Aircon vacuum and pressure test: what to ask installers
A new aircon cools on handover day whether or not the pipework was properly prepared. Two steps decide that, both finished before the refrigerant goes in, and neither leaves anything you can inspect once the job is closed.
By Team Snowflake | Updated 4 Aug 2026
The two steps that finish before any refrigerant goes in
Every split-system install ends with the same sequence. The pipework is joined, proved, emptied, and only then charged with refrigerant. The proving is a nitrogen pressure test. The emptying is evacuation, which most installers call pulling a vacuum. Both are done before the first gram of gas enters the circuit, and both vanish from view the moment the job is closed.
Neither step leaves anything behind for you to inspect. A system that was rushed through both looks identical to one that was done properly, and it cools on handover day either way. The consequences of skipping them are slow rather than immediate, so the finished job gives you no signal at all about which one you paid for.
That is also why they are the first two items to quietly leave a thin quote. Both cost time and no materials. An installer working to a fixed price and a full schedule saves more by shortening the vacuum than by using cheaper brackets. The shortcut is also one nobody will ever be able to point at. This is an incentive problem, not a character problem, which is exactly why asking about it works.
What does pulling a vacuum actually remove?
Evacuation removes air and moisture from the pipework so that only refrigerant is left inside. Open pipe ends sit in humid Singapore air for the whole install, and water vapour settles on the inner wall of the copper. A vacuum pump lowers the pressure inside the sealed circuit far enough that the water boils at room temperature, turns to vapour, and gets drawn out with the air.
Moisture left in the pipes does not sit there harmlessly. It reacts with the refrigerant and the compressor oil and turns the mixture acidic, and that mixture circulates through the compressor on every run cycle, attacking the windings and the internal valve surfaces. The same moisture can freeze at the expansion valve and choke the flow, so cooling fades in and out before anything fails outright.
The damage from a poor vacuum arrives long after the installer has been paid. Nothing about it points back to install day, so a compressor that gives up early gets written off as bad luck or a bad batch. Air left in the circuit adds a second, quieter penalty. It does not condense, so it raises the pressure the compressor has to push against, and the system draws more power for the same cooling.
How much moisture gets in depends on how the pipe was handled before the vacuum ever started. Copper delivered to site is capped for a reason. A run left open overnight between the piping day and the commissioning day collects far more moisture than a run sealed at both ends. Rain on an exposed ledge, a wet screed floor, and an afternoon of open flare ends all load the same circuit. The vacuum has to pull all of it back out, which is why a rushed pull on a badly protected run is worse than a rushed pull on a clean one.
A vacuum is a reading, not a judgement call
A vacuum is measured, and the measurement is the only proof it worked. The needle gauge on a manifold set is too coarse to show the last and most important part of the pull. A proper vacuum is read on a micron gauge that resolves what is actually left inside the circuit. Running the pump for a while and listening for the tone to change is not a measurement.
The decisive step is what happens after the pump is valved off. A circuit that is dry and sealed holds its reading once it is isolated. One that climbs back up is telling you something is still in there, and there are only two candidates. Either moisture is still boiling off, or air is coming in through a joint that does not hold. Both answers change what should happen next, which is the whole reason the standing test exists.
Ask what reading the vacuum reached, and whether it held after the pump was isolated. An installer who did the work answers with a number and a behaviour, because a number and a behaviour are what they watched on the gauge. Pulling down below 500 microns and then watching for a rise is common practice, so an answer in that shape is a good sign. An answer built around how long the pump was left running describes effort rather than result.
Where a vacuum goes wrong even when a pump was used
Owning a pump and using it well are different things, and the gap between them is where most weak vacuums live. A pump pulls hard at its own inlet. What matters is the reading at the system, and everything between the two can throttle it. A thin service hose, a long hose run, and valve cores left sitting in the service ports all restrict the path. The pump then reads deep while the circuit behind it is still wet.
Pump oil is the other quiet variable. The oil absorbs the moisture being pulled out of the pipes, and once it is loaded it stops being able to reach a deep vacuum at all. On a pump that has cleared several damp systems without an oil change, the gauge stalls at a level that looks acceptable and goes no further. None of this is visible to you on the day, which is the argument for asking about the reading at the system rather than about the equipment in the van.
The nitrogen hold, and what it proves that soap cannot
The pressure test comes first, at the point where the joints are the only variable in the system. The sealed circuit is filled with dry nitrogen, brought up to the pressure the system is rated for, and then left standing with a gauge on it. A steady gauge means the joints hold. A falling gauge means one of them does not, and the bad joint gets found while the pipe is still open and reachable.
Nitrogen is used because it is dry and inert. It carries no moisture into the pipes, it does not react with the oil, and it costs a fraction of what refrigerant costs to vent when a test fails. Testing with refrigerant instead means releasing the very thing the circuit is meant to hold, every time a joint turns out to be bad. That gives an installer a quiet reason to skip the test and hope, which is exactly the outcome the nitrogen step removes.
Watch for the answer that leaks were checked by brushing soap on the joints. Soap finds a fast leak at a joint someone can reach. It finds nothing on a slow leak, and nothing at all on a joint sitting inside a wall or above a ceiling. A held pressure reading covers the entire circuit at once, including the parts nobody can get to afterwards. Ambient temperature moves the gauge as the day warms, so a genuine test accounts for that before anyone calls it a pass.
A failed test is not a disaster, and it is worth saying so before you ask about it. It is the test doing exactly what it was there for, at the one point in the job where a bad joint is still cheap to reach. What you want to hear is that the joint gets traced and remade, then the circuit gets brought back up and held again, before anything else moves forward. An installer who treats a fail as normal is describing a process. One who has never had a fail is usually describing a step that does not happen.
The stakes change on a replacement that reuses the existing copper. Older pipe was often run for R22, and R32 works at higher pressures, so joints that were fine on the old system are being asked to do something new. Age, previous flares, and whatever the pipe has been through since it was installed all matter here. A nitrogen hold on reused pipe is the only honest way to find out whether the old run is still fit for the new system. It is also the test most likely to be waved away as unnecessary, on the grounds that the pipe was already there.
Why a long concealed run raises what is at stake
Both steps matter more as the pipe run gets longer and harder to reach. A longer run holds more air and offers more internal surface for moisture to settle on, so it takes a deeper and steadier pull to clear it. It also carries more joints, and every joint is one more place the nitrogen test has to prove before the gas goes anywhere near it.
In most Singapore condos the piping leaves the aircon ledge and runs through a false ceiling or a concealed chase to each room. Once that ceiling is closed and skimmed, a weeping joint that would have taken a few minutes to remake on open pipe becomes carpentry, repainting, and a return visit. HDB layouts are usually shorter and more exposed, which is part of why installers can price them predictably.
These are the situations where the two steps carry the most weight, and where they are worth confirming in writing before you sign:
- Piping concealed above a false ceiling or inside a chase, where any later repair means opening finished work
- Long runs from a shared aircon ledge to bedrooms at the far end of the flat
- Existing copper reused on a replacement system, where old joints meet the higher pressures of R32 for the first time
- Multi-split systems, where one condenser feeds several indoor units and the joint count multiplies
- Any run the installer cannot show you once handover is done
If the installation is already finished
A vacuum cannot be checked after the fact, and anyone who says otherwise is guessing. Once refrigerant is in the pipes, the evidence of how well the circuit was emptied is gone, and what remains is a system that either behaves normally or does not. That limit is worth knowing plainly, because it stops you paying for a verification that cannot be done.
What can still be checked is whether the circuit holds and how the system is running. Pressures and temperatures taken at the outdoor unit show whether the charge is behaving as it should. A circuit that is slowly losing gas declares itself in those readings well before it declares itself in the room. A proper pressure test remains available on a finished system, though it means recovering the refrigerant first. That makes it a sensible move when a leak is already suspected, and an unnecessary one when the system is cooling well.
Where the install is recent and still inside the workmanship cover, the more useful move is documentation rather than testing. Ask the installer, in writing, what the circuit was held at and what the vacuum reached. The answer, or the absence of one, tells you where you stand long before a fault appears, and it is a far easier conversation while the relationship is still open.
What to ask, and what a straight answer sounds like
Ask before the work starts rather than after. Both steps are cheap to include when they are part of the plan, and awkward to add once the installer is standing in your flat with the gas bottle open. Raising it early also tells the installer what kind of handover you expect, which tends to change how the rest of the job is run.
You are not testing the installer's knowledge. You are listening for whether the answer contains something that was measured. The table below sets out the questions worth asking, what an answer looks like when the work was actually done, and what a vague reply is telling you underneath.
None of these questions require you to understand the equipment. They only require the installer to answer with something that was read off a gauge, rather than a description of how the job usually goes. Most installers who work this way are glad to be asked. It is the part of the job a cheaper quote leaves out, and the part that never shows up in a photo of a neat trunking run.
| What to ask | A direct answer sounds like | What a vague answer is telling you |
|---|---|---|
| What to askWill the pipework be pressure tested with nitrogen before charging? | A direct answer sounds likeYes, held at the pressure the system is rated for, with a gauge left on it and the reading shown to you | What a vague answer is telling youA reply that the joints get checked usually means soap, which catches a fast leak at a reachable joint and misses everything else |
| What to askWhat vacuum reading are you pulling down to? | A direct answer sounds likeA figure in microns, read on a micron gauge rather than the manifold needle | What a vague answer is telling youAn answer measured in pump running time describes effort. It says nothing about what was left inside the pipes |
| What to askWill the vacuum be held after the pump is isolated? | A direct answer sounds likeYes, the pump gets valved off and the gauge is watched for a rise before charging starts | What a vague answer is telling youWithout a standing test, leftover moisture and a slow leak both pass unnoticed and reach the compressor |
| What to askWhat happens if the pressure test fails? | A direct answer sounds likeThe joint gets traced and remade before anything is charged, and you are told before that work starts | What a vague answer is telling youA quote with no answer here has not priced the possibility of a failed test, so the cost lands somewhere later |
| What to askCan I see the gauge at each stage? | A direct answer sounds likeYes, in person or as a photo of the gauge taken at the pressure test and at the vacuum | What a vague answer is telling youReadings nobody recorded cannot be produced afterwards, which is worth knowing before the ceiling is closed |
Ready to get started?
Tell us what’s going on. Symptoms, setup, photos, anything we should know. We’ll assess and come back with the right next step.