Heat pumps · Chillers · Sludge drying systems
Heat pump · Hot water · Heating · Cooling

Heat-Pump Noise: Source, Spectrum and Fixes That Work at Night

Most heat-pump noise complaints arrive at night. During the day, ambient sound in a residential courtyard may be 60 dB(A) and the unit is masked. At night the background falls to 35–40 dB(A), and the same unit becomes intrusive. The common response — wrap the casing in acoustic wool — often changes nothing, because the dominant path is structure-borne vibration through the roof slab, not airborne sound from the casing.

The correct order is: measure the sound pressure level and spectrum; identify the dominant frequency; match it to a component (compressor fundamental, fan blade-pass frequency, pump speed); then treat source, transmission path and receiver separately. The figures below are typical values for air-to-water units and are intended for diagnosis, not as a substitute for a site measurement.

Common questions

What spectrum should I measure?

Measure both A-weighted and low-frequency levels. A fan tone and a compressor tone need different fixes, and an A-weighted number alone hides the low-frequency component.

What path?

Source, structure and air. Measure at the source and the receiver to find the dominant path.

What night limit applies?

Check the local night limit at the nearest boundary. A unit that passes in the daytime can fail at night, when the background drops and the limit tightens.

Do I need an acoustic enclosure?

Only if airflow and service access are designed in. A sealed box overheats and trips the unit. An enclosure must be part of the acoustic design, not an afterthought.

What fan?

Low-speed fans and a larger coil reduce the tone. A speed change can move the tone, not remove it.

How do I control vibration?

Check the mounts, the pipe bridges and the structural connections. Structure-borne noise travels farther than airborne noise; isolators alone may not fix it.

Will a barrier help?

A mass barrier or a screen that breaks the line of sight helps, especially close to the source. A barrier far away or with gaps does little.

What should I check after installation?

Measure again at night, at the boundary, with the unit at full load. A design claim is not a commissioned result.

1. Four sources with different spectra

Compressor mechanical noise. The fundamental follows shaft speed. A two-pole motor on a 50 Hz supply runs at about 2,900 rpm, giving a mechanical fundamental near 48 Hz, with harmonics at 97 Hz and 145 Hz. Low frequencies penetrate walls and travel through structures; this is the “hum” that is hardest to remove.

Fan aerodynamic noise. The blade-pass frequency is speed in revolutions per second multiplied by the number of blades. At 900 rpm with four blades, 900 ÷ 60 × 4 = 60 Hz; at 700 rpm with three blades, 35 Hz. Broadband turbulence noise sits above this, heard as “whoosh” close to the unit.

Water and pipe noise. Pump speed frequency is usually 24–50 Hz. Valve throttling adds broadband hiss at 500–2,000 Hz. It becomes noticeable above 1.5 m/s pipe velocity and is worse at bends, tees and reducers.

Structure-borne sound. Compressor and fan vibration passes through the feet into the roof, then along beams and columns into rooms below. The signature is telling: standing beside the unit it does not sound loud, but the bedroom below feels a low throb. A sound level meter at the unit may show compliance while the room is unacceptable.

Source Typical frequency Dominant path Primary measure
Compressor 48 Hz fundamental, 97/145 Hz harmonics Structure and casing Isolation, inertia base, enclosure
Fan 35–60 Hz blade pass plus broadband Air Silencer, blade design, low night speed
Pump and valves 24–50 Hz; 500–2,000 Hz hiss Water pipe Lower velocity, flexible joints, isolation
Structure Below 50 Hz in rooms Roof, beams, columns Break the vibration path at the unit

2. Quantify before treating

Addition. Two equal sources add 3 dB; three add 4.8 dB; four add 6 dB; ten add 10 dB. Ten units at 65 dB(A) each produce about 75 dB(A). Reducing one unit by 3 dB changes the total by about 0.4 dB, which is inaudible. On a roof with multiple units, treat the group, not one machine.

Distance. In a free field, a point source loses 6 dB per doubling of distance. A unit at 70 dB(A) at 1 m gives 52 dB(A) at 8 m and 46 dB(A) at 16 m. A roof is not a free field: the surface reflects, a wall close to the unit adds 3–6 dB, and a corner can add up to 9 dB. That is why the same model is quiet on one roof and noisy on another.

Worked example. Four units on a hotel roof, each 68 dB(A) at 1 m, 18 m from the nearest residential wall. Four units add 6 dB, giving about 74 dB(A) at 1 m. Distance attenuation is 20 log(18/1) ≈ 25 dB, so the free-field level at the wall is about 49 dB(A). With roof reflection and wall orientation, expect 52–55 dB(A).

Under GB 12348-2008, a class 2 residential area has a night boundary limit of 50 dB(A) and a day limit of 60 dB(A). The example exceeds the night limit by 2–5 dB — the classic “slightly over but strongly complained about” case. When the low-frequency content is high, the A-weighted level can comply while the room still hears a hum. In that case, work backward from an indoor target of about 30 dB(A).

Spectrum. A phone sound-level app gives only the overall level. Use a meter with one-third-octave bands: measure at 1 m at 45° to the unit, and again inside the affected room. If the peak frequencies match, the path is the same. As a guide, peaks below 50 Hz are usually structure-borne; 60–500 Hz are compressor and fan fundamentals; above 1,000 Hz are air and valve noise.

3. Fixes, source by source

Isolation first. Fit spring isolators or rubber pads between the unit base and the support. Two parameters matter: the natural frequency of the isolator should be below one-third of the disturbing frequency (for a 48 Hz compressor fundamental, 5–8 Hz); and the static deflection should be 5–10 mm for rubber or 25–40 mm for springs. Mount the unit on an independent inertia block weighing 1.5–2 times the unit, with the isolators under the block. Isolation efficiency above 90% is achievable.

Use flexible couplings on the water and refrigerant connections, rubber pads under pipe supports, and a 20 mm gap at wall penetrations filled with mineral wool. This is the item most often omitted, and it is the one that decides whether the low-frequency path is broken at the mechanical room.

Silencers for air noise. Fit reactive-absorptive silencers on the fan inlet and discharge, 600–900 mm long, with an insertion loss of 10–15 dB(A) and a pressure penalty of 30–60 Pa. Match the silencer to the spectrum: reactive chambers for low frequencies, absorptive baffles for mid and high frequencies, or a combined unit. A steel box lined with foam is almost useless at 48 Hz.

Enclosures as a last resort. A full enclosure gives 15–25 dB(A) insertion loss but restricts cooling and access. Provide inlet and outlet openings with acoustic louvers, size ventilation for the heat rejection (roughly 0.05–0.08 m³/s per kW of cooling), and keep the internal temperature below 45 °C. Use 1.5 mm steel panels with 50 mm absorptive lining, seal the joints and make the compressor access door removable.

Location and operation. Keep the unit away from bedroom windows, preferably behind a parapet or on the service side of the building. For multiple units, a shared barrier is cheaper than individual enclosures. At night, lower the water set point by 3–5 °C and preheat earlier so the unit can reduce speed or stop after 23:00. Inverter units can have a night maximum-speed limit, typically worth 3–5 dB(A).

Measure Key parameter Expected effect Check
Spring or rubber isolator Natural frequency 5–8 Hz Structure-borne reduction, often 10–20 dB Manufacturer test report; installed deflection
Inertia block 1.5–2 × unit mass Stabilizes isolation at low frequency Block weight and fixing
Flexible pipe joint All water and refrigerant connections Breaks pipe-borne vibration Visual inspection; no rigid bypass
Silencer 600–900 mm; 10–15 dB(A) Fan noise reduction 1 m measurement before and after
Enclosure 0.05–0.08 m³/s per kW; below 45 °C 15–25 dB(A) Internal temperature and current
Night set point 3–5 °C lower; speed limit 3–5 dB(A) Operating log and sound check

4. Errors and boundaries

Acoustic wool without isolation. If the floor beneath the unit vibrates under your hand, the path is structure-borne; wrapping the casing will barely change the room level. Isolate first, then absorb.

Silencer chosen without a spectrum. A purely absorptive silencer gives only 3–5 dB at 50 Hz. A reactive silencer is equally ineffective at 2,000 Hz. Know the dominant band before buying.

Ignoring addition. Treating one of ten units changes the total by less than 1 dB. Design for the group.

An enclosure that cooks the unit. Insufficient ventilation raises the condensing temperature, the unit derates and capacity falls 10–15% in summer. The owner then blames the heat pump rather than the enclosure.

When equipment-side measures cannot win. If the unit is within 8 m of a dwelling and faces the windows, relocation or a ground-floor mechanical room is usually the only reliable answer. If the complaint is dominated by energy below 63 Hz, the building structure is part of the path; equipment-side treatment alone has limited effect.

Pipe velocity and the night boundary

Two numbers are often left out of the noise package: water velocity and the boundary limit. Both are cheap to fix at design stage and expensive to fix after the plant is running.

Item Limit How to check
Water pipe velocity Not above 1.5 m/s Flow divided by internal area; above this, valve and elbow noise rises sharply in the 500–2,000 Hz band
Night boundary, class 2 50 dB(A) GB 12348-2008; measure at the boundary, not at 1 m from the unit
Measurement conditions Dry, low wind Avoid rain, snow and strong wind; record background before and after

Water velocity above 1.5 m/s turns ordinary fittings into noise sources. A DN50 pipe at 3 m/s carries the same flow as DN65 at 1.8 m/s but produces a much higher hiss at the far end. If the client hears the noise at a bedroom wall, the pipe run and the supports are part of the path; replacing the pump with a quieter model will not fix it.

The 50 dB(A) night limit is for class 2 areas. A site in class 1 or near a hospital may have a lower limit, and an industrial zone may have a higher one. Check the local acoustic function area before promising a number. Measure at the boundary at the time of the complaint, not at midday: background is lower at night, so the same source reads higher.

Treat the water side and the air side as one system. Flexible joints at the pump, rubber pads under the brackets and a low night speed on the fan are usually more effective than an enclosure. If an enclosure is used, keep the internal temperature below 45 °C; otherwise the unit derates and the noise fix becomes an efficiency problem. Acoustic options for these units are summarized on the air-source heat pump page.

Related reading

Heat Pump Noise Control · Heat Pump Maintenance Schedule · Heat Pump Commissioning Checklist