We have a special prompt today from a listener named Marcus James. He just moved into a Jerusalem apartment next to a construction site in the excavation phase, and he sent us field measurements. Ninety-three point two dBA outside near the patio, sixty-eight point zero dBA inside with the balcony door closed. That's a twenty-five point two decibel drop. Sounds impressive until you realize sixty-eight dBA is louder than a vacuum cleaner and exceeds WHO daytime residential guidelines by thirteen decibels.
Modern acoustic glazing can hit STC fifty plus. So why does this apartment still feel like a construction site with the doors closed? The answer's in the frequency profile. Hydraulic rock breakers pump continuous broadband energy from a hundred and twenty-five hertz to two kilohertz — the exact band where standard double-glazing is weakest. You've got the mass-air-mass resonance on the low end and the coincidence dip on the high end, and the rock breaker sits right across both.
Marcus is living inside a physics demonstration of why windows fail at exactly the frequencies that matter most. He sent us five specific engineering questions. One — how do you estimate real attenuation from these field numbers? Two — is this airborne transmission through the glazing, or is the building itself shaking?
Three — what glazing specs actually work for this exact noise band? Four — why does hearing protection hit a physiological wall with low frequencies? And five — what psychoacoustic tricks can give you some relief tonight without touching the windows?
That last one's where it gets interesting. The answer to number four is that bone conduction puts a hard floor under what passive protection can do, and the answer to number five is a brown noise hack most people have never heard of. We're going to trace the whole chain — from the physics of the window to what you can do with a subwoofer and the right earplugs before you spend a dollar on construction.
Let's unpack what those numbers actually mean. It's not just "loud" — it's sustained loud in a frequency range human hearing is quite sensitive to. The A-weighting curve peaks around two to four kilohertz, so sixty-eight dBA of rock breaker noise is perceptually more intrusive than sixty-eight dBA of distant traffic rumble.
Marcus's spectrogram shows it's not impulsive. The hydraulic breaker is essentially a piston slamming into rock at a few hundred beats per minute, each impact exciting a whole spectrum. But because it's happening so fast and continuously, your brain doesn't get those micro-recoveries between events that make intermittent noise more tolerable.
With sustained broadband, there's no reset. The spectrogram shows energy just sitting there across a hundred and twenty-five hertz to two kilohertz, minute after minute. That's neurologically the worst profile — excavation-phase noise is uniquely punishing, and this is exactly why.
It's also the worst profile for building materials. Standard double-glazing has two failure modes that bracket this exact band. On the low end, the mass-air-mass resonance — the two glass panes and the air gap form a spring-mass system that resonates, typically between a hundred and two hundred fifty hertz. At that frequency, the window becomes transparent to sound. The panes vibrate in phase and transmission loss drops to near zero.
On the high end, the coincidence dip. At a certain frequency, the wavelength of sound in air matches the bending wavelength in the glass. The pane couples efficiently with the sound wave and radiates it into the room. For six-millimeter glass, that dip lands around two to three kilohertz — you lose five to eight decibels right there. The rock breaker's energy extends up to two kilohertz, so it's catching the tail of the mass-air-mass resonance and the leading edge of the coincidence dip simultaneously.
The window is failing in two different ways at two different parts of the spectrum, with no safe band between them. Compare that to speech, which sits mainly between three hundred hertz and three kilohertz — above the mass-air-mass resonance for most windows. A standard double-glazed unit does a reasonable job on speech. But the rock breaker spans the gap between those two failure pattern.
Which brings us to why the airborne-versus-structure-borne question matters so much. If the dominant path is airborne — sound hitting the window, vibrating the glass, re-radiating into the room — upgrading the glazing solves most of the problem. If it's structure-borne — vibration traveling through the ground, up the building frame, into the walls and floor — a new window does almost nothing, because the sound is being generated inside the room by the walls themselves vibrating.
The frequency profile gives us clues. Structure-borne transmission tends to be strongest below a hundred hertz, where the building's structural resonances live. You'd feel it in your feet. Marcus didn't mention feeling vibration — he's measuring airborne sound pressure levels. His octave-band data shows energy concentrated above a hundred twenty-five hertz, consistent with airborne transmission through the facade. That's a good sign, because airborne is solvable.
It also means we can frame everything around the window rather than chasing structural isolation, which is vastly more expensive. The five questions Marcus asked — attenuation estimate, transmission path, glazing specs, hearing protection, subjective mitigation — they stack in exactly that order. You figure out what your current window is doing, confirm it's airborne, specify what would actually work, deal with the physiological limits of ear protection, and finally bring in the psychoacoustic tricks for tonight.
The thread running through all of it is that hundred twenty-five hertz to two kilohertz band. Every recommendation traces back to why that band defeats standard solutions. It's not just "construction is loud." It's that this particular kind of loud lands exactly where our buildings and our ears are least equipped to handle it.
Let's start with the attenuation estimate. That twenty-five point two decibel number Marcus got — ninety-three point two outside minus sixty-eight inside — is not the window's STC rating. It's a field measurement taken during active drilling, and it includes everything between the patio and the living room. The building facade, the wall assembly, the room's absorption, the fact that the measurement position inside isn't in a diffuse sound field. You can't just take that number and say "the window is STC twenty-five.
STC is a lab rating measured under very specific conditions — controlled source spectrum, reverberant rooms, flanking paths sealed. Marcus measured real life, which is messier. So what's the actual window-only attenuation likely to be?
I'd estimate twenty-eight to thirty-two decibels, corresponding to an STC thirty to thirty-three assembly. That's typical for standard double-glazing with four to six millimeter panes and a twelve millimeter air gap. The field number is lower than the window's actual performance because the room absorbs some sound, which reduces the indoor level and makes the window look better than it is. But flanking paths through the wall and ceiling add sound inside, which makes the window look worse. Those two effects partially cancel, but the net is that field attenuation usually understates the window's lab STC by about three to five decibels.
Standard double-glazing lines up with what you'd expect in a typical Jerusalem apartment. Most residential glazing here is four or six millimeter float glass in aluminum frames, maybe a twelve to sixteen millimeter air gap, no lamination, no gas fill. It's a thermal window, not an acoustic one.
The mass law tells us why. Six millimeter glass has a surface density of about fifteen kilograms per square meter. Each doubling of surface density adds roughly six decibels of transmission loss above the critical frequency. To hit STC forty, you need about twenty-five kilograms per square meter — roughly ten millimeters of monolithic glass, or a laminated equivalent. Marcus's windows are probably less than half that mass per pane.
The mass law only works above the critical frequency of the glass. Below that, you're in the stiffness-controlled region, and below that you hit the mass-air-mass resonance. Adding mass doesn't give you a uniform six-decibel-per-doubling improvement across the whole spectrum. At some frequencies, thicker glass can actually perform worse.
This is the coincidence dip phenomenon, and it's the most counterintuitive thing in glazing acoustics. At the coincidence frequency, the pane couples with the sound wave and basically becomes a loudspeaker. For six millimeter glass, that dip hits around two to three kilohertz, and you lose five to eight decibels. Marcus's rock breaker has strong energy up to two kilohertz, so it's catching that dip.
Which is why ten millimeter monolithic glass can be outperformed by six millimeter laminated glass at two kilohertz. The thicker monolithic pane has a lower coincidence frequency and a deeper dip because there's no damping layer. Laminated glass has a PVB interlayer that dissipates energy through shear deformation, reducing the dip by three to five decibels. More mass isn't always better if you're buying it in the wrong configuration.
Now, the airborne-versus-structure-borne diagnosis. The evidence strongly points to airborne transmission as the dominant path. First, the frequency profile — energy concentrated above a hundred and twenty-five hertz. Structure-borne transmission peaks below a hundred hertz, where you'd feel vibration in the floor and walls. Marcus didn't report feeling vibration.
Second, the mass-air-mass resonance of standard double-glazing falls between a hundred and two hundred fifty hertz — exactly where Marcus's noise profile starts. If the window is resonating sympathetically with the noise, that's airborne coupling. Structure-borne transmission wouldn't show that same alignment with the window's acoustic properties.
Third, the coincidence dip alignment. The rock breaker's upper band reaches two kilohertz, overlapping the coincidence region for six millimeter glass. Again, an airborne coupling mechanism. If the dominant path were structure-borne, the window's acoustic properties would be largely irrelevant — the sound would be generated by the walls and ceiling vibrating independently.
Marcus can be cautiously optimistic. Airborne transmission is solvable with glazing upgrades. Structure-borne would mean floating floors, resilient channels, decoupling the entire room — orders of magnitude more expensive. The frequency profile gives a clear signal that the window is the weak link.
It's worth spelling out why a hundred twenty-five hertz to two kilohertz is the worst possible band for standard glazing. Below a hundred twenty-five hertz, you're below the mass-air-mass resonance, and transmission loss improves as stiffness dominates. Above two kilohertz, you're above the coincidence dip, and the mass law kicks in. But in between, you've got the resonance trough on one end and the coincidence dip on the other — the window's transmission loss is at its minimum across that entire band.
It's a valley. The transmission loss curve looks like a U — high at very low frequencies, dropping into the mass-air-mass resonance, recovering slightly, then dipping again at coincidence, then climbing. Marcus's rock breaker is pouring energy into the bottom of that U for hours at a time.
Now let's talk about what actually works. And the first thing to know is that the metric most people shop for is the wrong one.
The number stamped on every acoustic window brochure. For construction noise, it's borderline misleading.
STC weights frequencies from a hundred twenty-five hertz to four kilohertz, but it weights the higher frequencies more heavily. A window can score STC forty-eight by blocking high frequencies really well, while leaking like a sieve at two hundred hertz. I've seen windows with STC forty-eight that have an OITC of thirty-two — a sixteen-decibel gap. You're paying for a number that doesn't describe the noise you actually have.
What's the honest metric?
OITC — Outdoor-Indoor Transmission Class — extends down to eighty hertz and weights the low end more realistically for transportation and construction noise. For Marcus's situation, target OITC thirty-five to forty. If a manufacturer can't give you the OITC rating, they're either hiding something or they don't know it, and neither is good.
What glazing configuration actually hits OITC thirty-five plus across a hundred twenty-five hertz to two kilohertz?
Triple glazing with asymmetric pane thicknesses. If all three panes are the same thickness, they all have the same coincidence dip at the same frequency, and you've just spent more money to fail in the same way. What you want is something like six millimeter laminated for the outer pane, a twelve millimeter air gap, ten millimeter laminated in the middle, another twelve millimeter gap, and six millimeter laminated for the inner pane. Each pane has a different coincidence frequency, so no single frequency band is vulnerable across all three barriers.
The PVB interlayer shears when the glass tries to vibrate at the coincidence frequency, dissipating energy as heat. A ten point three eight millimeter laminated pane — two layers of five millimeter glass with zero point three eight millimeters of PVB — outperforms ten millimeter monolithic glass by about four decibels at one kilohertz. That's right in the middle of Marcus's problem band.
The PVB is basically a shock absorber sandwiched inside the glass. With three laminated panes of different thicknesses, you've spread your coincidence dips across the spectrum.
You've pushed the mass-air-mass resonances around too. With two air gaps of different widths, you get two different resonance frequencies instead of one. Neither lands in the hundred twenty-five hertz to two kilohertz band if you spec it right. A properly designed asymmetric triple-glazed unit with laminated glass can hit STC forty-five to fifty and OITC thirty-five to forty. That would bring Marcus's indoor level from sixty-eight dBA down to roughly forty-eight to fifty-three dBA — within WHO guidelines.
That's a full window replacement. What if Marcus is renting, or doesn't have the budget to rip out his existing windows?
You install a second window — a separate frame with its own glazing — inside the existing window, leaving an air gap of a hundred to two hundred millimeters. That large gap is crucial. It creates a mass-spring-mass system where the two windows are the masses and the air between them is the spring. With a hundred fifty millimeter gap, the resonance frequency drops to about fifty to eighty hertz — below the rock breaker's energy band entirely.
The whole problem band is above the resonance, in the attenuation region.
A well-sealed secondary glazing unit with laminated glass on the inner pane can add ten to fifteen decibels to the existing window's performance. That takes Marcus's indoor level from sixty-eight dBA to fifty-three to fifty-eight dBA. And it costs forty to sixty percent less than a full window replacement because you're not demolishing anything.
The catch being you need a proper seal. If air leaks around the secondary frame, you've built an expensive bypass.
Compression seals all around, and the secondary frame must be decoupled from the primary frame — no rigid connections that transmit vibration. Done right, secondary glazing is the highest impact-per-dollar option for airborne construction noise. It addresses the exact frequency band where Marcus's current windows are weakest.
All right, so we've covered the window. But even with the best glazing, Marcus is still living next to an active excavation site. What about hearing protection while he's figuring out the window situation?
This is where people get surprised. Standard foam earplugs with an NRR of thirty-three will give you maybe fifteen to twenty decibels of real-world attenuation in the hundred twenty-five hertz to two kilohertz range. Not thirty-three. The NRR is a lab number measured under ideal conditions. In the real world, people don't insert them perfectly, and more importantly, low-frequency sound bypasses the earplug entirely.
Through the skull.
At two hundred fifty hertz, even with a theoretically perfect earplug, about fifteen decibels of sound reaches the cochlea through the bones of the skull. The sound vibrates your head, and your head vibrates your inner ear directly. No earplug can stop that. This is why you can still "feel" jackhammering with foam plugs in — you're literally hearing it through your skeleton.
There's a hard physiological ceiling.
The bone conduction floor limits passive low-frequency attenuation to about thirty-five to forty decibels, no matter what you do. You can put the best foam plugs on the market and then the best earmuffs over them, and you'll only get three to five decibels more than the plugs alone. Double protection sounds like it should double the attenuation, but it doesn't — you're already up against what your own skull will transmit.
Which means for Marcus, sitting in sixty-eight dBA with earplugs in, the best he can hope for is maybe forty-eight to fifty-three dBA at the cochlea. Quieter, but not quiet.
Product selection matters more than NRR numbers suggest. For low frequencies, custom-molded plugs with a three-flange design outperform standard foam. The custom mold seals the ear canal more consistently and the flanges provide multiple impedance changes that reflect low-frequency energy back out. Something like the ACS Pro Twenty-Seven can actually outperform a foam plug rated at NRR thirty-three in the hundred twenty-five to five hundred hertz range because the fit is more reliable.
What about earmuffs?
The 3M PELTOR X5A is one of the few earmuffs tested and rated down to a hundred twenty-five hertz, and it's a solid choice. But for Marcus's situation — sixty-eight dBA of sustained broadband — I'd recommend custom-molded plugs as the first line. If he's going to be home for hours, the comfort difference alone means he'll actually wear them. A foam plug you take out after an hour because it's irritating is providing zero decibels of attenuation.
The practical hearing protection answer is custom-molded plugs with good low-frequency performance, and accept that you're getting fifteen to twenty decibels of real attenuation, not thirty-three. Double protection adds a little but not much. The bone conduction floor means you're never getting below about forty-eight dBA at the ear.
Which brings us to the fifth question — and this is where it gets genuinely clever. What can Marcus do tonight, without touching the windows, to make sixty-eight dBA of rock breaker noise subjectively less punishing? The answer is brown noise masking through a subwoofer.
This is not acoustic cancellation. This is psychoacoustic.
The idea is you introduce a masking sound that raises your hearing threshold at the frequencies where the construction noise lives, so your auditory system stops flagging it as salient. Brown noise has a spectral slope of minus six decibels per octave — energy drops by six decibels for every doubling of frequency. It's heavily weighted toward the low end, exactly where Marcus's rock breaker is concentrated.
Whereas white noise is flat — equal energy per hertz — and pink noise is minus three decibels per octave. White noise sounds hissy and bright. Brown noise sounds like a deep, heavy waterfall or a distant jet engine.
That rumble overlaps the perceptual space of the rock breaker's low-frequency energy. If you play brown noise through a subwoofer at forty-five to fifty dBA, it doesn't drown out the construction noise — it's quieter than the construction noise — but it raises the threshold at which your brain registers it as a distinct, attention-grabbing signal. The perceived loudness drops by ten to fifteen decibels through auditory masking, especially in the hundred twenty-five to five hundred hertz range.
You're not making the noise quieter. You're making your brain care less about it.
That distinction matters because it means this actually works even though the physical sound pressure level in the room hasn't changed. Your auditory system is now processing the construction noise against a background of brown noise occupying the same neural channels, and the salience drops substantially.
The subwoofer is important here. Laptop speakers or a small Bluetooth speaker won't reproduce the low end where brown noise does its work. You need something that can push real energy below two hundred hertz.
A modest home subwoofer, even a used one, will do it. Set the brown noise to loop — there are apps and YouTube tracks for this — position the subwoofer near where you're sitting, and set the level so it's audible but not intrusive. Forty-five to fifty dBA is about the level of a quiet conversation. You're not adding to the noise problem, you're reshaping the noise landscape.
This pairs with the earplugs. The earplugs knock down the airborne component by fifteen to twenty decibels — taking cochlear exposure from sixty-eight to roughly forty-eight to fifty-three dBA — and the brown noise masking raises the hearing threshold for what gets through. You're attacking the problem on both the acoustic path and the perceptual path simultaneously.
If you add all this up for Marcus, the recommendations stack by impact per dollar. Tier one, free to nearly free: brown noise masking through a subwoofer, combined with the best earplugs you can get — ideally custom-molded with good low-frequency fit. That dual approach takes the subjective burden from "I can't think in my own home" to "I can hear it but I can live with it.
Tier two, the highest-impact structural change: secondary glazing with a hundred fifty millimeter plus air gap and laminated glass on the inner pane. Ten to fifteen decibels of real attenuation added to the existing window, bringing indoor levels from sixty-eight to fifty-three to fifty-eight dBA. Costs forty to sixty percent less than full window replacement, and it's the best engineering match for the hundred twenty-five hertz to two kilohertz band because the large air gap pushes the mass-spring-mass resonance below eighty hertz — out of the problem zone entirely.
Tier three, if you're replacing windows anyway: specify OITC of thirty-five or higher, not just STC forty-five plus. Asymmetric triple glazing with laminated panes of different thicknesses. Six millimeter, ten millimeter, six millimeter, with PVB interlayers throughout. That spreads the coincidence dips so no single frequency band is exposed, and the lamination damps the dips that remain by three to five decibels. That'll get indoor levels down to forty-eight to fifty-three dBA — within WHO guidelines — but it costs significantly more than secondary glazing and may not outperform it for this specific noise profile.
The through-line for all of it is managing expectations. The bone conduction floor is the fundamental limit. No passive solution eliminates low-frequency construction noise. The goal is reduction from sixty-eight dBA to fifty to fifty-five dBA — not silence. You can make it tolerable. You cannot make it disappear.
Even with all these solutions, there's a fundamental limit we haven't talked about. And it points to what the future of urban noise control might look like.
Active noise cancellation for this exact problem. It's theoretically possible — hydraulic rock breakers have a quasi-periodic pattern, not truly random noise. If a system can predict the next impact cycle, it can generate an anti-phase signal. Prototype systems in labs have hit ten to fifteen decibels of cancellation in the hundred to three hundred hertz band. But no consumer product does this yet.
The catch is that quasi-periodic part. Traffic hum is steady-state, easy for ANC to lock onto. A rock breaker has a rhythm but it's irregular — the piston hits, the rock fractures unpredictably, the operator changes angle. The prediction window is tiny and the error penalty is adding more noise instead of canceling it.
That's the research frontier. Groups in Europe are working on adaptive ANC for construction sites using microphone arrays and machine learning to predict the impact pattern a few milliseconds ahead. The lab results are promising — ten to fifteen decibels in that low band where passive solutions hit the bone conduction wall. But it's probably three to five years from a product you can buy.
Which means for now, the stack we described is the ceiling. Secondary glazing plus brown noise plus custom plugs gets you to tolerable. ANC gets you to quiet, someday.
The broader point — Marcus's situation isn't unique. Cities everywhere are densifying. Jerusalem is growing fast, construction cycles are accelerating, and the conflict between building the city and living in it is only going to get sharper. Acoustic design isn't a luxury feature. It's a public health intervention hiding in a window specification.
The five-question framework Marcus gave us — measure, diagnose, specify, protect, mask — that's portable. It works whether you're next to a Jerusalem excavation or a highway in Los Angeles or a subway extension in London. The physics doesn't change. The frequencies shift slightly depending on the source, but the same principles apply.
If listeners are dealing with something similar, we'd love to see their measurements. Send them in. Outdoor dBA, indoor dBA, and a rough description of the frequency character — is it a low rumble, a high whine, an intermittent thump? The same analysis applies, and we might feature it in a future episode.
Marcus, thank you for sending this in. It's exactly the kind of question that makes the engineering come alive — a real measurement, a real apartment, and a real person trying to think straight while a rock breaker rewrites the acoustic properties of his living room.
That is Hilbert's daily fun fact.
Hilbert: In the nineteen fifties, researchers in Patagonia used photographic plates recovered from a single high-altitude balloon flight to confirm that cosmic ray muons arrive with a flux of roughly one per square centimeter per minute at sea level — and those same plates, stored in a Buenos Aires archive, are still used today as a calibration reference for modern detectors.
...right.
One plate, seventy years of calibration data. That's a good run.
This has been My Weird Prompts. Thanks to our producer Hilbert Flumingtop, and thanks again to Marcus James for the prompt. If you're enjoying the show, leave us a review — it helps other people find us.
I'm Corn.
I'm Herman Poppleberry.
We'll be back soon.
