What Sound Actually Is, and Why That Decides Everything

Why this matters

A lot of noise-control money in this trade goes to a path that was never carrying the problem. A shop hangs absorptive panels in a mechanical room because a tenant two floors down hears a hum, and the tenant hears exactly the same hum afterward. That is not a bad product or a bad installer. It is a category error that starts one step earlier, at the question nobody asked: which medium is actually carrying the energy from the machine to the ear. Answer that first and the shortlist of things that will work writes itself. Skip it and you can spend a whole budget treating a path that was never in play.

A pressure fluctuation riding on static pressure

Airborne sound is a small, fast fluctuation in pressure on top of the static atmospheric pressure that is already there. Nothing travels from the source to your ear except the disturbance; the air molecules oscillate about where they already were. That is why sound needs a medium and why a vacuum carries none of it.

The fluctuations are tiny. The reference pressure for airborne sound level, 20 micropascals, is roughly two ten-billionths of sea-level atmospheric pressure, and it is the anchor of every pressure level you will ever read. A level of 94 dB re 20 micropascals, which is a loud shop, corresponds to a pressure swing of about 1 pascal, or roughly one hundred-thousandth of atmospheric. The 140 dB peak figure in the footnote to Table G-16 at 29 CFR 1910.95(a) for impulsive noise, which the section states should not be exceeded, is about 200 pascals, still only about two tenths of one percent of atmospheric. The whole audible world lives in a rounding error on the barometer, which is exactly why a level is expressed as a ratio to a reference rather than as a pressure.

Speed belongs to the medium, not to the source

The speed of a sound wave is a property of what it is travelling through, not of the machine that made it. In air at about 20 C at sea level, that speed is close to 343 m/s; it rises with temperature and is essentially independent of barometric pressure. In water at about 20 C it is near 1,480 m/s. Longitudinal waves in steel run around 5,000 m/s.

That matters because the machine does not change its frequency when the path changes. A 120 Hz tone from a motor is 120 Hz in every medium it enters. What changes is the wavelength, because wavelength is speed divided by frequency:

• 120 Hz in air, at 343 m/s: about 2.9 m
• 120 Hz in steel, at roughly 5,000 m/s: about 42 m

Two facts follow, and they are the ones that decide jobs. First, the structural path attenuates far less per unit distance than the airborne path does, so a hum can leave a roof curb and appear in a suite three bays away with no obvious airborne route. Second, one qualifier: the bending waves that actually make a panel or a duct wall radiate back into a room are slower than that longitudinal figure and are frequency-dependent, so use the longitudinal speed to understand why structural energy travels, not to predict how much a particular panel radiates. That second question belongs to the assembly's own test data or to an engineer.

Wavelength decides what an object does to a wave

Whether a physical object blocks, scatters or ignores a sound wave depends on the object's size compared with the wavelength, not on how solid it looks. At 343 m/s in air, a 63 Hz octave band centre is about 5.4 m long, 125 Hz is about 2.7 m, 500 Hz is about 0.69 m and 1 kHz is about 0.34 m.

A 1.2 m tall barrier is a substantial fraction of a wavelength at 1 kHz and does real work there. The same barrier is a fifth of a wavelength at 63 Hz and the wave simply diffracts around it. This is why the same screen that fixes a whining complaint does nothing for a rumbling one, and why almost every disappointing barrier job is a low-frequency job. State the frequency before you buy the object.

The two paths behave nothing alike

        ROOFTOP UNIT on a steel curb
         |                        |
   structure path airborne path
   steel curb, beams, roof deck, then
   slab, then the the ceiling
   ceiling radiates cavity air
         |                        |
         v v
       SUITE AIR at the listener's ear
   resilient isolation cuts the left path only
   cavity absorption reaches the right path only

Every control you can buy attaches to one path or the other. A resilient isolator interrupts the mechanical connection and does nothing whatever to airborne energy crossing the deck. Absorption placed in a cavity converts airborne energy to heat inside that cavity and does nothing to energy already travelling in the steel. Mass and sealing in the deck assembly resist airborne transmission and do very little for a structural connection that bypasses the assembly entirely.

Blocking and absorbing are two different jobs

Blocking (transmission loss) is about how much energy an assembly stops from getting to the other side. Absorbing is about how much energy is converted to heat rather than reflected back into the room it is already in. They are separate properties, measured by separate test methods, and buying one when you needed the other is the most common acoustic mistake in this trade.

The practical test: name the room the complaint is in. If the complainant is in the same room as the source and the problem is that the room is loud and washy, absorption is on the list. If the complainant is in a different room, absorption in the source room is nearly beside the point, because you are trying to reduce what crosses a boundary, and a lined mechanical room reduces the reverberant build-up on the source side only. Both are legitimate purchases. Neither substitutes for the other.

Worked example: a hum two floors down

A tenant reports a steady low hum in a corner office, worst at night, present whether the office air handler runs or not. A packaged unit sits on a steel curb on the roof, two floors up and about one bay over horizontally.

Step 1, name the frequency before naming a cure. A narrow-band or octave-band reading at the complaint position shows the energy concentrated in the 125 Hz octave band. The unit's compressor is driven at line frequency, so 120 Hz is a plausible mechanical origin, and 125 Hz is the band that contains it.

Step 2, compute the wavelengths that follow, holding the medium's speed at its stated condition. At 343 m/s in air at about 20 C, 120 Hz is about 2.9 m. In steel at roughly 5,000 m/s, the same tone is about 42 m. The office is about 4 m across. In air, the room is a little over one wavelength wide at this frequency; in the structure, the whole floor plate is a fraction of one wavelength, so the steel is not attenuating this tone appreciably across the distances in play.

Step 3, separate the paths by a test, not by argument. With the roof unit locked out and its stored energy relieved before anyone opens anything, the hum stops, which only tells you the unit is the source, not which path it uses. The useful test is selective: run the unit's fan alone with the compressor off, then reverse it. In this case the hum tracks the compressor and is unaffected by the fan, which points at a discrete mechanical excitation entering the curb rather than a broadband airborne fan signature crossing the deck.

Step 4, check the answer against the paths. If the airborne path across the roof deck were carrying it, the hum would be strongest directly beneath the unit and would fall off as you moved along the floor away from that footprint. It does not. It is as strong one bay over. That distribution is consistent with the structural path, where the 42 m wavelength in steel means the whole floor plate moves nearly together at this frequency.

Step 5, choose the control the path allows. Absorption hung in the ceiling cavity addresses the airborne term the test just showed is not dominant. Adding mass to the deck addresses the same airborne term. The path in play is the curb, so the candidate is a resilient break in the mechanical connection, specified by whoever will stand behind the deflection and the load, with the note that an isolator selected wrong can make a structural complaint worse rather than better.

Step 6, state the failure mode. A shop that skipped steps 3 and 4 buys ceiling absorption, the tenant hears no change, and the shop has now spent the budget and burned the credibility it needed to sell the isolation work. The tell is that nobody could say, before the purchase, which medium was carrying the tone.

Confirming you named the right medium

Three checks, each of which can be run before you commit money.

Break one path and watch the other. Anything that interrupts a mechanical connection without changing the airborne route, even temporarily, separates the two. If the complaint does not move, the path you interrupted was not the one.

Follow the spatial pattern. Airborne energy from a compact source falls off with distance from that source. Structure-borne energy re-radiates from large surfaces, so it tends to be roughly as loud across a whole room, and often loudest where a wall or ceiling plane is largest and thinnest.

Compare the band content to the mechanism. A discrete tone at a shaft, blade-pass or line-related frequency is a mechanical excitation. Broadband hiss is aerodynamic. Reading a spectrum back to a mechanical cause is a family of its own in this library and it starts where this article stops.

Safety notes that attach to the acts above, not to the article in general: work on a roof to reach a unit is fall-exposed, and general industry triggers fall protection at 4 feet under 29 CFR 1910.28(b)(1) while construction work triggers it at 6 feet under 29 CFR 1926.501, so name which Part your job falls under before you step out. Any lockout to stop a unit for a selective test is mechanical isolation and stored-energy control under 29 CFR 1910.147, including verifying the fan has coasted to a stop before a guard comes off; electrical work at the panel to do it falls outside 1910.147 by its own carve-out and under 29 CFR 1910.333(b)(2) instead. Do not remove a fan guard to place a microphone closer, and do not stage a shutdown of equipment serving a life-safety or process-critical load without the load owner's agreement, because a test shutdown is a process change.

References

• 29 CFR 1910.95, Occupational noise exposure, including the 140 dB peak figure for impulsive noise in the footnote to Table G-16 at (a)
• 29 CFR 1910.147 for mechanical isolation and stored energy, and 29 CFR 1910.333(b)(2) for electrical work, which 1910.147 expressly excludes
• 29 CFR 1910.28(b)(1) for general industry fall protection and 29 CFR 1926.501 for construction, whichever Part your job falls under
• Manufacturer documentation and the isolator supplier's own selection data for any resilient mounting recommendation
• See related: Reading a Vibration or Sound Signature as Its Own Diagnostic Method; What a Vibration Isolating Connection Can and Cannot Do