How loud is air conditioning, really?

Inside: quieter than the fridge

This is the part people ask about and the part that turns out not to matter much. A modern inverter wall unit on its lowest fan speed is around 19 to 21 dB(A) at a couple of metres — below a domestic fridge, below a quiet library, and well below the level at which most people notice a sound at all in a room that already has a window in it.

Two honest caveats. That is the lowest setting — pulling a hot room down on a July afternoon runs the fan harder and it is audible, more like 35 to 40 dB(A). And a correctly sized unit spends most of its life on low, while an oversized one cycles between off and working hard, which is more noticeable than a steady quiet hum. Sizing and noise are the same conversation.

Outside: the number that decides things

The outdoor unit is a fan and a compressor, and it lives on a wall or a slab near somebody's house. It is typically 46 to 49 dB(A) at one metre — a quiet conversation, if you were standing beside it, which nobody is.

How sound behaves on the way next door

Decibels are logarithmic, which is why none of this works the way intuition suggests. Three rules cover almost everything that matters.

Every doubling of distance

−6 dBFrom a point source in the open. This is the big lever

A second identical unit

+3 dBNot double. Two 46 dB units together are 49 dB

A hard surface behind it

+3 dBA wall reflects. An internal corner reflects twice

A solid barrier in the path

−5 to −10 dBA fence or an outbuilding, if it breaks line of sight

So a 49 dB(A) unit measured at a metre is roughly 43 dB at two metres, 37 dB at four, and 31 dB at eight — before any barrier, and before the wall behind it adds some back. That is the whole reason the assessment is done at the neighbour's window rather than at the unit: the same machine passes comfortably in one position and fails in another six feet away.

The 37 dB test, in plain terms

Since 28 May 2026 an MCS 020(a) calculation is the only route to installing under permitted development in England, and since 1 June 2026 in Wales too. Scotland and Northern Ireland remain on the earlier MCS 020 at 42 dB(A) — a different threshold in a similar calculation, not a licence to be careless. It is not a measurement taken on the day: it is a calculation done beforehand, from the unit's declared sound power and the geometry of your property.

1. 1Start from the declared sound power of the specific unit — not the marketing figure, the declared one.
2. 2Apply the distance from the unit to the assessment position.
3. 3Add corrections for reflective surfaces: the wall behind it, the ground, a corner.
4. 4Subtract for any barrier that genuinely breaks the line of sight.
5. 5Compare the result against 37 dB(A) LAeq at one metre from a habitable-room door or window of a neighbouring property.

"Habitable room" is doing real work in that sentence: a bedroom or a living room counts, a garage, a bathroom or a hallway does not. And LAeq is an average over a period rather than a peak, which is why a brief noise like a defrost cycle starting does not fail an assessment on its own.

If the calculation comes out above 37 dB, the install is not banned — it just is not permitted development, so it needs a planning application like any other building work. More often the answer is to move the unit, add a barrier, or choose a quieter one, and the calculation is what tells you which of those is enough. The rest of the permitted development rules are here.

What actually makes one sound worse

Declared figures are taken under steady conditions. Four things in real life are louder than the number suggests, and none of them are faults.

• Defrost cycles. In cold damp weather the outdoor coil ices up and the unit reverses briefly to clear it. There is a change in tone and sometimes a short rush of air. It lasts a few minutes and it is the unit working correctly.
• Start-up. The compressor coming up to speed is the loudest moment of an otherwise quiet cycle.
• Very cold weather. Less ambient heat means the unit works harder for the same output, so it runs faster and louder in February than in October.
• An undersized unit, permanently. It never reaches temperature, so it never throttles back to the quiet part of its range — which is the same argument the sizing guide makes about efficiency, arriving at noise instead.

There is also a difference between loud and annoying. A steady low hum at 45 dB is easier to live beside than an intermittent one at 40, which is why a correctly sized unit that modulates gently beats an oversized one that cycles. Sizing has more effect on this than most people expect.

What we do about it

The MCS 020(a) calculation is run for every install we do, before anything is ordered, and you get the result in writing whether it passes or not — including if the answer is that your property needs a planning application. It is the evidence your permitted development rests on, so it is not a formality and it is not something to find out about afterwards.

The siting decision is made with that calculation in hand rather than after it. Where the outdoor unit stands relative to your neighbour's windows is a design decision, and it is one of the few on a job like this that cannot be revisited cheaply once the pipework is in.

If you want the numbers for your own room, the quote flow sizes it in about a minute and shows its working. The noise calculation itself needs your address and the position of the nearest neighbouring windows, so it happens at survey rather than online — and it is included in the price, not an extra.