Last spring I stood in a food plant near Izmir while two contractors argued about a duct system that had already been insulated. The specification said leakage class C. The test rig said the system was leaking at roughly 2.4 times the class C limit, which is not a rounding error, it is a different class altogether. The installer's defence was that the duct had passed a factory test on the blanks. It had. What it had not passed was a test with 96 transverse joints, 40 branch take-offs and about 1,200 metres of mastic bead applied by three different men over five days.
That is the gap this article is about. A duct leakage test does not measure how well your machine forms a flange. It measures how well a crew sealed a system, and how well the designer left them room to do it. I have run or witnessed a few hundred of these tests now, on rectangular galvanised systems from 120 to 2,400 square metres of surface, and the pattern is depressingly consistent: the machine work is rarely the problem, the joints almost always are.
If you are buying equipment and want to know what a line can hold before anyone touches a tube of mastic, that is a separate question, and it starts with the flange system and the seam quality. A TDF flange forming machine that holds flange height within half a millimetre gives the crew a pocket that a corner insert can actually seat into, and a Pittsburgh lock former with correct roll clearance gives them a seam that does not whistle. But once the duct leaves the shop, the test is a joint test.
Most international tenders quote either EN 12237 (rectangular ductwork) or a SMACNA leakage class, and occasionally both, which is where the arguments begin. They are not the same scale and they do not convert by a single factor at every pressure.
EN 12237 classifies tightness as a leakage rate per square metre of duct surface, related to test pressure by a power law:
f = k × p0.65
where f is in 10-3 m³/(s·m²), p is the test pressure in pascals, and k is the class constant: 0.027 for class A, 0.009 for B, 0.003 for C and 0.001 for D. Classification pressure is 400 Pa. Run the numbers at 400 Pa and class A allows about 4.8 m³ per hour per square metre of duct surface, class B about 1.6, class C about 0.53 and class D about 0.18.
That spread matters commercially. Class C is roughly nine times tighter than class A, and in my experience class A is what an unsealed but neatly built system achieves, class B is what you get with mastic on transverse joints, and class C needs mastic plus attention to branch taps and damper penetrations. Class D is a specialist job with sealant inside the joint, gasketed access doors and a crew that has done it before.
SMACNA expresses leakage as cfm per 100 square feet of duct surface at 1 inch water gauge, in classes 3, 6, 12, 24 and 48. Class 3 at 249 Pa works out to about 0.55 m³/(h·m²). Correct that to 400 Pa with the same 0.65 exponent and you land near 0.75, which puts SMACNA class 3 slightly looser than EN class C and considerably tighter than class B. If a consultant writes "SMACNA class 6 or better, tested to EN 12237", ask which scale the pass mark is on. I have seen a system accepted at 4.1 m³/(h·m²) that should have been rejected at 0.53.
| Class | Limit at 400 Pa, m³/(h·m²) | Allowance for a 380 m² system, m³/h | What usually achieves it |
|---|---|---|---|
| EN A | 4.8 | 1,824 | Neat Pittsburgh seams, joints pushed home, no sealant |
| SMACNA 3 (≈) | 0.75 | 285 | Mastic on transverse joints, corners seated |
| EN B | 1.6 | 608 | Mastic on transverse joints, some tap sealing |
| EN C | 0.53 | 201 | Mastic plus sealed taps, gasketed access, tidy corner inserts |
| EN D | 0.18 | 68 | Sealant inside the joint, factory gaskets, supervised crew |
One more thing before you buy a gauge: the test pressure is not the design pressure. EN 1507 asks for 1.5 times the design static pressure, capped for the classification test. A system designed for 500 Pa gets tested at 750 Pa, and leakage rises faster than pressure, so the pass mark at 750 Pa is not the number printed in the class table. Recalculate with the exponent, or you will fail a system that is perfectly fine.
The most common field mistake is a blower that cannot deliver the allowable leakage at the test pressure, which means you cannot even measure a pass. Size from the limit, not from the duct volume.
Take the 380 m² system at class C. The allowance is 201 m³/h at 400 Pa. The blower needs to move at least three times that at the test pressure so that the flow meter sits in a readable part of its range and so you can find the leak if the reading is double the limit. A 750 to 1,000 m³/h centrifugal blower at 800 Pa is the right size, with a variable speed drive or a bypass damper. For small systems under 100 m², a 400 m³/h ring blower is plenty and much easier to carry up a ladder.
Do not use a compressor and a needle valve. I have watched it done, and the flow is unsteady, the air is hot, and the reading depends on how far someone opened a ball valve.
A digital differential manometer with 0 to 2,000 Pa range and 1 Pa resolution covers almost every job. A U-tube water gauge reads in 10 Pa steps, which is 2.5 percent of 400 Pa, and the leakage error scales with the pressure error, so a 10 Pa misread is roughly 1.6 percent on flow. That is acceptable on a big system and fatal on a small one where the limit is 60 m³/h.
For flow, a thermal mass meter or a calibrated orifice plate with a known discharge coefficient. Rotameters are cheap and drift when the air is dirty. Whatever you use, calibrate it against something traceable once a year; I have found two field rotameters reading 15 percent low in the same season, both owned by contractors who were failing tests they should have passed.
If the rig leaks, you are measuring the rig. Everything here is about eliminating that possibility before the reading means anything.
Use 1.2 mm galvanised sheet caps with a folded 25 mm edge, taped and mastic-sealed on the inside. Plywood is tempting and it warps under 800 Pa; I have seen a 19 mm ply cap bow 6 mm and start leaking around the screws. On large openings, brace the cap with a 40 × 40 angle across the middle.
Every open end must be capped, including the ones the contractor "was going to connect tomorrow". A single uncapped 400 × 400 branch at 400 Pa passes several hundred cubic metres an hour, and it will make a good system look like a disaster.
One static tap per test section, at least two duct diameters or 100 mm downstream of the cap, on a flat face away from a seam. Drill 6 mm, deburr both sides, and seal the hose with a rubber boot, not tape. If the section is long and has a damper, tap both sides of the damper and test with it open, or better, remove the blade and blank the shaft.
Run the blower and watch the gauge for five minutes before writing anything down. A reading that is still climbing after five minutes usually means a flexible connection is inflating or a cap is creeping. A reading that oscillates by more than 5 Pa means the blower is hunting and you need to damp it.
After enough tests you stop looking at the duct and start looking at six specific places.
Transverse joints. Four corners and two flange pockets per joint. An unsealed TDF or TDC corner at 400 Pa passes somewhere between 0.5 and 4 m³/h depending on how well the insert seated, and a joint where the pocket has been over-formed passes more. On a 96-joint system that is 384 corners; at 2 m³/h each, 768 m³/h of leakage, which is nearly four times the class C allowance for the whole system. This single item explains most failures I have seen.
Longitudinal seams. A Pittsburgh lock is reasonably tight when the roll gap is right, and a snap lock on light gauge is not. On 0.6 mm duct the seam is usually a few percent of total leakage; on 0.5 mm spiral or on a lock formed with worn rolls, it can be 20 percent.
Branch take-offs. The joint between a spin-in collar and the duct wall, especially where the hole was cut with a nibbler and never deburred. Sealant will not bridge a 3 mm gap reliably; a gasket collar will.
Damper shafts and access doors. A 12 mm shaft through a 14 mm hole leaks more than people believe, and an un-gasketed access panel is a guaranteed fail at class C.
The last 300 mm of each run. Installers work away from the ladder, not towards it. The bead that stops short of the joint is so common that I now check it before I connect the blower.
Flexible connectors and canvas boots. Clamped, not sealed, and rarely tight. On a class C job these should be isolated with blanks during the test, then tested separately.
Log the ambient temperature, the barometric pressure and the duct surface area before you start. Leakage is a volumetric flow at actual conditions, and if the consultant is checking against a limit derived at 20 °C while your duct is at 38 °C in a plant room, the difference is around 3 percent. Not fatal on a comfortable pass, decisive on a marginal one.
Hold the test pressure for five minutes minimum, and take three readings a minute apart at the end. Record all three. A test that is reported as a single number with no time stamp is the kind of document that gets rewritten during a dispute.
Then walk the duct with a smoke pencil or, cheaper, a squirt bottle of soapy water on the low-pressure side. On a positive pressure test, air leaves the duct and you will not hear it; a soap film on the joint shows it in seconds. This is also how you build the punch list that makes the retest pass in one go instead of three.
The size of the failure tells you what to look for, which saves a lot of walking.
Between 1.2 and 2.5 times the limit, it is almost always unsealed transverse joints, with a contribution from taps. Between 3 and 6 times, expect an uncapped branch, an open access door, or a whole run of joints where the mastic was applied to the outside of the joint instead of the inside of the pocket. Above 10 times, stop looking at the duct and check the rig, because a system with that much leakage would be audible from the doorway.
One more diagnostic: plot leakage against pressure at four points, say 200, 400, 600 and 800 Pa, and fit the exponent. A real duct system sits near 0.65. An exponent near 1.0 means you are measuring flow through an orifice that is not the duct, usually the rig or a wide-open damper. An exponent near 0.5 means something is deforming as pressure rises, and on light gauge duct with 2.5 metre joint spacing that is a stiffness problem, not a sealing problem.
Every report I have seen accepted without argument contains six things: a sketch of the test section with the surface area calculation, the tap locations, the instrument serial numbers and calibration dates, the pressure and flow readings with time stamps, the ambient conditions, and a calculation showing the measured leakage against the class limit at the test pressure. Photographs of the caps help more than people expect, because the argument is usually about what was included, not about the number.
If you are the buyer and you are specifying this on a new plant, put the test in the contract with the class, the pressure and the retest cost borne by the installer. The cost of a retest is not the blower hire; it is the insulation that has to come off, and on a large job that is ten times the test fee.
Can I test a system before the insulation goes on? You should, and it is the only sensible sequence. Once the duct is wrapped, a leaking joint is hidden and you are testing the vapour barrier as much as the duct. Test bare, fix, retest, then insulate. On chilled water and low-temperature systems, test again after insulation if the spec asks, but expect the numbers to be similar.
Does a good auto duct line guarantee a good leakage result? No, and it is worth being clear about this with clients. A line with properly set TDF rolls and sharp corner inserts gives the crew a joint that can be sealed. Whether it is sealed is a labour question. What the machine does control is flange height consistency, corner pocket fit and seam tightness, and those three set the floor of what is achievable. If you are specifying a new line for a market where class C is written into every tender, put the flange and seam tolerances in the equipment order, and read the specification list on an auto duct line 5 against the job you actually intend to run.
How much sealant does a class C system use? On 1,000 mm duct with 1.5 metre joints, expect roughly 1.2 to 1.8 metres of bead per joint, which on a 380 m² system of about 96 joints is 45 to 60 tubes of 310 ml mastic. Contractors who budget 20 tubes always run short in the last third of the building, which is exactly where I find the leaks.
What is a realistic leakage figure for a shop-built system? With TDF flanges, seated corner inserts, mastic on transverse joints and sealed taps, I expect 0.6 to 1.0 m³/(h·m²) at 400 Pa on a first test. That passes EN class B and sometimes C. Below 0.5 needs a second pass with sealant inside the joints, and it is worth doing only where the spec genuinely demands it.