How long to pull a vacuum on an air conditioner – the hold test decides, not the clock
Roman Oškrobaný · · 5 min read
"How long should I pull a vacuum?" is a question every technician answers differently: 15 minutes, half an hour, an hour. The truth is that time is not the goal. The goal is a dry, tight circuit – and the clock can't show you that. The vacuum depth and the hold test after closing off the pump can. This post covers what the makers ask, why you can't trust a manifold gauge for vacuum depth, how to read the hold test and what makes evacuation quicker.
Why evacuate at all
Air in the circuit doesn't condense: it raises the condensing pressure, makes the compressor work harder and costs capacity.
Water in the circuit breaks POE oil down into acids once it is hot, and the acids attack the circuit.
The pump lowers the pressure until the water boils off at the pipe temperature. That is why the target sits far below the pressure at which water boils at the coldest point: 611 Pa at 0 °C (about 4,600 µm), about 17,500 µm at 20 °C.
What the makers ask
What the installation manuals say (a selection)
Maker
Vacuum
Time and test
Mitsubishi Electric
500 µm (−0.101 MPa)
after closing off, hold 1–2 min
Midea, Inventor
−0.1 MPa
pump 30 min (if not reached, 20 min more, then look for a leak); the needle must not move for 5 min
Daikin
−100.7 kPa (5 Torr)
check 4–5 min; if it rises, dry for 2 h more, then hold 1 h
Samsung AR50F–70F
−0.1 MPa
about 15 min; hold 2 min
Samsung AR CXKA
500 µm
4,000 → 2,000 → 500 µm with nitrogen breaks; hold 10 min
The manual for the actual unit always comes first. The vacuum calculator in the app knows the makers' rules too.
Notice that the times differ up to tenfold, and that most makers measure in MPa – that is, with a manifold gauge. And that is the problem.
A manifold gauge is not enough
500 µm is about 0.67 mbar absolute. A manifold gauge reads around −1 bar and can't tell 5,000 µm from 500 µm – to its needle both are "on zero". The "5 Torr" Daikin and Samsung give is 5,000 µm, ten times the usual trade target.
That is why you need a micron gauge (vacuum gauge). Fit it at the far end, at the unit, not on the pump: the hoses and the small service port hold back the flow, so a gauge at the pump shows a better vacuum than the circuit really has.
The hold test is the real result
Reaching 500 µm with the pump running proves little – the gauge mostly sees the pump. Close the circuit off from the pump (with a ball valve or the valve on the core tool), switch the pump off and watch what the reading does over the next minutes:
Rises a little, then stays low (under about 500 µm): the circuit is tight and dry. A small rise right after closing off is normal while the pressure evens out.
Rises, then levels off below atmospheric pressure: moisture is still boiling off. A little water levels off around 1,000–1,500 µm; a really wet circuit at water's boiling pressure, 20,000–25,000 µm at 22–27 °C. Keep evacuating.
Keeps rising in a straight line towards atmospheric pressure: a leak. Pumping longer won't help – pressurise the circuit with nitrogen and find the leak.
With a micron gauge, the ClimaToolkit app counts the circuit as done when the reading at the end is under 500 µm for at least 10 min after closing off the pump. With a manifold gauge it uses the maker's hold time. Enter the readings in the vacuum calculator and it tells you whether the circuit holds, still has moisture, or leaks.
What others accept (gauge makers and trade bodies)
Who
Pump down to
Then
JB Industries
300–400 µm
close off, wait 5–20 min: under 500 µm = done
Emerson / Alco
≈ 250 µm
a rise to ≈ 500 µm is acceptable
ACCA
500 µm
must stay under 1,000 µm for 5 min
The unit maker's manual comes first; these are the figures of gauge makers and trade bodies.
What makes it quicker
Take the valve core out. The service port's valve core blocks up to 90 % of the opening. In Appion's own test the same hose took 40 minutes with the core in and 19 with it out.
Short, wide, tight hoses. Ordinary charging hoses let air through at deep vacuum and their seals creep – it looks like a leak.
Fresh pump oil. The oil collects the water the pump pulls out. Closed off on its own, the pump should reach about 50 µm; if it doesn't, change the oil.
Gas ballast open at the start on a wet circuit, so the water doesn't condense in the oil, and closed at about 3,000 µm.
Break a wet circuit with nitrogen. After rain in the pipes or a long open job, the circuit dries much faster if you break the vacuum with dry nitrogen two or three times and evacuate again. Standing water has to be blown out with nitrogen first – a pump removes very little liquid water.
Pressure test first. A vacuum is not a tightness test. Do the nitrogen pressure test before the vacuum; the nitrogen test calculator judges it.
So how long?
Until the circuit reaches the vacuum it needs (500 µm or the maker's figure) and passes the hold test after closing off the pump. On a dry, tight split with the core out and good hoses that is often quicker than you'd expect; on a wet or cold circuit it can take hours. If the maker asks for a minimum time (Midea and Inventor 30 minutes), keep to it – but stop on the micron gauge, not the clock.
The unit converter converts vacuum units (µm, mbar, Pa, Torr, bar).
E1 or EL01 on a Midea, Vivax or Inventor split: the indoor and outdoor units can't talk to each other. How to read the code, what usually causes it and what to measure, in order.
What applies in 2027 to installing and servicing air conditioners: EU Regulation 2024/573, the Slovak Act 286/2009, the Czech Act 73/2012 and its amendment, certificates, records and bans.