Spa Center Acoustics: The Noise Numbers That Decide It

Everything below follows from those 3 figures. Partition build-up, door seals, glass thickness, pipe velocity, duct velocity, spring deflection under a pump and absorbent area in a tiled room are all consequences of the gap between what a room generates and what it is allowed to hold.

Table of contents

• Why does noise decide how a spa center is used?
• What is the difference between airborne sound and structure-borne sound?
• How does flanking transmission defeat a wall that tested well?
• What background noise level should each room in a spa center hold?
• How does an NR curve relate to a single dB(A) reading?
• What does the mass law predict for a partition?
• What sound reduction index does a treatment room wall need?
• How much do resilient bars, cavity depth and mineral wool add?
• Why is the door the weakest part of any partition?
• What does glass do to a sauna or steam room wall?
• How is pump and fan vibration stopped at the mount?
• What water velocity is quiet, and what is water hammer worth in bar?
• What duct and grille velocities keep an air handling unit inaudible?
• What noise does a steam generator make that nothing else does?
• What sounds does a sauna cabin make on its own?
• Why does a tiled wet room ring, and how long is its reverberation?
• Which absorbers survive a wet area, and why is open-cell foam ruled out?
• How is acoustic performance in a spa center measured and signed off?
• What does this article not decide about a spa center?

Because the quietest room sets the standard and the loudest machine sits about 12 m away from it. A spa center asks a resting body to stay still for 30 to 60 minutes in a room held near 30 dB(A), while a plant room 1 wall away runs a pump at 75 to 85 dB(A) continuously.

That is a gap of 45 to 55 dB across a building whose floors, walls and pipes are mostly hard, heavy and rigidly connected. No other room type in commercial construction asks for that spread in that little space.

The Google Search Console record for this site shows 494 impressions and 0 clicks on the query spa center over 16 months at an average position of 29.5, and almost none of the competing pages publish a single decibel figure.

Airborne sound starts in air and has to push a wall into motion to cross it, losing 40 to 55 dB on the way. Structure-borne sound starts in the building itself, travels in steel and concrete at about 3,500 to 5,000 m/s, and radiates back into air 3 rooms away having lost almost nothing.

The distinction decides what the fix is. A voice in a corridor is airborne and is stopped by mass, cavity and seals. A pump bolted to a slab is structure-borne, and adding a second layer of board to the wall beside it changes the result by roughly nothing, because the sound is arriving through the floor rather than through the wall.

Almost every acoustic complaint we are called in to diagnose in a working commercial spa is of the second kind. The partitions were built correctly and the machines were hung wrongly, so the measured airborne performance is fine and the building still transmits every pump start.

By going round it. A partition rated Rw 55 dB in a laboratory routinely measures R’w 45 to 50 dB on site, because sound also travels through the floor slab under it, the ceiling void over it, the continuous screed, and every pipe and duct that crosses it. The weakest of those paths sets the result.

The arithmetic of parallel paths is brutal and it is the same arithmetic as a leaking door. Sound power adds, so the total transmission coefficient is the area-weighted sum of every path. One flanking route at the equivalent of 40 dB will hold a 55 dB wall down to about 41 dB no matter how many layers of board go on the face.

Service penetrations are the flanking path that is easiest to measure and most often missed. A single unsealed 30 mm sleeve through a partition is an open hole, and an open area of 1% of the wall caps the whole partition at 20 dB, because a transmission coefficient of 0.01 gives a sound reduction index of exactly 10 log (1/0.01), which is 20. Every sleeve is packed with mineral wool and sealed both sides with a flexible sealant.

Between NR 25 and NR 45 depending on what happens in the room, which is a spread of 20 dB across a single floor plate. A resting area sits at NR 25, a treatment room at NR 25 to 30, a changing area at NR 35 to 40, a wet area at NR 40 and a fitness area at NR 40 to 45.

The table is also the partition schedule in disguise. Subtract the target of the quiet room from the source level of the room beside it and the difference is the sound reduction index the wall between them has to deliver, before flanking is allowed for. A plant room at 80 dB(A) beside a resting area at 30 dB(A) is asking for 50 dB of it.

An NR curve is a set of 8 octave-band limits from 63 Hz to 8 kHz, and the rating is the highest curve the measured spectrum touches. For the broadband spectra that building services produce, the dB(A) value lands roughly 5 to 6 dB above the NR number, so NR 25 reads as about 30 dB(A).

The curve exists because a single A-weighted figure hides low-frequency noise. A-weighting discounts 63 Hz by about 26 dB, so a fan whose energy is almost all below 125 Hz can measure a comfortable 32 dB(A) and still be clearly audible as a hum in a room where nothing else is happening. The octave-band curve catches that; the single number does not.

That the sound reduction index of a single limp panel rises by about 6 dB every time its surface mass doubles, and by about 6 dB every time frequency doubles. The working form is R = 20 log (m f) − 47 dB, with m in kg/m² and f in Hz, and it is the floor under every partition decision.

Put numbers through it. Two layers of 12.5 mm plasterboard weigh about 20 kg/m² together, so at 500 Hz the prediction is 20 log (20 × 500) − 47, which is 80 − 47, or 33 dB. A plastered 100 mm dense concrete block wall at about 200 kg/m² gives 20 log (100,000) − 47, or 53 dB, from mass alone.

Two limits sit on top of the prediction. Every panel has a coincidence frequency where it goes transparent, roughly 12,700 divided by thickness in millimetres for glass, so 8 mm glass dips near 1,590 Hz. And no partition performs above what its weakest path allows, which is why the mass law is a starting point rather than an answer.

Rw 50 to 55 dB between 2 treatment rooms, and Rw 55 to 60 dB between a treatment room and a plant room, corridor or fitness area. Those are laboratory figures; the site requirement is written as R’w and sits 5 to 10 dB lower, so a laboratory specification of 55 dB is asked to deliver 45 to 50 dB in place.

The number comes from subtraction rather than from habit. Normal speech at 1 m is about 60 dB(A), raised speech about 66 dB(A), and a treatment room specified at NR 25 holds about 30 dB(A) of background. Speech is inaudible when it arrives below the background, so 66 minus 30 gives 36 dB as a bare minimum, and 50 dB gives the margin that makes a conversation not merely inaudible but unrecoverable.

Plant separation works the same way from a much worse starting point. A pump room at 80 dB(A) next to a resting area at 30 dB(A) needs 50 dB of on-site performance, which means a laboratory rating of 55 to 60 dB, a sealed and rated door, and no shared screed. In a hotel spa design the same wall usually separates the spa from guest bedrooms as well, and the bedroom target is the stricter of the 2.

Mineral wool in the cavity is worth 5 to 8 dB, resilient bars on one face are worth 5 to 10 dB, and opening the cavity from 50 to 100 mm is worth 3 to 5 dB at mid frequencies. Together they take a 40 dB single-stud wall to 55 dB without any change in mass.

The wool works by damping the cavity resonance rather than by blocking anything. An empty air gap between 2 leaves behaves as a spring that couples them at a mass-spring-mass frequency, and unfilled it makes the partition worse than a single leaf at that frequency. Filling the cavity to about 60 to 80% of its depth with 45 kg/m³ mineral wool removes that resonance; packing it to 100% and compressing the wool starts to reconnect the leaves.

Beyond about Rw 60 dB the economics change and the answer becomes 2 fully separate structures. A thermal suite built against a fitness studio is the usual case: the studio floor is generating impact energy no partition can catch, so the wall gets a structural break rather than another layer.

Because it is a 2 m² hole in a 10 m² wall with a movable seal round it. Put an Rw 25 dB door into an Rw 55 dB wall and the composite performance is 32 dB, calculated from the area-weighted transmission coefficients: 2 m² at 0.00316 plus 8 m² at 0.00000316, divided by 10 m², gives 0.000635, or 32 dB.

Upgrading the door is worth more than anything done to the wall. The same wall with an Rw 35 dB acoustic doorset comes out at 42 dB, a 10 dB gain for a change to 20% of the area, while doubling the mass of the wall leaf in the same situation moves the composite by under 1 dB. The weakest path owns the result.

Gaps are worse than leaves. A 10 mm undercut under a 900 by 2,100 mm door is 0.009 m² of open area in a 1.89 m² leaf, which is 0.48% open, and that alone drags an Rw 30 dB door down to about 22 dB. Drop seals, compression seals on 3 edges and a rebated frame are not detailing preferences; they are the majority of the door’s rating.

It caps it. Monolithic glass runs Rw 31 dB at 6 mm and about 34 dB at 10 mm, an acoustic laminate at 8.8 mm reaches roughly 36 dB, and a sauna or steam room front is normally 8 mm toughened at about 32 dB. No cabin front made of glass is an acoustic barrier.

The glazed door is a deliberate exception that cannot be sealed. A sauna door opens outward and carries no latch that can hold it shut, so it is hung with clearance on every edge, and the gap round it is a design requirement rather than a defect. That single detail means a commercial sauna cabin leaks sound both ways whatever its front is made of.

The consequence is that the acoustic line moves outward. The barrier is the masonry or stud wall of the room the cabin stands in, not the cabin envelope, and that wall is where the Rw figure is specified, where the penetrations are sealed and where the door is rated. Inside a commercial steam room the same logic applies to the glass front and to the door beside it.

With spring isolators chosen by static deflection, not by look. A spring compressing 25 mm under load has a natural frequency of about 3.2 Hz, from the relationship 15.8 divided by the square root of deflection in millimetres, and a pump running at 1,450 rpm excites it at 24.2 Hz.

Isolation is then a ratio. Transmissibility is 1 divided by the square of the frequency ratio minus 1, so at 24.2 Hz over 3.2 Hz the ratio is 7.6, its square is 58, and transmissibility is 0.017, meaning about 98% of the vibration is stopped. The same pump on 5 mm rubber pads, whose natural frequency is about 7.1 Hz, gives a ratio of 3.4 and transmissibility of 0.09, or 91%.

Flexible connectors are only as good as the pipe beyond them. A braided connector at a pump outlet is short-circuited by the first rigid clip downstream, so the first 3 or 4 supports after the machine are resilient-lined and the pipe does not touch the structure until the energy has dropped. Pumps, fans and generators are manufacturers’ equipment and their own data give the operating speeds these calculations start from.

Below about 1.2 m/s in pipework crossing occupied areas, and up to 1.5 m/s elsewhere. Above 2 m/s the flow generates its own broadband hiss at every bend and valve. A fast valve closure at 1.5 m/s produces a surge of 1,000 × 1,200 × 1.5 pascals, which is 1.8 MPa, or about 18 bar.

That surge is the Joukowsky relationship: density times wave speed times the change in velocity. Wave speed in water inside a steel pipe is roughly 1,200 m/s, so the pressure spike arrives in a few milliseconds and travels the length of the pipe as a bang that every rigid clip injects straight into the structure. Halving the velocity halves the surge.

Roughly 4 to 5 m/s in main ducts, 3 m/s in branches and no more than 2 m/s at the face of a grille serving an NR 25 room. Regenerated noise at a fitting rises with about the fifth to sixth power of velocity, so halving the velocity is worth 15 to 18 dB without any attenuator at all.

That exponent is the most useful number in ventilation acoustics. Going from 4 m/s to 5 m/s at a grille, a change nobody would notice on a drawing, adds 10 log (1.25 to the sixth), or about 5.8 dB, which is the difference between NR 25 and NR 31. Oversizing the terminal is cheaper than silencing it afterwards.

Three things, on a timer rather than on demand. The fill solenoid closes with a metallic clack lasting under 100 ms, the water flashes to steam with a broadband hiss concentrated above 1 kHz, and the automatic blowdown discharges hot water into a drain on a cycle the controller sets, which is the loudest of the 3.

The blowdown is the event guests notice because it is unpredictable. It arrives as a rush lasting several seconds at a time nobody expects, and if the generator shares a wall with a resting area the difference between 30 dB(A) of background and a 60 dB(A) discharge is the whole complaint. It is solved by location before it is solved by lagging.

The generator itself belongs in a ventilated plant space with a lobby, not in a ceiling void above an occupied room. Steam generators and their controllers are manufacturers’ products; any CE marking or EN 60335 conformity on that equipment belongs to its manufacturer, and their data sheets give the blowdown interval and sound level this positioning is based on.

Four, and 3 of them are electrical. The contactor switching an 8 to 12 kW element bank clicks every few minutes as the thermostat cycles, the elements tick as they expand through 60 to 80 °C of temperature change, a ladle of water on stones produces a broadband burst of 2 to 3 seconds, and the timber creaks as it moves.

The contactor is the one worth engineering. It is a mechanical relay switching a large current, it is often mounted inside the cabin wall or immediately behind it, and a hard-mounted contactor on a stud radiates through the panelling into a room where the background is deliberately low. Mounting it in the plant space, on a resilient backing plate, removes the sound entirely.

Because glazed tile and polished marble absorb about 1 to 2% of the energy that reaches them. Take a hammam 5 m by 4 m by 3 m: the volume is 60 m³ and the surface area is 94 m², so total absorption at 0.02 is 1.88 m², and Sabine gives 0.161 × 60 ÷ 1.88, or 5.1 seconds.

Five seconds is a cathedral. Speech becomes unintelligible above roughly 1.2 seconds in a room that size, and the room measures far louder than the sum of its sources because the energy has nowhere to go. The complaint in a wet area is almost never the plant; it is the reverberation.

The target is 1.2 to 1.8 seconds, and reaching it is an area calculation. To get the same room to 1.5 seconds the total absorption has to be 0.161 × 60 ÷ 1.5, or 6.44 m², which is 4.56 m² more than the tile provides. A ceiling material absorbing 0.65 supplies that over about 7 m², which is roughly a third of the ceiling.

Micro-perforated panels and acoustic plaster, in that order. A micro-perforated sheet with holes of 0.5 to 1.0 mm at an open area of 0.5 to 2%, spaced 50 to 150 mm off the substrate, reaches an absorption coefficient of 0.6 to 0.9 across about 2 octaves with no fibrous material anywhere in the build-up.

It works as a resonator rather than as a sponge. Air forced through holes that small loses energy to viscous friction at the hole walls, and the cavity behind sets the frequency the system is tuned to, so the absorber is made of the same aluminium, glass or coated panel the rest of the room is made of and can be steam-cleaned. Cavity depth is the tuning control: deeper cavities move the peak downward.

Acoustic plaster is the alternative where a seamless surface is wanted. Applied 15 to 30 mm thick over a porous backing board it gives 0.5 to 0.7 at mid frequencies, and it accepts a curved ceiling in a way a panel system does not, which matters in a domed room.

Open-cell foam and mineral fibre are ruled out because they are porous by definition. A material that absorbs sound by letting air pass through its pores also lets water in, and once wet its pores are full and its absorption collapses towards zero while its weight climbs. It cannot be wiped down, it cannot be steam-cleaned, and it will not dry between sessions in a room held at 45 °C and near-saturated air.

With a Class 1 sound level meter, calibrated before and after each session and accepted only if the drift is under 0.5 dB, reading in octave bands from 63 Hz to 8 kHz. Three quantities are measured: background level against the NR curve, field sound insulation between rooms, and reverberation time in each hard-surfaced space.

Each has its own method. Background noise is measured with the building running and the room empty. Airborne insulation between rooms is measured to ISO 16283-1 and rated to ISO 717-1, which gives the R’w figure the specification was written in. Reverberation time is measured to ISO 3382-2, using either an interrupted noise source or an impulse, averaged over several source and microphone positions.

The survey also has to be the last thing switched on rather than the first. Plant that has not been commissioned runs at the wrong speed, and a background measurement taken against an uncommissioned air handling unit is a measurement of nothing. Anyone working through how to open a spa center should put the acoustic survey after commissioning and before the soft opening, with time left to fix what it finds.

The other invisible service in these rooms is light, and it is settled with the same kind of arithmetic. The lux levels room by room, the colour temperatures, the colour rendering floor and the measuring grid are set out in the lighting numbers behind a spa design.

Most of the building. Acoustics sets partition ratings, mount deflections in millimetres, pipe velocities in m/s and absorption areas in m², and it is silent on everything else. At least 5 other disciplines set numbers for the same rooms and several of them pull in the opposite direction to this one.

Ventilation is the clearest conflict. Air change rates are set by humidity and by occupancy, not by noise, and the duct sizing that satisfies an NR 25 room at 2 m/s face velocity is larger than the duct an air-change calculation alone would produce. The 2 requirements are resolved together or the acoustic target is quietly lost at the terminal.

Lighting, corrosion resistance, waterproofing, floor loading and fire separation are decided elsewhere and are outside this article entirely. Where they touch acoustics they do so as constraints: a fire-rated partition has its own build-up, a waterproofing layer changes what can be fixed to a wall, and a drained floor build-up changes what a resilient layer can sit on.

Sauna Dekor has built thermal facilities since 1987, is now in its 40th year, works from its own Istanbul facility with 19 employees under TS EN ISO 9001:2015, and has delivered projects in 41+ countries. On acoustics that means we set partition build-ups, door specifications and machine mounts ourselves.

What we do not make is equally clear. Steam generators, sauna heaters, control panels, pumps and extract fans are not our products; they come from the equipment manufacturers we buy from, and any CE marking or EN 60335 conformity carried by that equipment belongs to its manufacturer rather than to us. The TS EN ISO 9001:2015 certificate is ours; the sound-power data and the operating speeds in this article come from those manufacturers’ own published sheets.