A contractor I work with in Doha took over a fit-out where the previous ductwork crew had been paid on square metres and had sealed nothing. The system was a 1200 Pa low velocity supply for a hotel podium, around 3400 square metres of galvanised rectangular duct. The commissioning agent connected a test rig to the first 30 metre run and the fan could not get the section above 780 Pa. The leakage was so high that the flow meter read almost nothing at the nozzle. That is not a sealing problem, that is a ductwork system that never had a chance.
What followed was six weeks of going back over every joint with mastic and a brush, in a plant room with the services already installed. The resealing cost more than the original sheet metal work. I tell that story to every customer who asks me whether the sealant stage on an auto line is worth automating, because it is the cheapest insurance in the whole fabrication process and it is the first thing shops cut when they are busy.
Duct leakage is not spread evenly over the surface. On a typical rectangular system, the distribution looks roughly like this: 35 to 45 percent at the transverse joints, 20 to 30 percent at the longitudinal seams, 10 to 15 percent at the corner inserts on TDF and TDC systems, and the rest through the duct wall itself, mainly at fastener holes, access doors and any place where the sheet has been creased hard enough to open a pinhole.
That distribution matters because it tells you where to spend money. Painting mastic over the middle of a duct panel does nothing. The money goes into the transverse joint and the corner, and the corner is where almost every shop gets it wrong, because the corner is where the gasket ends.
On a TDF or TDC joint, the gasket runs around the flange and stops at the corner insert. The insert is a separate moulded piece that is supposed to bridge the gap. In practice the insert sits in a pocket that is a few tenths of a millimetre oversize, and that gap is a leak path. On a 500 Pa system you can lose 15 percent of the joint's sealing performance at the four corners alone.
The fix is not more gasket. The fix is a bead of sealant applied to the insert before it is pressed home, and a check that the insert bottoms in the pocket. I ask fitters to fit the insert, pull it out, and look at the witness mark. If the mark only covers half the insert, the pocket is shallow or the insert is the wrong size for the profile, and no amount of mastic will compensate.
A Pittsburgh lock is a folded seam, and a folded seam is not airtight. The fold traps a thin continuous capillary path along its whole length. On a 1.5 m duct that is 1.5 m of leak path per seam, and a rectangular duct has four of them. The usual fix is a sealant bead run into the pocket before the seam is closed, which is why higher-end lock formers have a sealant injection port. The cheaper fix, and the one most shops use, is to run a brush bead over the outside of the seam after forming.
Neither is as good as a seam that is mechanically tight. A Pittsburgh lock with the correct roll gap has a seam that is closed to within 0.1 mm along its length, and leakage through it is small. The same lock with an open roll gap has a visible gap you can see light through, and it will leak no matter what you paint on it. If your leakage numbers are bad and your mastic consumption is high, check the lock former before you buy more sealant.
There are five methods in common use. Each has a place, and the mistake I see most often is using the cheapest one everywhere.
| Method | Typical cost per metre of joint | Leakage class achievable | Labour time | Best used on | Main weakness |
|---|---|---|---|---|---|
| Solvent-based mastic, brush applied | Low | Class B to C | 15 - 20 s per joint | Transverse joints, site touch-up | Shrinks and cracks after 3 - 5 years; solvent smell delays handover |
| Water-based mastic | Low | Class B | 20 - 25 s per joint | Interior duct, occupied buildings | Poor adhesion to oily or zinc-heavy surface; needs 24 h cure |
| Pressure-sensitive tape (foil or fabric) | Medium | Class B | 10 - 15 s per joint | Accessible straight runs | Fails at 60 C or above; adhesion drops on dusty galvanised |
| Pre-formed gasket in flange groove | Medium (material) | Class A with good corners | Near zero at assembly | TDF, TDC and slide-on flange systems | Gasket compression set after repeated dismantling; groove depth critical |
| Sealant injected into lock pocket | Medium | Class A on seams | Zero extra (in-line) | Longitudinal seams on automated lines | Needs a sealant pump and clean port; blocked ports are common |
The cost numbers are indicative and depend heavily on local labour rates. In a market where a fitter costs 4 USD an hour, tape and mastic beat everything. In a market where the same fitter costs 25 USD an hour, the in-line sealant injection pays back within one mid-size project, because it removes a whole manual operation.
Most international specifications reference a leakage class at a stated test pressure. The classes are usually expressed as allowable leakage in litres per second per square metre of duct surface at a reference pressure, and the three that come up in commercial work are roughly: tight systems at 0.4 to 1.0, medium at 2 to 6, and loose at 8 to 12. The exact numbers vary with the standard being cited — SMACNA, EN 12237, DW/144 and AS 4254 all define the bands slightly differently — so the first thing I do on a job is read which standard the specification actually names, because the difference between them can be a factor of two on the same duct.
What matters on the shop floor is not the class number but the test pressure. A system tested at 400 Pa will pass with gaskets and care. The same system tested at 1000 Pa will find every corner that was assembled with a hammer. Above 750 Pa I insist on mastic at every transverse joint in addition to the gasket, regardless of what class the specification asks for, because the gasket alone has too many variables.
The light test — a lamp inside the duct in a dark room — catches open seams and missing sealant, and nothing else. It is fast and it is worth doing on every run before the duct leaves the shop. The pressure test with a calibrated fan and nozzle catches everything but takes time and needs the section capped. The smoke test is useful for finding a leak you know exists but cannot locate.
My routine on a new shop is: light test every section on the bench, then pressure test one section in twenty at the specified pressure. If the sampled sections pass with margin, the batch is probably fine. If one fails, the batch gets tested section by section until the cause is found, because a leakage failure is almost never a one-off — it is a setup error that repeats.
There is a strong argument for doing as much sealing as possible in the shop. Site access is worse, the duct is already installed, the sealant has to cure in a building with no ventilation, and the labour rate is usually higher. Anything you can do at the bench costs less.
That is why I recommend shops buying a new line specify the sealant injection option on the lock former, and a flange machine with a gasket groove cut to the gasket they actually intend to use. It is also why a duct beading machine earns its place: a bead rolled across the panel at 300 to 400 mm centres stiffens the sheet, which reduces oil-canning, and a panel that does not flex does not crack its sealant. Beading does not seal anything directly, but it stops the seal from being destroyed by movement. On 1.0 mm panels over 800 mm wide I would not run a job without it.
The same logic applies to the seam locker. A hydraulic seam locker closes the Pittsburgh seam with a controlled force along its full length, which produces a much more consistent seam than a hand tool, and consistency is what leakage performance actually is. You can see the difference on a light test: hand-closed seams show a dotted line of light, machine-closed seams show nothing.
If you are the buyer and you want a system that passes, put four things in the specification: the leakage class and the standard it is measured against, the test pressure, the permissible sealing methods by joint type, and the requirement that corner inserts be sealed. That last one is the item everyone forgets, and it is the one that shows up in the commissioning report.
If you are the fabricator, price the sealing honestly. A tender that assumes 20 seconds per joint and actually needs 45 is a tender that loses money on every duct. The hydraulic Pittsburgh seam locker and an in-line sealant system are the two pieces of equipment that most directly reduce that time, and they are usually cheaper than the labour they replace within a year on any shop above 1500 square metres a month.
At the same pressure and the same corner quality, they are close. TDF has a slightly larger gasket contact area and a deeper groove, which helps at high pressure. TDC has a more rigid flange, which helps resist distortion when the duct is large. The difference in leakage is smaller than the difference made by how well the corners are fitted. I have seen TDF systems pass Class A and TDC systems fail on the same project, built by the same crew, because of corner assembly.
You can, and sometimes you have to, but the cost is three to five times the shop cost. Access is the limiting factor. Above a suspended ceiling with services already in, a fitter may only be able to reach two sides of a joint, and the two sides he cannot reach are usually the ones leaking.
For a rectangular duct with transverse joints every 1.5 m, expect roughly 0.9 to 1.2 metres of sealant bead per square metre of duct surface, which works out at about 0.15 to 0.2 kg of mastic per square metre. If your consumption is double that, you are either over-applying or your joints are too far apart. If it is half, check the corners.
Rarely. Pressure-sensitive tape depends on sustained contact pressure, and at 1000 Pa the duct panels flex, which peels the tape edge. Above 750 Pa I would use mastic on every joint and treat tape as a temporary repair only. Below 500 Pa, on an accessible straight run, good foil tape is a reasonable choice.
Whatever method you choose, the numbers that decide whether a system passes are set long before the sealant is opened: the lock gap on the former, the flange groove depth, the corner pocket depth, and the flatness of the panel. Those are machine settings. Get them right and sealing becomes a ten second operation. Get them wrong and you will be chasing leaks for the life of the building. The auto duct line 5 configuration is the one most of my customers specify when leakage class is written into the contract, mainly because the line's notching and beading stations let them build the stiffness in before the duct is formed.