Duct leakage and pressure testing field guide for commissioning

Duct air leakage testing, and the air the design already paid for

Duct air leakage testing, sometimes written DALT, pressurizes a sealed length of ductwork and measures how much air escapes through its joints, seams, and connections at a set test pressure. The result is reported in cfm per 100 sq ft of duct surface, and it tells you the one thing the static pressure gauge never will: how much of the air the fan moves is leaving the duct before it ever reaches a register, a diffuser, or a server inlet.

Leakage is wasted air, and wasted air is wasted money. The design sized the fan, the coil, and the duct to deliver a certain airflow to the space. Every cfm that bleeds out of a supply trunk through an unsealed transverse joint is a cfm the building paid to move and condition and never used. On a leaky system you can hit design static at the air handler and still starve the far rooms, because the air went into a ceiling cavity instead of the room.

The number people miss is that leakage scales with pressure and with surface area, not with how the duct looks. A trunk that looks tight can leak badly at every joint, and the only way to know is to pressurize it and measure. You do not eyeball duct tightness. You test it.

What does a duct leakage test prove?

A duct leakage test proves how tight the installed duct actually is, measured as the airflow lost through the assembly at a known pressure, compared against the leakage class the specification called out. It is a pass or fail against a number, not a judgment call.

What it does not prove is whether the system moves design airflow. That is a different test. A duct can pass a leakage test and still deliver the wrong air to the wrong rooms because the balance was never set, and a duct can move plenty of air at the registers while quietly dumping a quarter of it into the plenum. Leakage testing isolates one failure: air escaping the duct envelope. Test and balance handles distribution.

The reason the test exists is that leakage hides. You cannot hear most of it over a running system, you cannot see it, and it does not show up as a single dramatic symptom. It shows up as high fan energy, rooms that never satisfy, and a system that runs at full tilt and still falls short. The test turns the invisible loss into a measured number you can hold a contractor to.

How does the SMACNA leakage class work?

SMACNA ties leakage to pressure with one equation: F equals CL times P to the 0.65 power, where F is the leakage rate in cfm per 100 sq ft of duct surface, P is the test pressure in inches of water column, and CL is the leakage class. The leakage class is the value of F at a test pressure of 1 in wg, so a class 4 duct leaks 4 cfm per 100 sq ft at 1 in wg, and more as you push the pressure up.

The 0.65 exponent is not arbitrary. It comes from the physics of air pushing through the cracks and slots of sheet metal joints, which behave somewhere between a sharp orifice and a long crack. Double the pressure and leakage does not double, it rises by about a factor of 1.57. That is why the test pressure has to be stated with the class. A leakage class means nothing without the pressure it was measured at.

Lower class is tighter. Sealed metal duct built to SMACNA standards commonly lands in the single digits to the mid teens depending on shape and construction, with round duct tighter than rectangular for the same effort. The exact class assigned to a given construction has changed between editions of the manual, so pull the leakage class from the edition the spec references rather than from memory.

F

Leakage rate in cfm per 100 sq ft of duct surface at the test pressure

CL

Leakage class, the value of F at a 1 in wg test pressure; lower is tighter

P

Test pressure in inches of water column, raised to the 0.65 power

Seal class A, B, and C

Seal class tells the installer where to put the sealant, and it is set by the duct's pressure class, not chosen at the bench. SMACNA defines three. Seal Class A seals everything: transverse joints, longitudinal seams, and all penetrations of the duct wall. Seal Class B seals transverse joints and longitudinal seams but not the penetrations. Seal Class C seals transverse joints only.

The higher the operating pressure, the tighter the seal class the construction standard calls for. Lower-pressure duct has historically been allowed a looser class, with the higher-pressure classes pushed to Seal Class A. The trend in recent guidance and in a lot of specs is to require Seal Class A across the board regardless of pressure, because the cost of sealing everything is small next to the cost of chasing leaks later. Read the spec. If it calls Seal Class A, every seam and every penetration gets sealed, full stop.

Seal class and leakage class are related but not the same thing. Seal class is the instruction. Leakage class is the measured result. Sealing to Class A is how you hit a low leakage class, but a sloppy Class A job can still fail the test, which is the whole reason you test instead of trusting the note on the drawing.

How much duct leakage is allowable?

Allowable leakage is set by the specification, expressed as a leakage class, and converted to a hard number in cfm per 100 sq ft at the test pressure through the SMACNA equation. The spec names the class, you compute the allowable F for the test pressure, multiply by the tested surface area, and that total cfm is the line the measured leakage has to stay under.

Worked the other way, the same allowable is often sanity-checked as a percentage of system airflow. Many specs aim for total system leakage on the order of a few percent of the design airflow, with figures around 5 percent common and tighter numbers on critical or high-pressure systems. That percentage is a target the class is chosen to meet, not the acceptance criterion itself. The acceptance criterion is the measured cfm against the computed allowable cfm for the section you tested.

Do not confuse the class with the limit. A leakage class is a rate per 100 sq ft at a reference pressure. The allowable for your actual test is that rate scaled to your test pressure and your surface area. Get the surface area wrong or the pressure wrong and the allowable is wrong, and a failing duct passes on paper.

The test rig: fan, orifice, and manometer

The test apparatus is a calibrated blower, a flow-measuring orifice, a manometer, and a way to seal the section. The blower pushes air into the sealed duct, or pulls it out, until the duct holds the target test pressure. The orifice plate or flow nozzle on the blower measures the airflow the blower is delivering, and at steady state that airflow equals the leakage, because the only place the air is going is out through the cracks.

The manometer does double duty. One reading is the static pressure inside the duct, which you bring up to and hold at the test pressure. The other, across the orifice, gives the flow through the calibrated opening, which converts to cfm off the orifice's calibration curve. Modern rigs read both digitally and compute the leakage for you, but the principle is old: meter the air in, hold the pressure, and the air in equals the air out.

Sealing the section is half the job. Every opening that is not part of the duct, every open branch, every register boot, every test port, gets capped or blanked with a gasketed plate or an inflatable bladder. A single open branch you forgot reads as enormous leakage and sends you chasing a leak that does not exist. Seal the boundary first, verify it, then pressurize.

Which ducts get tested, and at what pressure?

The duct's pressure class sets the test pressure, and the specification sets which ducts get tested at all. Not every system is leakage tested. High-pressure supply duct upstream of the terminal boxes is the usual target, because that is where the pressure is highest and the leakage costs the most. Low-pressure duct downstream of the boxes is often exempted, but the spec decides, not habit.

The test pressure is commonly the duct's static pressure class, the same number the duct was built to. The SMACNA pressure classes run 1/2, 1, 2, 3, 4, 6, and 10 in wg. A duct in the 3 in wg class is typically tested at 3 in wg. Some specs test at the design operating pressure instead, which can be lower than the class. Read which one the spec names, because testing at the wrong pressure changes the allowable through the P to the 0.65 term.

Supply duct is tested at positive pressure, because that is how it runs. Return and exhaust duct is tested at negative pressure, a vacuum, because that is how they run. The leakage physics is identical; the blower just pulls instead of pushes. The convention matters for the test setup and for which side of the joints you can reach to find the leak.

Testing in sections vs the whole system

You test in sections, not all at once, on anything bigger than a small system. A section is a length of duct you can isolate and seal at both ends, sized so the blower can actually bring it up to test pressure and hold it. Try to test too much duct at once and either the blower cannot make pressure or the allowable leakage is so large that a real leak hides inside the tolerance.

Section size is a balance. Bigger sections mean fewer setups and less capping, but a higher total allowable that can mask a single bad joint. Smaller sections find the bad joint but cost time in setup. The practical move is to test by the riser, the main, or the floor, matching the section to what you can seal cleanly and to where you would want to localize a failure anyway.

Account for the surface area of the section you actually tested, including the duct you sealed off at the ends but not the caps and bladders themselves. Get the section's surface area right, because the allowable cfm scales straight off it. Claim more surface area than you tested and you inflate the allowable and pass a leaky duct. That is one of the most common ways a marginal section gets signed off.

How do you find a duct leak?

Once a section fails, you find the leaks with your hand, your ears, and a little theater. With the section under test pressure, run a bare hand slowly along the joints and seams and you feel the jets of escaping air at the bad spots. On a positive test the air blows out at you; on a negative test it pulls in, so a tissue or a wisp of smoke shows the direction.

For the leaks you cannot feel, use smoke or soap. A smoke pencil or a theatrical smoke machine charged into the duct streams out of every leak and makes them visible, which is the fast way to map a long run. Soap solution brushed on a suspect joint bubbles where it leaks, the same trick used on gas and refrigerant lines. Smoke shows you the pattern; soap pins the exact spot.

Leakage concentrates in predictable places. Transverse joints, the connections between duct sections, are the worst offenders, followed by longitudinal seams, then every penetration and connection: takeoffs, access doors, damper shafts, sensor wells, and the boots where the duct meets a register. Check the joints and connections first. The flat field of the duct rarely leaks. The places where two pieces of metal meet are where the air gets out.

Mastic vs tape: sealing the leaks

Seal with mastic, not tape, on anything that has to stay sealed. Duct mastic is a troweled or brushed-on sealant that cures to a flexible solid, bridges the gap at a joint, and moves with the duct as it heats, cools, and pressurizes. It is the seal that lasts, and it is what the SMACNA seal classes assume.

Tape fails over time, and that is not a brand problem, it is a physics problem. The adhesive dries out, the carrier shrinks, and the bond lets go at exactly the transverse joints that flex and pressurize the most. A taped joint that passes the test on the day of construction can be leaking within a few years, which is why so much old duct leaks at every seam despite looking taped and finished. If tape is used at all, use a listed foil tape rated for the duty, and back the critical joints with mastic anyway.

For a leaking joint found during the test, the fix is mastic worked into the gap, reinforced with fabric mesh embedded in the mastic on the wider gaps, then allowed to cure before you retest. Sealing a joint and immediately re-pressurizing tells you nothing if the mastic has not set. Plan the cure time into the schedule, because a rushed retest on wet mastic is a retest you will be doing twice.

Calculating the result

The calculation is three numbers: the surface area of the tested section, the allowable leakage rate from the class and test pressure, and the measured leakage off the orifice. Compute the section's surface area as the duct perimeter times its length, summed over every piece in the section. For rectangular duct that is two times width plus two times height, times length. For round it is pi times diameter times length. Express it in square feet, then divide by 100, because the leakage rate is per 100 sq ft.

Run the allowable next. Take the specified leakage class, plug the test pressure into F equals CL times P to the 0.65, and you get the allowable leakage rate in cfm per 100 sq ft. Multiply by the section's surface area in hundreds of square feet, and that is the total allowable cfm for the section.

Then compare. Measured leakage under the allowable is a pass. Over it is a fail, and you seal and retest. Record all three numbers, not just the verdict, because the next person needs to see the surface area you used and the pressure you tested at to trust the pass. A pass with no surface area written down is a pass nobody can check, and a marginal pass with an inflated area is the oldest trick in the book.

Field example: a 3 in wg supply main at class 6

Take a supply main in the 3 in wg pressure class, specified to leakage class 6, with 2,000 sq ft of duct surface in the section. The allowable rate is 6 times 3 to the 0.65 power, which is 6 times about 2.04, or roughly 12.3 cfm per 100 sq ft. Across 2,000 sq ft that is 12.3 times 20, about 245 cfm of allowable leakage at the 3 in wg test pressure.

Pressurize the section and the orifice reads 310 cfm. That fails, by about 65 cfm. A hand pass along the joints finds three transverse connections blowing and an access door with a tired gasket. Mastic on the joints, a new gasket on the door, a cure overnight, and the retest reads 180 cfm. Now it passes with room to spare, and against a system moving 10,000 cfm the 245 cfm allowable works out to under 3 percent, which is where the spec wanted it.

The lesson in the numbers is how much the joints carry. Three bad connections and one door were the whole 130 cfm difference between the failing 310 and the passing 180. The field of the duct was never the problem. The places where the pieces met were.

What if the duct fails the leakage test?

A failed section is normal on the first test, not a disaster. You seal it and retest the same section at the same pressure, and you keep the as-found and as-sealed numbers in the record. The first-test failure rate is high enough that a schedule assuming every section passes the first time is a schedule that slips.

Find the leaks under pressure with the hand-and-smoke method, seal the joints and penetrations with mastic, let it cure, and re-pressurize. If the second test still fails, the leaks are either somewhere you did not check, usually a penetration or a connection inside a wall or above hard ceiling, or the section boundary itself is leaking through a cap or bladder you trusted. Re-verify the boundary before you condemn the duct.

The trap is the section that passes only because the allowable is generous. A large section at a loose class can carry a real leak inside its tolerance, so it passes the number while a downstream room still starves. If the leakage test passes but the airflow is still wrong, the problem moved to balance, fittings, or a section that was never in the test scope. The leakage test clears one suspect. It does not close the case.

Leakage test vs the static pressure reading

The leakage test and the static pressure reading measure different things and get confused constantly. The external static pressure reading, the ESP a technician takes at the air handler, tells you the resistance the blower is fighting and, read against the blower table, the airflow it is moving. The leakage test tells you how much air is escaping the duct on the way to the registers. One is about the fan working too hard. The other is about the air going missing.

They are related symptoms of the same bad duct, though. A leaky supply trunk bleeds air, which actually lowers the static downstream of the leak while the fan still strains, so a tech can read a static number that looks acceptable while a quarter of the air is gone. That is why a commissioning agent does not stop at the static reading. The external static pressure guide on this site walks through measuring ESP and splitting supply from return to localize a restriction; the leakage test is the companion that catches the loss the static reading can hide.

Run both. Measure the static to see how hard the fan works and what airflow it makes, and leakage test the high-pressure duct to confirm that airflow actually reaches the space. Either one alone leaves a gap the other one fills.

Datacenter context: the plenum and supply integrity

In a datacenter the duct leakage problem becomes a plenum and supply integrity problem, and the stakes are higher because the leaked air is leaked cooling. A raised-floor cooling design pushes cold air into a sealed underfloor plenum, and the plenum pressure, often on the order of 0.05 in wg, drives air up through the perforated tiles to the server inlets. Every unsealed cable cutout, every gap at the floor edge, and every leak in the supply duct feeding the plenum is cold air going somewhere other than a server.

The mechanism is the same as a leaky supply trunk in a building, just laid out on the floor. Air that escapes through an open cutout into the wrong aisle, or short-circuits straight back to the return, cost energy to cool and did no work. That is bypass air, and it is the quiet capacity killer behind racks that overheat while the room average looks fine. The data center cooling overview on this site covers the airflow picture, and the raised-floor acceptance work is where the cutouts and the plenum seal get verified before the hall is loaded.

For ducted supply to CRAH units, in-row coolers, or overhead distribution, the SMACNA leakage test applies the same way it does in any building: pressurize the section, measure the loss, hold it to the spec. The difference in a datacenter is that the leaked cfm is not just wasted money. It is cooling a rack at the end of the row needed and did not get.

What to document

A leakage test nobody wrote up is a test you get to run again. The report is what the commissioning agent signs, the owner files, and the next engineer reads when a room or a rack runs warm and the question is whether the duct was ever tight.

Record the section identifier and what it covers, the tested surface area and how you figured it, the test pressure, the specified leakage class, the computed allowable leakage in cfm, the measured leakage, the pass or fail, and for a failure the as-found and as-sealed numbers with what you sealed. Note positive or negative test, the instrument and its calibration date, and who witnessed it. The calibration date is the one people skip, and it is the first thing an auditor asks for, because an uncalibrated orifice makes every number on the page a guess.

Common mistakes

• Claiming more surface area than the section actually contains, which inflates the allowable and passes a leaky duct.
• Testing at the wrong pressure, so the P to the 0.65 allowable does not match the duct's class.
• Leaving a branch, boot, or test port unsealed, then chasing a leak that is really an open boundary.
• Sealing transverse joints with tape and expecting the seal to last past a few seasons.
• Retesting on wet mastic before it has cured, so the pass does not hold.
• Reporting only pass or fail without the surface area, pressure, and class behind it.
• Testing a section so large that a real leak hides inside a generous allowable.
• Skipping the calibration date on the orifice, so the numbers cannot be defended.

Field checklist

Standards and references

The central reference for this work is SMACNA. The HVAC Air Duct Leakage Test Manual gives the leakage classes, the F equals CL times P to the 0.65 equation, the test procedures, and the apparatus, and the SMACNA HVAC Duct Construction Standards give the pressure classes and the seal classes the leakage classes are built to. The specific leakage class assigned to a given construction has changed between editions, so use the edition the project specification names rather than a number from memory.

ASHRAE Standard 90.1, the energy standard, addresses duct sealing and leakage by topic as part of the energy requirements, and the adopted edition and local amendments control. Test and balance and the broader airflow verification fall to the trade bodies that certify it, AABC and NEBB, whose procedures the commissioning agent witnesses. Above all of them sits the project specification, which names the ducts to test, the classes to hit, the test pressure, and the acceptance criteria. When the spec and the manual differ, the spec governs the job.

Cite the body that owns the point and confirm the current edition before putting a class or a number on a submittal. SMACNA owns the construction and the test, ASHRAE owns the energy requirement, the TAB bodies own the verification of airflow, and the contract documents control the limit.

Units, terms, and conversions

Duct leakage carries its own shorthand, and the same idea reads differently across a spec, a SMACNA table, and a metric drawing.

The leakage rate is cfm per 100 sq ft of duct surface in the field, while metric sources use liters per second per square meter. Test pressure is inches of water column, written in wg or in. w.c. or in. H2O, where 1 in wg is about 249 Pa. The leakage class CL is the rate at 1 in wg, and the seal class A, B, or C is the construction instruction, not the measured result.

Leakage class (CL)

Leakage rate in cfm per 100 sq ft of duct surface at a 1 in wg test pressure; lower is tighter

Seal class (A, B, C)

The construction instruction for where to apply sealant, set by the duct pressure class

cfm per 100 sq ft

The unit of leakage rate, airflow lost per 100 square feet of duct surface area

in wg / in. w.c.

Inches of water column, the test pressure unit; 1 in wg is about 249 Pa

Pressure class

The static pressure a duct is built and tested to: 1/2, 1, 2, 3, 4, 6, or 10 in wg

Transverse joint

The connection between two duct sections, the place leakage concentrates most

DALT

Duct air leakage test, the pressurize-and-measure procedure this guide describes

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is bypass airflow in a data center?
• How do I measure total external static pressure?
• What is a duct leakage test?
• What floor resistance passes ANSI/ESD S20.20?
• What is a chilled beam?