Insulating a Sloping Ceiling in a Room in the Roof

The board in your hand is the right board for the wrong plane

Pull one panel off a sloping ceiling in an old room-in-the-roof and put a tape in the rafter. A nominal 100 measures 97, sometimes 95, and on a roof built before regularised sawing it will change by five millimetres between one bay and the next. The 100 mm sheet of foil-faced board leaning against the dwarf wall was bought to fill that rafter, and it cannot, because fifty of those ninety-seven are already spoken for by moving air. That sheet has a job on this roof. Its job is under the rafters, not between them.

A sloping ceiling is the tightest thermal envelope in a house because both of its faces are pinned. Above it are battens, underlay and a covering nobody is lifting unless the roof is coming off anyway. Below it is a finished line that decides whether the room is legal to stand up in, and where the slope dies into the dwarf wall. Unlike a loft deck, depth cannot be bought here by piling more on: every millimetre added on the warm side is a millimetre taken permanently out of the room, and the person taking it is standing in the room while they do it.

So the job is a subtraction before it is an insulation job - written down in millimetres down the rake before a single sheet is ordered. What the air path takes, what the rafter leaves, what the continuous layer has to make up, what the finish and the services add back. Get that column of figures on paper and every product decision after it becomes obvious. Skip it, buy the thickness that was on offer, and the first thing to go wrong is the thing nobody can see until a winter has been through it.

The air path is spent before anything else gets any

In a cold roof with a ceiling that follows the pitch, the ventilation gap is not slack in the detail. It is a designed cross-section running from an inlet at the eave, up the full length of every rafter bay, to an outlet at high level, and it does one thing: it carries away moisture that got past the ceiling before that moisture reaches a cold underlay and turns into water. It is the first claim on the rafter depth and it does not trade against a better U-value, because the two failures are not comparable. An under-insulated slope costs money every winter. A blocked one rots the rafter feet and stains the ceiling, and by the time it shows, the plasterboard is on.

What sets the figure is the document adopted at the address, not habit. In the UK, BS 5250, Management of moisture in buildings - Code of practice, is what a building control officer will be working from for a roof where the ceiling follows the pitch; 50 mm free above the insulation, with a continuous inlet at the eave, is the detail most UK roofers have in their hands, and the current edition is where to confirm it. In North America the International Residential Code, Section R806, Roof Ventilation, requires a clear space between the insulation and the roof sheathing in a vented rafter assembly, with a minimum of one inch, and pairs it with a required net free ventilating area for the space as a whole and a split between the low- and high-level openings. Different numbers, same subtraction: whatever the clear dimension is, it comes off the rafter first.

A gap only exists if something physically holds it open. Rigid chutes stapled to the rafter sides, running from behind the eave inlet up past where the insulation will stop, are the whole of that job, and they have to go in before any insulation, because once a bay is full there is no way to slide one up it. Two details decide whether they survive: they need fixing to the rafter sides rather than to the underlay, and they need to start low enough that the eave inlet is behind them rather than under the first sheet of board. And the path needs somewhere to go at the top. A 50 mm channel that arrives at a sealed ridge, or dies into a collar with the flat ceiling insulated across it, is a stripe of cold air with no exit and it will condense at its top end.

Run the target backwards before choosing a thickness

Consent is granted against a U-value. Merchants sell millimetres. Convert once, at the very start, and carry a resistance figure around for the rest of the job instead of a transmittance. The unit trap on that conversion is worth stating flatly, because it is a factor of nearly six rather than a rounding error: a metric U-value in W/m²K and an imperial one in BTU/hr·ft²·°F differ by about 5.678. A target of 0.18 W/m²K is 0.0317 in imperial units, whose reciprocal is about R-31.5 hr·ft²·°F/BTU, the same assembly as 5.56 m²K/W. Divide the metric figure by 5.678 before it goes anywhere near an imperial calculator, or the answer comes back roughly five times too flattering and looks entirely plausible on the way past.

Which target applies is a question with a local answer. In England, Approved Document L, Conservation of fuel and power, Volume 1: Dwellings, sets the standards, and a room in the roof lands in a different row depending on whether the element counts as new, as a change of use, or as a retained element being upgraded, with the edition in force at the date of application deciding all of it. In the United States the International Energy Conservation Code and the IRC set it by climate zone. Take the number off the paperwork for this address, and treat any figure quoted in an article - the 0.18 used as arithmetic above included - as an example rather than as your requirement.

Writing the depth budget down

With the air path fixed and the target expressed as resistance, the rest is a column of subtractions, written in the order the layers sit down the rake so that the continuous layer stops being the afterthought that gets thinner every time the headroom argument comes up. Two figures feed it and both belong on a document rather than in anyone's memory. The first is the declared thermal conductivity of whatever board or wool is going in: rigid polyisocyanurate is commonly declared around 0.022 W/m·K, mineral wool between roughly 0.032 and 0.044 depending on density and product line, and the value that counts is the one on that product's certificate at that thickness. The second is the rafter, which is timber and conducts like timber - softwood at about 0.13 W/m·K puts a 100 mm rafter near 0.77 m²K/W, or about R-4.4, almost exactly what a 2x4 stud is worth in a wall.

Run those through the assembly and the shape of the problem appears immediately. Fifty millimetres of board at 0.022 is about 2.27 m²K/W. Against a 5.56 target that leaves roughly 3.3 still to find, which is another 72 mm of the same board, and it has to be continuous under the rafters because there is nowhere else for it. Add plasterboard, and a service batten if the specification has one, and the finished ceiling has come down somewhere between 85 and 110 mm from the underside of the rafter. That is the real cost of the job and it belongs in the conversation with the owner before the order is placed, not after the first ceiling is up and somebody notices the door head.

One arithmetic caution when totalling layers: internal and external surface resistances are part of the assembly and are worth having, but they are worth having once. The calculator below adds a combined air film allowance of R-0.85 - about 0.15 m²K/W - on top of the layers entered, so put materials in its three layer fields and let it supply the films rather than spending a field on them. Those three fields are imperial R and do not follow the metric toggle, which makes them the second place the 5.678 factor bites: every metric resistance above goes through it on the way in, so 2.27 m²K/W is typed as 12.9 and a 5.56 target is R-31.5.

Cutting board into a roof that was never square

Between-rafter board is measured bay by bay and cut bay by bay. Sawn rafters were never a parallel set, they have twisted since, and a roof carrying a purlin or a valley has bays that change width over their own length. Cut one template and repeat it and the slope fills with two-millimetre stripes that are, thermally, slots. Measure top and bottom of each bay, cut two or three millimetres oversize for a friction fit, and seal every piece's perimeter with foam or a compatible tape as it goes in - in a roof this thin the perimeter seal does as much work as the board face.

Ordering surprises people, because this layer is bought as full sheets and used as strips. A sheet ripped into bay-width pieces leaves a last rip that fits nothing, so the waste allowance is meaningfully higher than on a flat wall, and it climbs again on a roof with dormers, a valley or a roof window where a high proportion of pieces are one-offs. Count in sheets, then check the count against the bay schedule rather than against slope area. Rigid polyurethane and polyisocyanurate boards are specified under BS EN 13165, Thermal insulation products for buildings - Factory made rigid polyurethane foam (PU) products - Specification, and ASTM C1289, Standard Specification for Faced Rigid Cellular Polyisocyanurate Thermal Insulation Board; the facer named on that certificate matters twice, once for conductivity and again for vapour resistance.

Foam plastic in an occupied room needs its thermal barrier, and here the plasterboard is that barrier. A slope insulated and left unlined over a weekend is a genuinely different fire risk from one insulated and boarded the same day, which is worth saying out loud to whoever is sleeping in the house during the works. Confirm the lining against the board manufacturer's certificate, which names the thickness and the lining it was tested behind.

What the timber takes back

Nothing installed between the rafters improves the rafters. At 400 centres a 47 mm rafter is a little under twelve per cent of the ceiling area; at 600 centres, closer to eight. Those are the clean numbers, and a conversion is rarely clean: add trimmers around a roof window, valley timbers, a dormer's cheeks, the purlin and its struts, the collar and the doubled rafters either side of an opening, and the timber fraction climbs back towards what a stud wall carries. Count the timber on the actual slope, from the survey.

Put representative figures through the parallel-path arithmetic and the size of it stops being arguable. Fifty millimetres of board between the rafters is about R-13 in imperial terms, the 100 mm rafter beside it about R-4.4, and at twelve per cent framing the two together come out near R-10.5 rather than R-13. Close to a fifth of the between-rafter layer is handed back by the timber, and it is handed back regardless of which board went in the bay - which is the argument for the continuous layer stated as a number instead of as a principle. Parallel path is a simplification and says so; the version to defend a submission with is the combined method in BS EN ISO 6946, Building components and building elements - Thermal resistance and thermal transmittance - Calculation methods, with junction losses handled under BS EN ISO 10211.

One practical note on the calculator below: its framing fraction floors at 0.10, and a disciplined 600-centre roof with nothing in it genuinely sits under that. Check whether the slope really is that clean before taking the benefit - a single valley usually spends it.

Where wool is the better answer

Rigid board wins the between-rafter contest on conductivity and loses it on tolerance. Mineral wool will take up a bay that changes width over its length, close around a twisted rafter, and pack into the awkward triangle where a valley board meets a common rafter, all of which are places a rigid sheet leaves a gap that then has to be foamed. On a hundred-year-old sawn roof there are usually several such bays per slope, and the sensible answer is not to pick one product for the whole job.

It also has the better argument on the parts of the room that are not the slope. The dwarf wall behind the eaves, the flat ceiling between the collars, a dormer's cheeks and the studwork around the stair are standard framed cavities at standard centres, and they are where the room's fire lining and acoustic separation are being formed at the same time. Batts specified under BS EN 13162, Thermal insulation products for buildings - Factory made mineral wool (MW) products - Specification, or ASTM C665 in North America, and installed to the practice in ASTM C1320, cover those elements in one pass. Friction fit them, split them around cables rather than compressing over them, and never squeeze a batt into a narrower bay to save a cut.

The trade-off is depth, and depth is this roof's currency. Wool at 0.037 needs roughly two-thirds more thickness than board at 0.022 for the same resistance, so with 50 mm left after the air path it usually loses the slope and wins everywhere else. Order it against the coverage printed on the pack at the thickness actually being used, since coverage per package moves with both thickness and product line.

Which side of the insulation the vapour control belongs on

A room in the roof generates more moisture than the space it replaced and has less ventilation to deal with it. It is usually a bedroom, frequently with an en-suite, sitting at the top of a stair that carries warm humid air up from the whole house, with windows on one plane only. All of that arrives at a ceiling made of thin insulation with cold timber running through it, so a vapour control layer here is doing a harder job than the same membrane on a wall.

The placement rule is what the calculator below tests: keep the control layer warm enough that its own surface never falls to the dew point of the air arriving at it, which means most of the assembly's resistance sits outboard of it and only a small share inboard. Push it too far out - above the between-rafter board, say, or by relying on the foil facer of a board that has another insulating layer inside it - and the membrane becomes the coldest surface in the assembly with warm wet air pressed against it. The method is set out in BS EN ISO 13788, Hygrothermal performance of building components and building elements - Internal surface temperature to avoid critical surface humidity and interstitial condensation - Calculation methods; the psychrometrics behind the dew point are standard ASHRAE Handbook - Fundamentals material.

Two site conditions dominate the outcome and neither is on the drawing. The first is what is already sitting on the cold side. A roof with an old bitumen felt underlay has a vapour barrier outboard of the insulation, which is exactly why that roof must stay ventilated and why its air path cannot be traded for thickness. A vapour-permeable underlay changes the argument, but only where its own certificate says so and only where the eaves detail matches what was tested.

The second is the service zone. Every downlight, cable and shower feed that crosses a vapour layer is a hole in it, and a battened void beneath the membrane is what stops the electrician undoing a week of careful work. BS 5250 and BRE Report BR 262, Thermal insulation: avoiding risks, are the two documents to have open while deciding both of these.

Foil facers deserve their own line. A foil-faced board is already a high-resistance vapour layer, so an assembly with foil-faced board between the rafters, foil-faced laminate under them and a polythene sheet under that has three vapour checks with cold voids trapped between them. Pick which layer is the control layer, make that one continuous and sealed at every perimeter and penetration, and let the others be what they are.

The bays that are not the standard bay

The straight run of a slope is the easy part of the take-off. What generates the callbacks is everything the run is interrupted by - trimmers around a roof window, the cheeks and front plate of a dormer, the valley where a dormer or an extension meets the main roof - each of which creates bays narrower than the standard, bays blind at one end, and bays whose ventilation path is cut off partway up. So measure the centres rather than assuming them: a sawn roof drifts, an extension is often framed to a different spacing than the original, and a re-covered roof may have had rafters sistered. Take the run in sections between the interruptions and count each section separately, because a single division across the whole eave assumes a continuity this roof does not have. The two baffle counters on this site answer that division slightly differently - one takes the run over the spacing, the other adds one for the bay that closes the run. On a straight uninterrupted eave the first is fine; where a section is terminated at both ends by a cheek or a valley the closing bay is real, and the higher figure is the one to buy against.

Then handle the paths that are blocked rather than narrow. A bay that runs into a valley board has no route to the ridge and needs its own outlet, usually a tile or slate vent above the obstruction. A dormer roof is a small separate roof with its own inlet and outlet requirement. Neither is solved by ordering more chutes, and both are cheaper to solve while the slope is open than after it has been skimmed.