teaching U-value of 0.3 W/m²·K, would lose only about 16 W on the same night.

The last line is the sobering one. On a cold night, RL-24 loses about 21 times as much heat as the insulated roof it replaced. That is not a reason to leave it out; the daylight is worth having, and 350 W is a modest part of a whole house’s heating load. It is a reason to keep roof lights to the size the room needs, to consider double glazing in cold climates, and to include them honestly in the house’s energy assessment, which is required under the NCC and is a matter for an accredited assessor rather than this book.

The two numbers are not directly comparable, and it is worth saying why. The summer figure is a peak that lasts an hour or two around the middle of a clear day; the winter figure is a rate that lasts all night. Over a whole day, a roof light in winter also gains some useful solar heat while the sun is up, which partly offsets the night’s loss, and over a whole summer day the gain falls away in the morning and afternoon as the sun moves lower. An energy assessor’s software accounts for all of this, hour by hour, for the site’s climate. The point of the rough numbers is to show the designer and owner the size of each effect and which way a decision pushes them, before the assessment confirms it.

Shading

The most effective way to cut summer gain is to stop the sun before it reaches the glass. An external shade over a roof light is unusual on a house, because it is hard to maintain and spoils the view of the sky, but it works. Internal blinds are common and cut glare, but they stop the heat only after it has passed through the glass, so much of it is still released into the room. A performance laminate, as on RL-24, is the practical middle course: it reduces the gain permanently, at the cost of some daylight. Whichever is chosen, the solar heat gain coefficient of the whole assembly, not just the glass, is what matters for the summer numbers.

Movement

Aluminium expands about 23 millionths of its length for each degree of temperature rise; glass about 9 millionths. Those coefficients are small, but a roof light’s frame can range from near 0 °C on a frosty dawn to 70 °C or more in still summer sun, particularly in a dark finish. The shop uses 70 K as its teaching temperature range for a roof light frame.

Figure 2Thermal movement in RL-24 over a 70 K range

3.86 mmLong side of the eaves frame, 2,400 mm aluminium23 × 10⁻⁶ × 2,400 × 70

1.93 mmRidge and short frame sides, 1,200 mm aluminium23 × 10⁻⁶ × 1,200 × 70

1.07 mmOne common bar, 662 mm aluminium23 × 10⁻⁶ × 662 × 70

0.42 mmGlass along the same 662 mm9 × 10⁻⁶ × 662 × 70

0.65 mmDifference between bar and glass over a pane lengthtaken up in the 4 mm edge clearance

This is an illustration using expansion coefficients of 23 × 10⁻⁶ per K for aluminium and 9 × 10⁻⁶ per K for glass, from The Engineering ToolBox. Movement equals coefficient × length × temperature range.

Movement between glass and bars

Aluminium moves about two and a half times as much as glass for the same temperature change. Over a pane’s 662 mm length the difference is only about 0.65 mm, easily taken up by the 4 mm edge clearance from Chapter 3 and the gaskets. That clearance is also why glass must never be packed hard against a bar’s web: a pane fitted with no clearance on a cool morning will be squeezed by the bars on a hot afternoon, and laminated glass under edge pressure can crack.

Movement between the frame and the kerb

The frame’s long sides grow by nearly 4 mm over the temperature range, and the timber kerb they sit on hardly moves at all with temperature. If the frame is screwed rigidly to the kerb at close centres, that 4 mm has nowhere to go. The frame bows, the screws work loose in the timber or tear their holes, and the joints at the frame’s mitred corners, the stiffest points, are pulled open and shut every day.

RL-24’s frame is fixed so that it can slide. Each long side is fixed through a round hole at its midpoint, which locates it, and through slotted holes elsewhere, the slots running along the frame. The ends of a 2,400 mm side move about 1.93 mm each way from the fixed midpoint, so slots of the screw diameter plus about 6 mm give comfortable room. The screws have bonded sealing washers and are tightened to snug, not crushed, so the frame can slide under them. The short sides are treated the same way, with smaller slots. The frame’s corners are factory-mitred and mechanically cleated, so they move with the frame rather than against it.

Sealant joints that move

Wherever movement is taken up by a sealant joint, at a frame corner, between the frame and a flashing or between two lengths of capping, the joint must be wide enough for the sealant to stretch and squeeze without tearing. Every sealant has a movement capability, published by its maker as a percentage of the joint’s width: a sealant rated at ±25%, for example, can accommodate movement of a quarter of the joint width either way. The minimum joint width is then the expected movement divided by that capability.

For a joint that must take the full 3.86 mm growth of RL-24’s long frame side with a sealant rated at ±25%, the joint must be at least 3.86 ÷ 0.25 ≈ 15.4 mm wide, which is far wider than the hairline joint a mitred corner naturally has. That is why RL-24’s frame does not rely on sealant at its corners to take movement: the slots let the whole frame slide on the kerb, the corners are cleated rigid, and the sealant at a corner only has to seal, not stretch. The same rule applied to a capping joint, where two lengths of ridge capping meet, sets the gap left between them before the joint cover goes on. Joint depth matters too: the sealant maker usually specifies a width-to-depth ratio and a backing rod, so that the sealant bonds to the two sides of the joint and not to its base, and can stretch freely between them.

Hot glass, cold edges

Glass itself suffers from uneven temperature. In sun, the centre of a pane heats quickly, while its edges, held in the bars’ rebates and shaded by the cappings, stay cooler. The hot centre expands and pulls on the cool edge, putting the edge in tension. If the difference is large and the edge has a flaw, the pane can crack: a thermal fracture, which typically starts at the edge, at right angles to it, and runs inwards. Tinted and coated glass absorbs more sun and heats more than clear glass, and partial shading, from a nearby wall, a tree or an internal blind pulled tight against the glass, makes the temperature difference worse.

The glazier assesses thermal stress risk for the chosen glass, frame colour and shading, and may specify heat-strengthened plies, clean-cut or arrissed edges, or limits on blinds close to the glass. For RL-24’s performance laminate in a dark frame, the glazier checked the risk and specified edge treatment for every pane. It is another reason the glass is chosen by someone qualified to do so, not by price alone.

Holding it down

Wind blowing over a roof speeds up over the ridge and edges and sucks upwards on the roof surface. Chapter 3 used a teaching uplift of 1.2 kPa over the roof light. The real value comes from AS/NZS 1170.2 for the site’s wind region, terrain category, topography, the building’s height and the roof light’s position on the roof, and it is often much higher near roof edges and corners. A roof light near a corner of a roof may see well over twice the pressure of one in the middle.

Figure 3Uplift on RL-24 and its fixings to the kerb

uplift = pressure × plan area load per fixing = (uplift − self-weight) ÷ number of fixings

Uplift: 1.2 kPa × 2.88 m² ≈ 3.46 kN
Self-weight resisting it: about 48 kg glass + 25 kg frame and bars ≈ 0.72 kN
Net uplift: 3.46 − 0.72 ≈ 2.74 kN
Fixings: 7,200 mm perimeter at 300 mm centres = 24
Load per fixing: 2,740 ÷ 24 ≈ 114 N

Each screw must hold about 114 N in withdrawal from the kerb timber, with a margin from the loading standard's factors

This is an illustration with teaching values. A real design applies the load factors and combinations of the loading standard, including a reduced factor on the self-weight that resists uplift, and uses the screw maker's published capacity in the kerb timber.

About 114 N is a modest load for a suitable screw into sound treated timber, and the screw maker’s withdrawal capacity for the timber species and embedment will normally exceed it by a wide margin. The weak link is more often the kerb itself: a kerb built from offcuts and skew-nailed to the rafters may pull away as a whole. RL-24’s kerb is fixed to the trimmers and rafters with structural screws and brackets designed for the full uplift, because once the roof light is on its kerb, the kerb is the only thing holding it to the building.

Edges and corners

Near the edges and corners of a roof, uplift pressures are higher. If RL-24 were near a corner of the roof where, say, the pressure was 25% higher, the uplift would rise to about 4.32 kN, and the fixings and kerb would be checked again. Designers are sometimes tempted to place a roof light near the edge of a roof to light a room’s perimeter; doing so is possible, but it raises the wind load, and the fixing design must follow.

A failure case: screwed tight

This is an illustration. On an earlier flat-topped roof light with a 3,000 mm aluminium frame in a dark bronze finish, the installer screwed the frame to its timber kerb at 200 mm centres with plain round holes, tightening each screw hard to compress the sealing tape under the frame.

In its first summer, the owners heard ticking from the roof light on hot afternoons, and by autumn the sealant at two of the frame’s mitred corners had split. The first heavy rain of winter found the splits. The detection came from the ticking, which the owners mentioned in passing, and from the split corners seen on inspection. A calculation showed that the 3,000 mm frame was trying to move about 23 × 10⁻⁶ × 3,000 × 70 ≈ 4.8 mm over its seasonal range, and the round holes allowed none of it.

The correction was to lift the frame, redrill the holes as slots away from a fixed midpoint, refit with sealing washers tightened to snug and reseal the corners with a joint that could accommodate movement. The shop’s fixing schedule now shows round holes at one point on each side and slots everywhere else, with slot lengths calculated from the frame’s length.

How this applies to a small Australian business

Show the owner both the summer gain and the winter loss, and compare them with the roof the roof light replaces. Choose the glass and any shading for the whole assembly’s performance. Calculate thermal movement for every long member and allow for it: edge clearance between glass and bars, slotted fixings with a fixed point on each frame side, and corner joints that move with the frame. Get the uplift from the loading standard for the roof light’s actual position, size the fixings and, above all, the kerb’s own fixing to the building, for it.

Exercise: a longer frame

A roof light for a commercial building has an aluminium frame 3,600 mm long, fixed at its midpoint with slots elsewhere. The uplift pressure at its position is 1.5 kPa, its plan area is 3.6 × 1.2 = 4.32 m², and its self-weight is 1.05 kN. Using this chapter’s teaching values, how far does each end of the long side move over a 70 K range? How many fixings at 300 mm centres are there around its perimeter, and what does each carry in net uplift?

Worked answer. The long side grows by 23 × 10⁻⁶ × 3,600 × 70 ≈ 5.80 mm, so each end moves about 2.90 mm from the fixed midpoint, and slots must allow at least that each way plus a margin. The uplift is 1.5 × 4.32 = 6.48 kN, and the net uplift after self-weight is 6.48 − 1.05 = 5.43 kN. The perimeter is 2 × (3,600 + 1,200) = 9,600 mm, giving 32 fixings at 300 mm centres, each carrying about 5,430 ÷ 32 ≈ 170 N. The loading standard’s factors and the screw maker’s capacity then decide whether 300 mm centres are close enough.