How technicians find an aircon gas leak: methods and limits

Refrigerant is not fuel. It carries heat around a closed loop and returns to where it started, so the charge should never fall on its own. When the charge is low, it went out through an opening. Adding gas refills the loop and says nothing about that opening.

The search is separate work because it needs separate tools and separate time on site. Gas goes in at the service port through a hose and a set of gauges. Finding where the gas went can mean isolating the circuit, raising pressure inside it, sweeping every reachable joint, opening casings, and returning for a second look. The two jobs do not carry the same effort, and treating them as one is how a leak survives a visit.

A top-up on its own is a wait-and-see, not a finding. There is nothing wrong with it so long as it is presented that way. The gas is restored, the system is watched, and there is a plan for what happens if the cooling fades again. It is something else when it is offered as the conclusion. Refrigerant that leaves once leaves again, so a visit with no search behind it books the next visit.

This is also why the sequence matters more than the tools. Prove the loss, size it, then narrow it to a section, then place it. Skip a step and the later methods return results nobody can act on, because there is no way to tell a genuine hit from stray vapour or an old stain.

The first job is proving the circuit leaks at all. A low reading at the gauges has more than one explanation, and an active leak is only one of them. The test that separates them is a standing pressure test. The circuit is isolated, filled with dry nitrogen above its normal working pressure, and left alone with a gauge on it. If the needle holds, the circuit is tight. If it falls, refrigerant has a way out.

Nitrogen is used instead of refrigerant for two reasons. It is inert, so a test that fails does not vent refrigerant into the air. It can also be pushed harder than the system ever runs, which turns a slow escape into a falling needle. A loss too small to notice in normal service becomes a number that will not sit still.

The standing test answers whether, not where. A circuit that will not hold has told you the search is worth doing and roughly how hard it will be. A fast fall means an opening wide enough that bubble solution will probably find it. A slow creep means the search moves to the sensitive methods, and it may take more than one attempt. Neither result puts a finger on the spot.

The rate also decides what a visit can promise. A circuit that empties quickly can usually be closed out in one attempt, because the opening is large enough for the direct methods to catch. A circuit that bleeds down slowly may hold its charge through an entire search without giving up one bubble or one repeatable alarm. Knowing the rate before the search starts is what lets a technician say which of those two visits this is going to be.

Where the pipe run is long, the circuit can be split and tested in sections. Valves at the outdoor unit let the indoor coil, the pipe run and the condenser be isolated from each other and pressurised on their own. Whichever section will not hold is the section worth opening. On a leak buried in a wall, sectioning is often the only way to narrow the search without cutting anything.

No single method covers the whole system. Each one trades sensitivity against where it can be applied, and that trade sets the order they are used in.

The order runs on likelihood first and reach second. Mechanical connections are swept before anything else, because they are both the most common failure point and the easiest surface to test: flare nuts at each unit, valve cores under the service caps, and the caps themselves. Brazed joints along the run come next. The coil block is last, not because coils never leak, but because reaching one costs the most and every cheaper answer deserves to be excluded first.

Bubble solution is the first pass on anything you can reach and see. A thick soap film is brushed over flare nuts, valve cores, service ports and brazed joints while the circuit sits under pressure. Escaping gas pushes the film into a growing bubble or a patch of foam. The evidence is direct. The bubble forms exactly where the opening is, so there is nothing left to interpret.

What soap misses is everything it cannot touch. A joint behind trunking, a tube inside a coil block, a pipe cast into a wall: none of these can be painted. It also misses seeps too slow to lift the film while someone stands there watching. Soap finds fast leaks in reachable places, which is a narrower category than most people expect.

An electronic detector reads refrigerant vapour in the air around the probe tip. It reaches into gaps a brush cannot, and it alarms on escapes far too slow to bubble. That sensitivity is also the problem. Refrigerant is heavier than air and pools in casings, in drain trays, at the base of a service cupboard, and around a ledge with no breeze. The probe alarms and the source may still be somewhere upwind.

Working around that means slowing down. The area is cleared of standing vapour, then the probe is swept low and slowly, moving upward from beneath the suspect surface because vapour falls. A hit is re-tested from clean air to see whether it repeats in the same spot. Cleaning sprays, aerosols and some plastics set the alarm off as well, so one beep proves nothing until it repeats.

UV dye moves the search onto the surface of the metal. A small quantity of fluorescent dye is introduced into the oil that circulates with the refrigerant. The system then has to run, so the dye reaches the leak, seeps out with a trace of oil, and dries as a bright stain under a UV lamp. The stain marks the spot and does not fade, which makes dye the method for seeps too slow to catch inside one visit.

Dye asks for two things in return: a second visit, and a line of sight. The system has to run long enough for the dye to travel and escape, so the technician leaves and comes back to read the result. The lamp then has to see the surface, which rules out concealed pipe runs and the inside of a coil block. Some manufacturers also decline dye in systems still under warranty, so it is a decision to check before it is a method to use.

Some leaks are findable in principle and unreachable in practice. Every method above needs to get near the metal. Three places defeat that: inside a coil block, inside a concealed pipe run, and behind trunking or a false ceiling. In each case the circuit can be proven to leak in that section, and the exact point stays hidden until something is opened.

A coil leak hides inside the fin pack. The tube losing gas sits behind rows of aluminium fins, so there is no surface to brush and no clear path for a lamp. The circuit can be pressurised and the coil bagged so vapour collects, then sampled at the opening to confirm the coil is the source. Confirming which tube, and whether it can be repaired at all, means stripping the unit down.

A concealed run is a different shape of problem. Pipe cast into a wall or lifted above a false ceiling can be isolated at both ends and tested as its own section, which proves the loss sits inside it. Placing the point means opening the trunking, lifting ceiling boards, or cutting. At that stage the question stops being diagnostic and becomes a scope decision: chase the point, or replace the run.

A joint behind trunking sits between the two. Trunking comes off, so the surface can be reached, but only once the covers are stripped along enough of the run to expose the joint. That is labour on a ladder rather than damage to the building, and it is worth doing after a section test has narrowed which stretch to open. Pulling trunking blindly along a whole run to look for a stain is the point at which a search stops being a search.

An honest report says which of these it is. A technician who names the section, shows the pressure result for that section, and explains what opening it would involve has done the work. A precise leak point announced for a concealed run, with nothing opened and no section test behind it, is worth one question. Ask what showed it.

A standing test holds one pressure at one temperature, and some leaks do not exist under those conditions. In service, the discharge line runs hot and the suction line runs cold. Copper expands and contracts across that swing, joints load and unload, and the whole assembly shakes each time the compressor starts. A crack that seals shut when the metal is cool and still can open once the line heats and moves.

That is why a first visit can prove a leak and fail to place it. The search then continues under running conditions. The system is brought to load and the sensitive methods are used with the pipe hot rather than at rest. Where that is impractical, dye is left in the circuit and the technician returns to read the stain once the system has been in normal use.

The complaint pattern often points at this. Cooling that fades only in the afternoon, or only after a long run, hints at an opening that appears with heat. A loss that follows a spell of heavy use hints at movement at a bracket or a poorly supported pipe. These patterns locate nothing on their own. They tell the search where to look and under what conditions.

Saying this before the work starts is part of the job. A leak search is a search, and searches produce results that are not yet answers. The useful version of that is specific: this section holds, this section does not, the seep is slow enough that dye is the next step, and here is what the return visit will read. The version worth questioning is the one where nothing was tested and the plan is to add gas and see.