How it works
Why a trap loses its seal
Two inches of water stand between the drain and the room. Pull on them, push on them, and find out which one empties a trap; then see what an S-trap and a vent do.
A trap holds a few inches of water in a U-shaped bend, and that water is all that separates the drain from the room. It is not fixed in place. It sits there because the air pressure on both sides is the same, and a difference of a couple of inches of water, about 500 pascals, is enough to move it out of the way. Everything a vent does, and every way a trap fails, follows from treating the trap for what it is: a manometer [1].
The trap is a manometer
The weir is the lip where the trap’s outlet turns toward the drain; water stands level with it when the trap is full. The dip is the top of the lower bend. The vertical distance between them is the trap seal, d, normally 2 to 4 inches [2]. If the trap has lost some water, the depth that remains is s, and both sides stand d − s below the weir [2].
A column of water h tall balances a pressure difference of ρgh, which is why plumbers and engineers measure drain pressures in inches of water column: one inch is 249 Pa [3]. So when the air in the drain is at p inches of water relative to the room, the two water surfaces in the trap end up exactly p inches apart [1]:
Here z is each water surface’s height, p is negative for suction in the drain, and both are in inches. Equation (1) is the whole of the hydrostatics. The interesting part is what the trap’s shape does with it.
Suction: a full trap breaks at its own depth
Pull on a full trap from the drain side. Equation (1) wants the outlet surface to rise, but the outlet is already level with the weir, so the water spills over into the drain. The inlet surface falls instead, by the whole suction a = −p. When a reaches d, the inlet surface reaches the dip, and the next increase lets room air through the bend in a gulp. That is the gurgle, and it is exactly the failure the National Bureau of Standards described, with the inlet leg “pulled down to the level of the dip” and air passing through [2].
A trap that has already lost some water is different. Its outlet surface sits below the weir, so the first part of the suction lifts the outlet side without spilling: the two surfaces share the movement, half each. In general the suction that breaks a seal of remaining depth s is
A full 2-inch trap breaks at 2 inches; a trap left with 0.6 inches breaks at 1.2. Drag the “water left” slider in the figure down and watch the red mark on the suction side move toward zero.
Back-pressure: twice the depth
Push from the drain side instead, as happens when a slug of water coming down a stack drives air ahead of it. Now both surfaces are free to move: the outlet side goes down and the inlet side comes up, equal amounts in a trap whose two sides have the same bore. The drain side reaches the dip when the surfaces are 2s apart:
So a full trap holds twice as much back-pressure as suction. That asymmetry is why most of the engineering of vents is about suction, and it is also why the codes still insist on a path for air to leave the drain: when back-pressure does break a seal, it is drain air that comes up into the room, and the International Plumbing Code describes positive pressure pushing a seal back and spilling water or odour [4].
What a pull leaves behind
Back-pressure moves water but, in the ideal picture, loses none: release it and the levels return. Suction is different, because every inch the outlet side would have risen above the weir has gone down the drain. When the suction is released, the remaining water settles level again, so a pull of a inches on a full trap leaves
A 1-inch pull on a 2-inch trap costs half an inch. And the damage accumulates: a trap left with 1.5 inches still breaks at 2 inches of suction by equation (2), but a further pull of 1.5 inches drains it to 1.25, and so on. A seal that has been pulled holds less back-pressure (equation 3), and once less than half of it is left it breaks at less suction too.
Self-siphonage and the S-trap
Where does a suction of several inches come from? The most direct source is the fixture’s own water. When a basin empties, its discharge can fill the drain below the trap completely, and a column of water falling full-bore through a pipe needs air behind it. If no air can reach the top of that column except through the trap, the column hangs on the trap like the long leg of a siphon, pulling with up to its own height in inches of water. The NBS study of this self-siphonage found the loss depends on the trap’s diameter and seal, the drain’s diameter, length and slope, the vent fitting and the rate of discharge [2].
The chart has the shape equation (4) predicts. With no vent, a short fall pulls less than the seal resists and costs a little water; once the fall passes the seal depth, the seal breaks and the trap is left with at most half its water, however long the leg. An S-trap is the extreme case, a trap built with its own siphon leg, and the Uniform Plumbing Code prohibits it [5]. A vent at the top of the leg changes the curve completely, because air now reaches the falling water without passing the trap. The code’s answer is geometric: every trap gets a vent within a set distance, and the vent opening may not sit below the trap’s weir [5], so there is no unvented full-bore drop for the water to hang on.
Why traps are 2 to 4 inches deep
Equation (2) says a deeper seal resists more suction, so why not make traps 12 inches deep? Partly because depth is a poor defence against the real threat: a siphon leg is many inches long and beats any practical seal, and a vent removes the threat outright. Partly because the water in a trap has to be swept through by every use. The codes require traps to be self-cleaning and cap the seal at 4 inches while requiring at least 2 [5]; a deeper bend holds more standing water and more of whatever settles in it. Inside that range, deeper is better, which the NBS tests also found: a deeper seal lost more water but kept more [2].
The vent: keeping the pressure inside the band
Water draining through a building’s pipes drags air along with it, both behind falling slugs and as a core of air pulled down a stack inside the falling film of water [6]. A vent is the low-resistance path that supplies that air. How much pressure it costs is ordinary pipe-flow physics: the Darcy–Weisbach loss plus the losses at fittings, entries and exits, the method the NBS used to size vents [6]:
For 5 litres a second of air through 20 feet of 1½-inch pipe, with a friction factor of 0.025 and fitting losses of 1.5, equation (5) gives about 64 Pa, a quarter of an inch of water: well inside the ±1 inch band that vent tables are built on [6][4]. The pressure grows with the square of the air flow and, for the same flow, with the fourth power (or more) of a shrinking diameter, which is why a partly blocked vent fails fast once it starts. When the vent cannot supply the air, the pressure keeps falling until the next path opens, and the next path is the weakest trap on the branch.
Wind across the roof vent
A vent ends in the open air, and moving air is at lower pressure than still air. Bernoulli’s equation gives the scale: the dynamic pressure of the wind, ½ρv2 [1]. Some fraction of it, set by a pressure coefficient Cp that depends on the shape of the terminal and the roof around it, appears at the top of the vent:
At 55 mph (24.6 m/s) the dynamic pressure is about 364 Pa, almost an inch and a half of water; with an illustrative Cp of −0.5 the vent sees about −0.73 inches. That stays inside the band, but it rises with the square of the wind speed and it comes in gusts, so the seal rocks and spills a little each time.
Evaporation: the slow failure
The last way to lose a seal needs no pressure at all. Water evaporates from both surfaces of a trap that nobody uses, and the seal falls at the evaporation rate e:
The rate is the uncertain part. Experiments published in 1884 found from about one-seventh to over one-third of an inch a day in vented traps, fastest in heated winter air, slowest in damp summer weather [8]. Those traps had a ventilating draught across them; without one the same tests found evaporation “exceedingly slow” [8]. Long before the trap is empty it becomes easy to break: by equations (2) and (3), a seal half gone holds only half the back-pressure, and one three-quarters gone breaks at half the suction. That is why a dry trap is the most common cause of indoor sewer odour [9], why a weekly splash of water fixes it [10], and why codes call for trap primers on floor drains that see little use [5].
How we know: glass pipes and a street sewer in a laboratory
Much of what codes say about traps and vents was measured at the National Bureau of Standards in Washington in the years after the Second World War, for the federal housing programme. For the self-siphonage study, French and Eaton connected standard fixtures to complete drainage systems that emptied into an 8-inch “street sewer” in the laboratory, whose flow they could vary up to 300 gallons a minute. The stacks and some fixture drains were built of transparent methacrylate tubing so they could watch what the water and air did; the numbers they reported came from repeat tests in ordinary metal pipe [2]. They filmed air bubbling up through a transparent trap after a basin’s discharge ended abruptly, the gulp at the end of step 2 on the home page [2].
They also asked how much loss is too much. Their answer was a rule of thumb still worth knowing: what keeps sewer air out is the seal that remains, and a remaining seal of 1 inch was the limit commonly used [2]. A decade later, Wyly and Eaton sized vent stacks so that pressures stay within ±1 inch of water, computed with exactly the pipe-friction method of equation (5) [6]. That band is the amber stripe in every figure here.
The research did not stop there. Later work, notably at Heriot-Watt University, treats the air in a drainage system as carrying pressure waves that travel and reflect along the pipes, and fits measured single-stack systems with expressions for the pressure at critical points [7]. Those transient models are what designers of tall buildings use; the one-pressure model in these figures is the quasi-static limit of the same physics, good for seeing why a trap fails, not for designing a stack.
How it is calculated
The trap model is equations (1) to (4) for two connected columns of equal bore, with water that rises above the weir spilling into the drain. The branch pressure is found by balancing air: the air the water takes away equals the air that comes in through the vent (equation 5), an air admittance valve (closed until its opening suction, about 0.28 in, then a square-law flow sized to pass 7.5 L/s at 1 in, an illustrative rating [11]), and the trap once its seal has broken. A siphon leg cannot pull more than its own height. Wind adds equation (6) at the vent terminal. All pressures are gauge pressures in inches of water relative to the room; 1 in. w.c. = 249.0889 Pa [3].
Before you open anything. Sewer gas carries hydrogen sulfide, which deadens the sense of smell at high concentrations [12], and can carry methane, which is explosive between 5 and 15 percent in air [13]. Vent and stack work, roof access and changes to drain layout are jobs for a licensed plumber, under codes that differ from place to place. Nothing on this page says what your local code allows.