The Air Permeability Test on a Building You Have Nearly Finished

The divisor is the envelope, and the envelope includes the floor

Almost everyone arrives at this test carrying a number in air changes, because that is the currency a house gets talked about in and the one the North American protocols and the certification standards report against: flow divided by the volume inside. A non-dwelling compliance result is not that. It is normalised by envelope area, and that single difference reorganises the whole exercise. Air permeability, written q50, is the flow needed to hold the building at fifty pascals divided by the total area of the internal surfaces bounding the measured volume — every wall on its room side, the underside of the roof, and the ground floor slab. The floor is the one left out of a first take-off, and on a single-storey shed it is close to a quarter of the total.

Take a plain box: a hundred metres square on plan, twelve metres to the mean internal height. Floor ten thousand, roof ten thousand, walls four hundred metres of perimeter by twelve, which is four thousand eight hundred. Twenty-four thousand eight hundred square metres of envelope, enclosing a hundred and twenty thousand cubic metres. At a permeability of five cubic metres per hour per square metre, the fan has to move a hundred and twenty-four thousand cubic metres an hour — about seventy-three thousand cubic feet a minute, or thirty-four thousand litres a second.

Now put that same flow in the other currency. Over a hundred and twenty thousand cubic metres of volume, it is a shade over one air change per hour at fifty pascals. On a house, one air change at fifty pascals is certification-grade airtightness people build carefully for years to reach. On this shed it is unremarkable, for a purely geometric reason: a large box holds an enormous amount of inside per square metre of skin, so the metric referenced to volume flatters it and the metric referenced to area does not. Anyone arriving at a warehouse with a scale calibrated on dwellings reads every number wrong in the same direction.

Which currency your project is written in follows from the document. In the United Kingdom the measurement is to BS EN ISO 9972 under ATTMA Technical Standard L2, the non-dwellings standard — a tester whose registration covers TSL1 holds a certificate for the wrong building type, and building control will say so. Compliance sits with Approved Document L Volume 2 in England and the devolved equivalents elsewhere, and the limiting value has moved between editions, so read the one in force. In North America the whole-building test route sits in Section C402.5 of the IECC and the envelope provisions of Section 5.4 of ASHRAE 90.1, is expressed in cubic feet per minute per square foot of envelope, and is run at 0.3 inches water gauge — seventy-five pascals, not fifty. Federal work under the US Army Corps of Engineers Air Leakage Test Protocol is tighter again.

Those two reference pressures are the trap. A result at fifty pascals is not a result at seventy-five with a correction factor applied, because flow through an enclosure follows a power law whose exponent is a property of the leaks rather than a constant. Converting between them honestly needs the exponent from the tester's own multi-point fit, and a specification quoting one pressure while the tester works to the other is a conversation to have before the rig is built.

Last, and least glamorous: get the envelope area take-off agreed in writing before the day, with whoever will accept the certificate. The two versions only agree once the boundary is settled — the substation, the pump house, an external plant enclosure, the office block if it is a separate zone. A boundary moved afterwards moves the published result by the same ratio without anybody remeasuring a thing, which is the commonest quiet error on a non-dwelling certificate.

Reading a result that came in over

The headline arrives as a permeability figure against a target, and the first question is not whether you failed but by how much and in what shape. A result a few percent over is a diffuse problem — a general standard of sealing across a large area — and it usually yields to a day's work on the longest joints and a defensible retest. A result forty or sixty percent over is not diffuse. Something is open, probably one or two things, and both will be findable in an hour with the fan still running: an argument for keeping the tester on site rather than for scheduling a meeting.

Convert the flow into the volume currency as a cross-check, because the two say different things and both matter downstream. Permeability tells you about the skin. Air changes at fifty pascals tell you how fast the enclosed volume turns over, which is the figure the ventilation designer, the smoke strategy and anyone modelling heating demand will reach for. A shed and its office block can sit at the same permeability and be an order of magnitude apart in air changes, and a number handed to the wrong discipline in the wrong currency is how a mechanical design gets sized against a fabric figure that never described it.

The useful output buried in the report is the flow exponent from the multi-point fit. It describes the character of the leakage rather than its size: near one half it behaves like large sharp-edged openings, near one like long narrow cracks. That single number tells you whether to send two people looking for a hole or a gang looking for four hundred metres of joint, and it sits on the certificate everybody files without reading past the first line.

The hour when the fan is still running

The most valuable part of the day is not the number. It is the period afterwards with the building held at pressure and somebody walking it with a smoke pencil, a thermal camera or the back of a wet hand. Depressurised, every leak pulls inward, so it is findable from inside — which is where your access, your lighting and your MEWPs already are. Book that hour explicitly in the tester's day and say what it is for, because a tester engaged for a compliance test will reasonably pack the rig down the moment the reading is recorded.

Have the materials on site before the test so the hour turns into a fix rather than a snag list. Profile fillers matched to the actual sheet profile, not the one on the drawing; gun-grade sealant and backer rod; air barrier tape and its primer; and the fixings to hold any of it in place at height. The tape is what most often fails on the day, for reasons that have nothing to do with the tape: acrylic and butyl air barrier tapes want a clean, dry, sound substrate above the minimum application temperature their maker states, and a dusty profiled sheet at eight on a February morning meets none of those. Read the manufacturer's instruction before assuming primer is optional, and warm the surface if the sticking has to happen that morning.

1. Agree with the tester before the day that the building will be held at pressure for leak detection after the reading, and that you will have people and access ready for it.
2. Walk the long linear joints first — eaves, verge, corner flashings, wall to floor — because they are where a diffuse failure lives.
3. Then the discrete holes in descending order of size: shutters, dock pits, penetrations, dry traps, vent upstands.
4. Mark every find on the floor and on a plan at once, since the plan is what survives the fan being switched off.
5. Fix what can be fixed while the rig is standing, and re-check those items under the same pressure before the fans come down.
6. Book the retest before leaving site, and put the remaining items against named trades rather than on a general snag list.

The ductwork is a separate test with separate paperwork

"The air test" on a programme almost always means the envelope, and the mechanical specification almost always also calls for ductwork leakage testing. Different standards, different pressures, different sample proportions, different certificates — and a site manager who has booked one has not booked the other. This gets discovered at handover with impressive regularity.

In the United Kingdom the practical document is BESA DW/143, with the leakage classes defined in DW/144 against the duct's pressure classification. In North America it is the SMACNA HVAC Air Duct Leakage Test Manual, where the allowable rate per hundred square feet of duct surface is the leakage class multiplied by the test pressure raised to the power 0.65. Note the pressure straight away: a duct is tested at its own operating class, in inches of water gauge or in pascals well above the envelope's fifty, so a number from one test cannot be read across to the other even in principle. How much of a system gets tested is set by its class and by the specification, not by a rule of thumb.

The two tests meet at exactly one place. A duct that crosses the envelope — a kitchen extract, a smoke vent duct, an air handling unit connection through a plant room wall — is an envelope leak whenever its damper is open or its casing joints sit on the outside face of the air barrier. A leaky duct entirely inside the envelope does not move the permeability result at all, because the air it loses never leaves the measured volume. Which is why a building can pass one test and fail the other with a single defect properly the cause of neither, and why the ductwork drawings are worth laying over the envelope boundary before either test is booked.

What the finished building does when its own extract runs

Once the certificate exists, the leakage figure behind it becomes useful for something the certificate never mentions. The building has extract systems — welfare, a canteen kitchen, a workshop or a battery charging bay, a mechanical smoke ventilation system that has to run on test — and each pulls the enclosure negative against the leakage you just measured. The tighter the envelope, the more pressure the same flow produces, which is the part that surprises people on a building just congratulated for being tight.

The relationship is the same power law the test is built on, run backwards: a measured flow at a known reference pressure, plus the exponent, gives the pressure the building settles at for any other net flow. Take the shed above, leaking about thirty-four thousand litres a second at fifty pascals. Six thousand litres a second of everyday extract with nothing supplied back puts it a shade over three pascals below outside — barely perceptible, and still straight off the margin a door closer was set with. Run twenty thousand litres a second of smoke extract on commissioning and the same building sits above twenty pascals negative, which across a normal leaf is enough added force to matter against the maximum opening forces in Approved Document M or ICC A117.1.

Treat the result as a screen rather than an answer. It ignores wind and stack effect, both of which can dominate on a tall building on a cold or exposed day; it treats the whole building as one zone, which a shed with an office block behind a closed door plainly is not; and it assumes the makeup air you enter is genuinely interlocked rather than a note asking somebody to prop a door. Where it says you are close to a limit, the answer is a manometer, not a smaller number in a spreadsheet.