Are air-to-air heat pumps noisy? Real decibel figures, the 37 dB(A) rule, and how to keep the neighbours happy

Not very, if the installer does the job properly. The indoor unit of a modern wall split runs at about 19 dB(A) on its lowest fan speed, which is quieter than a whisper across the room. The outdoor unit is the part that matters for heat pump noise: a typical 2.5 to 3.5 kW unit is quoted at 45 to 50 dB(A) sound pressure at 1m, and about 58 to 61 dB(A) sound power. That is a quiet office when you stand next to it, and a hum you have to listen for at a neighbour's window a few metres away.

The rule that decides whether your unit can go in without planning permission is the MCS 020 a) calculation: 37.0 dB(A) or less at 1m from the nearest neighbour's habitable-room window. This guide gives the real figures, walks through the 37 dB(A) test, and covers the night-time problems that actually generate complaints. It is about air-to-air heat pumps, the reverse-cycle splits that heat and cool a room.

Real decibel figures for air-to-air units

The indoor figure tells you whether you can sleep with it running. The outdoor figure tells you whether your neighbour can.

Indoor figures are sound pressure on the quietest fan setting in heating mode, from the manufacturers' UK product data; outdoor figures are sound pressure at 1m unless stated. Manufacturers do not all publish outdoor sound power on their consumer pages, so ask your installer for the databook figure, which is the one the MCS 020 a) calculation uses.

Every premium wall split at the popular 2.5 and 3.5 kW sizes now hits 19 dB(A) indoors on its lowest setting, which is why these ranges lead our bedroom air conditioning guide.

Bigger units are louder. Panasonic's figures climb from 19 dB(A) on the Z35 to 30 dB(A) on the Z50 indoors, and from 45 to 52 dB(A) outdoors. If you only need a 2.5 kW unit, do not let anyone sell you a 5 kW one "to be safe".

And heating is louder than cooling on the same outdoor unit: Panasonic's Z35 is 48 dB(A) cooling and 50 dB(A) heating, and Daikin's databook shows the same gap. More on why below.

Sound power vs sound pressure: why brochures disagree

This is the biggest cause of confusion about air conditioning unit noise, and why one website says a unit is "48 dB" and another says "61 dB".

Sound pressure is what a microphone measures at a spot, usually 1m from the unit under test conditions. It describes what you would hear standing next to it. Sound power is the total acoustic energy the unit gives off, wherever you stand. It is always the bigger number, typically 10 to 13 dB(A) above the 1m sound pressure figure for a small outdoor unit, and it is the figure on the energy label and the one MCS uses in the planning calculation.

Neither is wrong; they answer different questions. Compare like with like, and treat any brochure with a single unlabelled "dB" figure with suspicion. Our glossary defines both terms and LAeq.

The usual comparison points: 20 to 30 dB(A) is a whisper or a quiet bedroom at night, 40 dB(A) a library, 50 dB(A) a quiet office, and 60 dB(A) normal conversation at 1m. Every doubling of distance from a small source drops the sound pressure by about 6 dB, and a 10 dB change sounds roughly twice or half as loud.

The 37 dB(A) rule: heat pump noise regulations in the UK

In England, an air-to-air unit that both heats and cools can be permitted development under Class G, provided it meets the conditions. The noise condition is compliance with the MCS Planning Standards, and since the transitional period in SI 2025/560 ended on 28 May 2026, MCS 020 a) is the only accepted route. A cooling-only unit is not permitted development at all and needs a planning application. The other Class G conditions (1.5 m³ per unit, two units on a detached house, one on a semi, terrace or flat, listed buildings and conservation areas) are in our air conditioning planning permission guide. Wales adopted the same standard from 1 June 2026; Scotland and Northern Ireland differ, so always check your local planning authority.

MCS 020 a) sets one number: 37 dB LAeq,5mins at every assessment position, "ignoring the effect of the façade". LAeq,5mins is the A-weighted average over five minutes, so a brief defrost burst does not sink you but steady running does. The assessment position is 1m outside the centre of any door or window to a habitable room of a neighbouring property: bedrooms and living rooms, not bathrooms, hallways or utility rooms. The limit applies whether or not your installer is MCS certified, and passing it does not on its own give you permitted development rights.

How the assessment works

Nobody turns up with a meter. It is a desk calculation from four inputs:

1. Sound power of the outdoor unit, from the manufacturer's declared figure. The standard says the "low noise mode" figure shall not be used, so quiet mode cannot rescue a borderline site.

2. Reflecting surfaces within 1m of the unit: Q2 for ground only, Q4 for ground plus the wall it hangs on (the normal case), Q8 for a corner. More than three surfaces cannot comply.

3. Distance in a straight line from the centre of the unit to the assessment position.

4. Barriers. A masonry wall or solid fence at least 18mm thick, gap-free and extending 1m beyond the unit each side, earns minus 10 dB if it fully blocks the line of sight, minus 5 dB if partly. A thinner solid fence earns minus 5 and minus 2.5 dB. Hedges and gappy panels earn nothing.

There is no background-noise correction, which is the main difference from the old 42 dB(A) version. A quiet cul-de-sac gets no allowance for being quiet.

Will one outdoor unit pass?

Run a 60 dB(A) sound power unit on a wall (Q4) with no barrier through the MCS formula and the level at the neighbour's window comes out at about 45.5 dB(A) if that window is 3m away, 41 dB(A) at 5m, 37.0 dB(A) at 8m and 35 dB(A) at 10m. A 58 dB(A) unit passes at about 7m.

That sounds tough, but the distance is to the neighbour's window, not the fence. On a typical semi, a unit 2 to 3m from the boundary is often 7 to 10m from the neighbour's nearest bedroom or living-room window, and it passes. A solid 1.8m close-board fence that blocks the line of sight gets a 60 dB(A) unit through at 5m; a brick wall that hides it, at 3m.

Where it fails is predictable: a narrow gap between terraces with a bedroom window 3 to 4m away, a unit fixed high on a wall that no fence can screen, a corner with three reflecting surfaces, or a 5 to 7 kW unit with a sound power in the mid 60s. The fix is a quieter unit, a different wall, a proper barrier or enclosure, or a planning application.

Passing the 37 dB(A) test is not a defence against a nuisance complaint. Councils must investigate under the Environmental Protection Act 1990 regardless of planning status.

If a unit "unreasonably and substantially" interferes with a neighbour's enjoyment of their home, the council's environmental health team must investigate and, if it agrees, serve an abatement notice; ignoring one means prosecution and fines. A unit can pass MCS 020 a) and still be a nuisance if it rattles on a bad bracket under a bedroom window all night.

Heat pump noise at night and in cold weather

The outdoor unit is not one fixed number. In heating mode on a cold night it works hardest: the compressor and fan speed up to pull heat from air at 0°C, which is why manufacturers' heating figures sit one to two decibels above their cooling figures, and every so often the unit reverses for a minute or two to defrost the coil, with a whoosh and a change in pitch. Our guide to air-to-air heat pumps in cold weather explains the defrost cycle.

Most premium units have an outdoor quiet or night mode that caps fan and compressor speed: Daikin's Silent function cuts the outdoor unit by about 3 dB(A), and Toshiba quotes the Haori's outdoor unit down to 39 dB(A). The trade-off is output, so put it on a timer for bedtime hours rather than leaving it on permanently.

The practical fixes are about placement and mounting:

• Keep the outdoor unit away from bedroom windows, yours and the neighbours'. A side wall facing a road often beats the back wall facing three gardens.

• Never in a narrow alley between houses. Every hard surface within 1m makes the unit louder at the window opposite.

• Use proper anti-vibration mounts on wall brackets and ground feet. Many complaints are about a low hum through the brickwork, not airborne noise.

• Screen it with something solid. A gap-free close-board fence or a masonry wall counts in the calculation; a hedge does not.

• Consider an acoustic enclosure on a tight site, but only one designed for the airflow the unit needs, or it runs harder. Keep the coil clean for the same reason.

Are air-to-water heat pumps louder?

Usually a little, yes. An air-to-water unit moves enough heat for a whole house through radiators, so it uses a larger fan and heat exchanger and runs for long stretches on winter nights. A single air-to-air split heats one room with a smaller fan doing less work. The MCS 020 a) test is the same for both, but the margin is narrower on a tight plot.

Figures correct as of September 2026 from manufacturers' UK specifications and MCS 020 a) Issue 1.0. Check the exact model's spec sheet before you buy.

Our wall-mounted split system reviews list indoor and outdoor noise figures for every unit we cover, and the heating guides hub has the rest of the series.