Steam heating and trap commissioning field guide for HVAC

A steam system, and where it still runs

A steam heating system carries heat by boiling water in a boiler, pushing the steam out to the terminals, and letting it condense where the heat is wanted. The steam gives up its heat by condensing back to water, and that water, the condensate, drains back to the boiler to be boiled again. The heat is in the phase change, not in the temperature of the water.

Steam heating is old, and people assume it is gone. It is not. It still runs in the places that were built when steam was the only practical way to move a lot of heat a long way: the prewar apartment buildings and high-rises, the campuses with a central plant and miles of distribution, the hospitals, the schools, and the industrial plants that need steam for process and heat the building off the same loop. If you work commercial or institutional mechanical, you will meet steam, and the systems you meet are often eighty years old with sixty years of bad repairs on them.

A steam system comes in two basic shapes, one-pipe and two-pipe, and the trap belongs to the two-pipe kind. The boiler startup and commissioning guide covers firing the vessel and proving its safeties. This guide is about everything downstream of the header: getting the steam out, getting the air out of the way, getting the condensate back, and the small device that makes a two-pipe system work or quietly bankrupts it.

Why steam carries so much heat

The reason steam moves so much heat is the latent heat of vaporization. To turn a pound of 212°F water into a pound of 212°F steam at atmospheric pressure takes about 970 Btu, and the steam carries that 970 Btu with it until it condenses. When it reaches a cold radiator and condenses back to water, it dumps all of that latent heat into the room in one phase change, at constant temperature. A pound of hot water cooling 20°F gives up about 20 Btu. A pound of steam condensing gives up roughly 970. That ratio is why a steam pipe can be small and still heat a building.

Steam temperature and pressure are locked together. Saturated steam has one temperature for every pressure, and you cannot change one without the other. At 0 psig, atmospheric, steam is 212°F. At 15 psig it is about 250°F. Raise the pressure and you raise the temperature, and the steam tables list the pair for every point. As pressure climbs the latent heat actually drops a little, from about 970 Btu/lb at atmospheric to about 946 Btu/lb at 15 psig, while the sensible heat in the water rises. The tradeoff is real but small across the heating range.

Building heat runs low pressure, generally under 15 psig and often only ounces above atmospheric. Higher pressure does not buy you more heat per pound. It buys you a hotter pipe and more standby loss. High-pressure steam belongs to process and to district distribution, where you need the temperature or you need to push the steam a long distance and then reduce it at the building. For a heating system, lower is better, and many old systems run far higher than they need to because nobody ever turned the pressure control down.

What is the difference between one-pipe and two-pipe steam?

In a one-pipe system the steam and the condensate share the same pipe. Steam rises through the single riser into the radiator, the air ahead of it escapes through an air vent on the radiator, and the condensate drains back down the same pipe it came up. There is no trap. The radiator vent is the whole story: it lets the air out so the steam can get in, then closes when the steam reaches it. One-pipe is simpler, cheaper, and found in smaller and older buildings, and it is fussy about pitch because the condensate has to run back against the steam coming up.

In a two-pipe system the supply and the return are separate. Steam comes in one connection, condensate and air leave through the other, and a steam trap sits at the outlet of every terminal. That trap is what keeps the two pipes doing two jobs: it passes the condensate and the air into the return but blocks the live steam from blowing straight through. Two-pipe is what you find in larger buildings and anywhere the system is bigger or the control is tighter, because the separate return lets the system run at lower pressure and heat more evenly.

The quick field tell is to count the connections at the radiator. One connection with a vent on the side is one-pipe. Two connections and no vent is two-pipe with a trap on the return side. The distinction matters because the failure you chase is different. On one-pipe you chase air vents. On two-pipe you chase traps.

What does a steam trap do?

A steam trap is an automatic valve that passes condensate and air but holds back live steam. That is the whole job, and it is harder than it sounds, because the trap has to tell the difference between water and steam, or between hot air and steam, and open or close fast enough to keep up with a load that changes all day. The trap is the heart of a two-pipe system, and it is also the part most likely to be dead.

Think about what the trap is protecting. Steam gives up its heat by condensing in the terminal, which means the terminal is constantly making condensate, and that condensate has to leave or it piles up and floods the heat transfer surface. The trap lets it leave. At the same time, if live steam were allowed past the trap into the return, you would send your most expensive Btu straight back to the boiler unused, and you would pressurize the return so the other traps cannot drain. The trap holds the steam back until it has condensed and given up its heat.

A trap fails two ways and they are opposites. Failed open, it stops holding steam and blows live steam into the return, which wastes fuel and can back-pressure the whole return. Failed closed, it stops passing condensate, the terminal waterlogs, the heat dies, and you have set up the slug that causes water hammer. A good trap does a quiet, invisible job. A bad one is either an open wallet or a cold room, and from the floor they look identical.

There is one more thing the trap does that people miss: it vents air. On startup the terminal is full of air the same as the mains, and many traps, especially the thermostatic and F&T types, open on that cool air and let it out ahead of the steam. A trap that has failed closed traps that air, and the terminal stays cold even with steam at the door. So a dead trap can present as a cold radiator that looks like an air problem, and the only way to know which it is is to test the trap, not guess from the symptom.

Steam trap types and where each one goes

There is no universal trap. You pick the type for the load, the pressure, and whether the application makes a lot of air. Get the type wrong and even a new trap fails early or never works right.

Thermostatic traps sense temperature. They hold until the condensate has cooled a little below steam temperature, then open, which makes them good on radiators and small terminals and good at venting air. Float and thermostatic traps, the F&T, combine a float that rides the condensate level with a thermostatic air vent built in; the float gives continuous drainage that follows a changing load, so the F&T is the common choice on heat exchangers, coils, and anything that modulates. Inverted bucket traps are mechanical, working on the density difference between steam and water, durable and tolerant of dirty condensate, but they need water to seal and can lose their prime. Thermodynamic disc traps work on the velocity difference between fast steam and slow condensate in a small rugged body that takes superheat, freezing, and vibration, which is why they live on main drip legs and outdoor lines rather than on terminals.

The application drives it. A modulating coil wants an F&T. A drip leg on a steam main wants a thermodynamic or an inverted bucket. A radiator wants a thermostatic. The one rule that crosses all of them: the trap has to be sized for the actual condensate load at the actual pressure differential. A trap sized off the pipe size instead of the load is a trap that will fail to keep up or short-cycle itself to death. Match the trap to the manufacturer's selection data for the application, not to what was on the shelf.

How do you test a steam trap?

You test a steam trap three ways and you trust the combination, not any one alone: temperature, sound, and sight. The fastest is temperature, because a trap that is stone cold on the inlet is not passing anything and the terminal behind it is waterlogged. But temperature alone lies to you, because flash steam reads as hot as live steam, so a hot trap can be a healthy trap cycling or a failed-open trap blowing through, and a thermometer cannot tell them apart.

Sound separates them. An ultrasonic tester listens at the trap, and a person who knows the trap type can hear the difference between the intermittent snap of a trap cycling normally and the continuous rush of live steam blowing straight through a failed-open trap. Acoustic testing is the core of a real trap survey because it catches the failed-open trap that temperature misses. Sight is the third leg: where you can see the discharge, into a flash tank or an open return, you can watch whether the trap dribbles condensate and flashes a little, or roars a steady plume of live steam.

Use all three and you can classify a trap as good, failed open, or failed closed with confidence. A trap survey program does this to every trap on a schedule, tags each one, and feeds a replace-or-leave list. The single biggest reason facilities waste steam is not that traps fail. It is that nobody is testing them, so the failures sit for years. Test cold, test by sound, and confirm by sight where you can.

One field note on the survey itself: test the trap, then read it against the trap type. The normal sound and cycle of a thermostatic trap is not the same as an inverted bucket or a thermodynamic disc, so the person doing the survey has to know what type each trap is before calling it good or bad. A thermodynamic disc that cycles a few times a minute is working; the same pattern would be wrong on a different type. This is why the survey list carries the trap type next to the tag, and why a survey done by someone who does not know the types produces a list nobody can trust.

What a failed-open trap costs

A failed-open trap is a hole that runs around the clock at full boiler pressure. It does not trip anything, it does not make a noise the tenant notices, and it bills you every hour of the heating season. The numbers people put on it are large for a reason: a single failed-open trap can run into the tens of thousands of dollars a year in wasted fuel, and across a building the waste compounds.

The reason it stays hidden is that a blowing trap often heats fine. The room is warm, the tenant is happy, and the steam roaring uselessly into the return never shows up as a complaint. It shows up only on the gas bill, which nobody ties back to a small valve in a closet. That is the trap survey's whole argument: the failures are invisible from the comfort side, and only the test finds them.

The fleet-level number is the one that gets a budget approved. In buildings with no testing program, surveys routinely find 15 to 30 percent of traps failed, and a system left unmaintained for a year can have half its traps in a failed state. By one common estimate, leaking traps waste more than 20 percent of the steam a plant produces. Best-run plants hold failure under 5 percent by surveying on a schedule and replacing on a schedule. The gap between 30 percent and 5 percent is real money, every year, and it is the easiest energy savings in the building.

The condensate return, and why you keep every drop

Condensate is hot, treated, near-distilled water, and it is worth real money, so the return system exists to get it back to the boiler instead of dumping it. After the steam condenses in the terminal and the trap passes it, the condensate drains by gravity to a low point and collects in a condensate receiver, and a condensate pump sends it back to the boiler or the feed tank when the level rises. On bigger plants a deaerator strips the oxygen out before the water goes back to the boiler.

Lose condensate and you pay three times. You pay to make new water, you pay to treat it, and you pay to heat it from cold instead of from 180°F or hotter. Every gallon of condensate you fail to recover is a gallon of cold makeup the boiler has to bring up from scratch, carrying dissolved oxygen that attacks the boiler from the inside. A return with failed traps blowing steam into it, flooded lines, or a dead condensate pump is bleeding money and shortening the boiler's life at the same time.

Watch the flash steam. When hot condensate at pressure drops to a lower pressure in the receiver or the return, some of it flashes back to steam, and that flash steam is normal, not a trap failure. Good systems capture flash steam and use it. Bad ones vent it to atmosphere through the receiver vent, which is wasted heat and the visible plume people mistake for a problem. Know the difference before you condemn a trap for the steam coming off the receiver.

Air binding and the air vent

Air is the enemy of steam heat, and the first thing that has to happen on every cycle is the air getting out of the way. A cold system is full of air. When the boiler makes steam, that steam has to push the air out ahead of it, and if the air cannot escape, the steam cannot get in. The radiator stays cold, not because there is no steam, but because the air is blocking it. That is air binding, and it is the most common reason a one-pipe radiator does not heat.

On a one-pipe system the air leaves through the radiator vent and the main vents on the steam mains. The radiator vent lets air out and closes when steam reaches it. A vent plugged shut leaves the radiator air-bound and cold, and a vent stuck open spits steam and water into the room. Main vents at the ends of the steam mains are the ones people forget, and they matter more than the radiator vents, because venting the mains fast is what gets steam to every riser at once instead of the near radiators cooking while the far ones stay cold.

On a two-pipe system the air leaves through the trap and into the return, which is one more job the trap is doing, and a trap that has failed closed traps the air as well as the condensate. Either way the rule is the same: the air has to get out for the steam to get in. When a radiator will not heat, check whether it is air-bound before you blame the boiler.

What causes water hammer in steam?

Water hammer is a slug of condensate picked up and thrown down the pipe by moving steam, and the bang is that slug slamming into a fitting, a valve, or an elbow. Steam moving over a pool of condensate in a low spot drags the surface, builds a wave, and then closes the pipe off into a piston of water that the steam launches at speeds that can pass 100 feet per second. When it hits something solid it stops dead, and all that momentum becomes a hammer blow that can split a fitting or knock a valve apart. It is dangerous, not just loud.

The condensate has to be there for it to happen, so water hammer is a condensate-management failure. A line that sags between hangers, is pitched the wrong way, or has no way to drain at a low point holds a pool of water waiting to be picked up. A trap that has failed closed backs condensate into the terminal and the line. And the classic trigger is warming a cold main too fast: full steam pressure hits a cold pipe, condenses violently against the cold metal, and the first wave of steam drives the resulting slug down the line before the traps can clear it.

Two-pipe and one-pipe hammer for related but different reasons. On two-pipe, the trap that failed closed is the usual culprit, backing condensate where the steam will find it. On one-pipe, the radiator pitched the wrong way holds water at the far end, and the steam coming up the single pipe slams it; the cure is often just shimming the radiator to drain back toward the supply valve. Either way the bang is condensate in the wrong place, and the fix is to get the condensate out of the steam's path.

The cure is condensate control, not bigger pipe. Pitch the lines to drain, drip the low points to a trap, keep the strainers clean so the traps do not starve, and warm the system up slowly so the steam front and the condensate stay close in temperature. A system that hammers on every startup is telling you it cannot clear its condensate. Find out why before something lets go.

Drip legs, strainers, and dirt pockets

A drip leg is a vertical pocket at a low point of a steam line that collects condensate and lets a trap drain it before it can pool in the running pipe. Every place a steam main reaches a low point, an end, a riser, or a point ahead of a valve or a piece of equipment, the condensate that forms in the pipe has to be caught and dropped out, and the drip leg is how you catch it. Skip the drip legs and the condensate has nowhere to go but down the main, where it becomes the slug that hammers.

Size the drip leg full pipe size where you can, not a skinny stub off the bottom. A small drip leg fills and overflows back into the main, which defeats the point, especially on warm-up when the condensate load is highest. Run the connection to the trap a few inches up from the bottom of the leg, and leave the bottom as a dirt pocket with a blowdown valve. That stub below the trap takeoff is where the scale and rust settle out, so they stay out of the trap.

A strainer ahead of every trap catches the debris that would otherwise wedge the trap open or closed. Old steam systems are full of loose scale, and a single chunk in a trap seat will hang it open and blow steam or hang it shut and waterlog the terminal. Blow the strainers down and clean them as part of the trap program. A trap that keeps failing in the same spot usually has a dirt problem upstream, not a bad trap.

Warm-up and seasonal startup

Bringing a steam system up from cold is the moment it is most likely to hammer, so you do it slowly and you drip the lines as you go. A cold main full of air becomes a cold main full of condensate the instant steam hits it, and if you throw the main valve wide open on a cold system you make condensate faster than the traps can clear it and you drive slugs down the pipe. The slow warm-up keeps the steam front and the pipe close in temperature so the condensation is gentle and the traps keep up.

Use the warm-up valve if the station has one, or crack the main valve and bring the pressure up in stages. Open the drip leg traps and let the startup condensate run off. Listen as you go. The first warm-up of the season is when you hear which lines hammer and which radiators stay cold, and it is the best free survey you get all year. A system that warms up quiet and heats every terminal is in good shape. One that bangs and leaves cold radiators is telling you where the work is.

On seasonal startup, before the cold weather, walk the system once it is warm: every trap, every vent, every radiator. The traps you find dead in October are the ones that would have wasted fuel all winter. This is also where you reconcile the water side with the boiler. The boiler startup and commissioning guide covers the firing side and the safeties; the steam-side warm-up is proving the distribution clears and the heat reaches every terminal without hammer.

Insulation on the steam and condensate lines

Steam and condensate piping has to be insulated, and on an old system it usually is not, or it was and the lagging is gone. Bare steam pipe condenses steam along its whole length, which means it makes condensate the traps then have to clear, wastes heat where you do not want it, and burns anyone who touches it. Insulation on the steam main is not about comfort. It is about keeping the Btu in the pipe until it reaches the terminal that is supposed to release it.

The condensate lines get insulated too, to keep the recovered heat in the water you are sending back to the boiler. The exception people forget is that you do not insulate the radiator itself or the terminal that is supposed to give its heat to the room. Insulate the distribution, release the heat at the terminal.

Watch for old pipe lagging on systems that age out. Asbestos was the standard steam insulation for decades, and disturbing it is a regulated abatement job, not a maintenance task. If the lagging is suspect, stop and get it tested before anyone cuts into it.

The boiler tie: water level, the low-water cutoff, and the PRV

The steam side and the boiler are one system, and the water level is what links them. The boiler boils water into steam, the steam goes out and gives up its heat, and the condensate comes back. The boiler water level rises and falls with how much steam is out in the system and how much condensate has returned. If the condensate does not come back, because a pump is dead or a return is flooded, the boiler runs low on water while the system is full of it.

The low-water cutoff shuts the burner off before the water drops far enough to expose and damage the vessel, and on a steam boiler it matters even more than on a hot-water boiler, because the level moves all the time with the steaming load. The boiler startup and commissioning guide covers proving the low-water cutoff and the rest of the safety chain in detail, and that work belongs to the boiler commissioning, not repeated here. What the steam-side tech owns is the connection: condensate coming back fast enough and clean enough to hold the level, makeup feeding correctly when it does not, and the float and probe cutoff devices kept clean so they read the real level.

Where the building reduces high-pressure steam to low pressure for heating, the pressure-reducing valve station is the other pressure boundary, and it carries its own safety. A proper station has a strainer and a drip trap ahead of the valve, a gauge on each side, and a relief valve on the low side sized to pass what the reducing valve could pass wide open. That relief is what stands between the low-pressure system and full upstream pressure if the PRV sticks open. It is an ASME-stamped device, like the boiler's relief, and it is not optional. Commission the station by proving the reduced pressure holds at setpoint across the load range and the relief lifts.

Commissioning a steam heating system

Commissioning a steam system proves it does the whole job: steam reaches every terminal, the air gets out, the condensate comes back, the traps hold steam, and nothing hammers. You can fire the boiler perfectly and still have a system that does not heat the top floor, so the steam-side commissioning is its own scope on top of the boiler startup.

Work it in order. Prove the distribution first: warm the system up slowly and confirm steam reaches every main and every riser, with the main vents and radiator vents clearing the air fast enough that the far terminals heat about when the near ones do. Then prove the terminals: every radiator or coil gets hot, and none stays cold from a closed trap, a plugged vent, or an air bind. Then prove the traps: test each one and confirm none is blowing through and none is waterlogged. Then prove the return: condensate gets back to the receiver and the pump returns it to the boiler without flooding or losing the water. Then prove there is no hammer through a full warm-up cycle from cold.

Document the trap survey as the baseline, because that is the record the owner inherits and the one the next survey is measured against. A steam commissioning that does not leave a trap-by-trap baseline gave the owner nothing to maintain against. Tag every trap, log its type, location, and test result, and hand over the list.

Keeping it running after turnover

The day the system is commissioned is the best shape it will ever be in unless the owner runs a trap survey program, and most do not. Traps fail at a steady rate, and a system left alone drifts from 5 percent failed toward 30 percent and beyond over a few years, with the fuel waste climbing the whole time. The single highest-return maintenance task on a steam system is surveying the traps on a schedule and replacing the failed ones.

Build the program around a route. Every trap is tagged and on a list with its type, size, location, and last test result. The survey runs the whole route on a cycle, commonly once a year for a heating system, testing each trap by temperature, sound, and sight, marking it good, failed open, or failed closed, and feeding the failures to a replacement list. The replacements get done, the list gets updated, and the failure rate stays low instead of creeping. That is the entire program, and it pays for itself in fuel many times over.

The rest of the maintenance hangs off the same route: clean and blow down the strainers, check and replace the radiator and main vents on one-pipe systems, confirm the condensate pump and receiver are working, and walk for hammer and cold terminals at seasonal startup. None of it is exotic. The skill is doing it on a schedule instead of waiting for a complaint, because the most expensive failures, the blowing traps, never generate a complaint at all.

One more reason the program matters: the failures hide from everyone but the meter. A failed-open trap heats the space fine and generates no complaint, so without a survey it never surfaces. A failed-closed trap shows as one cold radiator that gets a service call, gets blamed on the boiler or the vent, and the trap behind it never gets tested. The survey is the only thing that looks at every trap on purpose, which is why the buildings that run one hold their failure rate low and the ones that do not drift into wasting a fifth of their steam without ever knowing it.

Safety around steam

Steam burns are the hazard people underestimate, because steam is invisible until it condenses, and a leak of dry steam can be on you before you see it. Steam at 250°F carries far more heat than 250°F water, because it dumps its latent heat the instant it condenses on your skin, so a steam burn goes deep fast. Treat any steam leak as a serious burn hazard, and never put a hand near a suspected leak to find it. Use a board or a rag on a stick to locate it.

Pressure is the other hazard, and it ties back to the boiler. The boiler and its appurtenances are a pressure vessel under ASME rules, the relief valves are ASME-stamped, and the jurisdiction inspects them. Never lift, gag, or pin a relief valve, never isolate one with a valve, and never work on a trap or a fitting on a line you have not isolated and proven depressurized. A trap or a union opened on a live line is a steam release at face height.

Condensate is hot enough to scald even after it leaves the steam, and old systems carry their own hazards: asbestos lagging, lead-caulked joints, and decades of brittle fittings that fail when you put a wrench on them. Confirm the line is cold and dead before you open it, and when an old system fights you, slow down. The pressure and the heat do not forgive a shortcut.

What to document

The deliverable from a steam commissioning or a trap survey is the trap schedule, the line-by-line record of every trap with its type, location, test result, and the action taken. That record is what the owner maintains against and what the next survey compares to, and a survey with no record has to start over next year.

Capture each trap by a tag number tied to a location, its type and size, the test result and how it was tested, the action, and the date and who did it. Add the supporting items the system depends on: the main and radiator vents, the condensate pump and receiver, the PRV station setpoint and relief test, and any line that hammered on warm-up. The record below is the minimum that makes the next person's job possible.

Common mistakes

• Running the system at higher pressure than the heating load needs, which adds standby loss and buys no extra heat.
• Leaving failed-open traps in service, blowing live steam into the return and the fuel bill with no comfort complaint to flag them.
• Letting a terminal stay cold from a failed-closed, waterlogged trap and blaming the boiler instead of testing the trap.
• Running no trap survey program, so failures sit for years and the failure rate creeps past 30 percent.
• Warming a cold system too fast with no drip legs working, which sets up the condensate slug that hammers.
• Ignoring air binding: a plugged radiator vent or a dead main vent leaves terminals cold while the boiler runs fine.
• Losing condensate to atmosphere through flooded returns or a dead pump, then paying to make, treat, and heat cold makeup water.
• Sizing a trap off the pipe size instead of the condensate load and pressure differential, so even a new trap cannot keep up.

Field checklist

Standards and references

The framework on the boiler and pressure side is ASME. The boiler is a fired pressure vessel built and stamped to the ASME Boiler and Pressure Vessel Code, its relief valves are ASME-stamped, and the jurisdiction inspects it. The boiler startup and commissioning guide covers that side, and the PRV station's relief valve follows the same logic. Do not treat any relief device as optional or adjustable in the field.

For the traps themselves, the manufacturer's selection and application data govern. There is no single code number that tells you which trap goes where or how big it is. You size and select from the trap maker's published capacity data for the type, the pressure differential, and the condensate load, and you follow their installation requirements. ASHRAE handbook material on steam systems gives the design framework for distribution, sizing, and condensate return, and the Department of Energy steam best-practice material is the common reference for trap surveys and the energy case behind them.

As always, the project specification controls. Where a spec calls out a trap type, a survey interval, an operating pressure, or a documentation standard, that governs over any rule of thumb here. Confirm the trap selection against the manufacturer, the pressure-side safety against ASME and the jurisdiction, and the survey program against the spec and the DOE guidance. Never cite a section number you have not verified for the edition in force.

Units, terms, and conversions

Steam work has its own vocabulary and a couple of unit conventions worth keeping straight across a drawing set and a manufacturer sheet.

Pressure on a heating system is usually gauge pressure, psig, and often only ounces above atmospheric, where the gauge may read in ounces per square inch or inches of water column rather than psi. Temperature and pressure are tied by the saturation relationship, so a steam pressure implies a steam temperature off the steam tables. Heat content is in Btu per pound, split into the sensible heat in the water and the latent heat in the phase change. Condensate load and trap capacity are rated in pounds per hour of condensate at a stated pressure differential, not in gpm, and that pounds-per-hour figure is the number to match when you size a trap.

Latent heat

The heat absorbed or released in a phase change, about 970 Btu/lb when steam condenses at atmospheric pressure

Saturated steam

Steam at the temperature that matches its pressure; raise the pressure and the temperature rises with it

Condensate

The water that forms when steam gives up its heat and condenses, returned to the boiler to be reused

Flash steam

Steam that re-forms when hot condensate at pressure drops to a lower pressure, normal and not a trap failure

Steam trap

An automatic valve that passes condensate and air but holds back live steam

Drip leg

A pocket at a low point that collects condensate so a trap can drain it before it pools in the main

psig / ounces

Gauge pressure above atmospheric; heating steam often runs only ounces, well under 15 psig

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is boiler commissioning?
• How do you balance a hydronic system?
• How does steam heat work?
• What causes water hammer?
• Why does my AC condensate drain clog?