Why Intercom Audio Quality Tanks During Wind and Rain
Here is the complaint that lands on my desk two or three times a year: "The intercom worked perfectly at commissioning. Now the guard station can't understand anyone when the weather turns." The customer suspects a failing unit. The installer suspects the network. Nine times out of ten, both are wrong — the system is performing exactly as installed, and the installation was designed on a calm day for calm days. Outdoor audio is an acoustics problem wearing an electronics costume, and the industry keeps trying to solve it with DSP settings when the failure was baked in at the mounting bracket.
The contrarian position I will defend below: most outdoor intercom "audio quality" problems cannot be fixed in software, because wind noise is not noise in the signal-processing sense. It is signal. Understanding why changes how you design, what you buy, and where you put it.
The Outdoor Microphone Reality
An electret or MEMS microphone element does not know the difference between a human voice and turbulent air hitting its port. Voice arrives as acoustic pressure waves; wind arrives as direct pressure fluctuation on the diaphragm — and the wind's energy is concentrated below about 500 Hz, exactly where it swamps the 300 Hz-3 kHz band that carries speech intelligibility. The magnitude difference is brutal: a 20 mph gust across an unprotected mic port can generate low-frequency pressure equivalent to well over 100 dB SPL at the diaphragm, while the visitor standing two feet away is delivering maybe 65-70 dB SPL of speech. No gain structure survives a 30-40 dB deficit. The microphone is not picking up "background noise" — it is being physically buffeted, and the speech is a whisper underneath it.
Rain is a different mechanism with the same result: broadband impact noise from drops striking the faceplate, the housing, and whatever metal surface the unit is bolted to. A door station mounted on a hollow steel mullion becomes a drum head; every drop within a meter couples through the structure into the mic. I measured this on a parking-garage entry station once — the audio during moderate rain was dominated not by rain falling near the unit but by rain hitting the sheet-metal hood two feet above it, conducted straight down the mounting surface.
Why Echo Cancellation Doesn't Cancel Wind
This is the most common misdiagnosis. Modern intercoms advertise acoustic echo cancellation (AEC) and noise reduction, and integrators reasonably assume "noise reduction" covers weather. It does not, for a structural reason: AEC works by subtracting a known reference — the far-end audio the speaker is playing — from what the microphone hears. Wind has no reference signal. There is nothing to subtract against.
Stationary-noise suppressors fare little better. They model a steady noise floor (HVAC hum, traffic rumble) and subtract the estimate, which works because the noise is statistically predictable frame to frame. Wind is the opposite — non-stationary, gusting, spectrally overlapping speech. Aggressive suppression tuned against gusts starts eating the speech itself, which is why cranking the noise-reduction slider produces that underwater, artifact-riddled audio the guard describes as "worse than the noise was." DSP earns its keep on the last 6-10 dB of a problem. It cannot manufacture 30 dB of physics. The 30 dB comes from mechanical design and mounting, or it does not come at all.
Beam-Form vs Omni Microphones
Multi-microphone arrays help, and it is worth being precise about why. A two- or four-element array can form a pickup lobe toward the visitor position and reject sound arriving off-axis, buying real gains — typically several dB of signal-to-noise improvement in the speech band, more when the interference is directional traffic noise. Units engineered for harsh outdoor duty stack these techniques deliberately; Code Blue's emergency stations are a good reference point for what purpose-built looks like — the microphone is recessed and mechanically shielded by the housing itself, because their units live on open campuses and parking structures where wind is the operating condition, not the exception. Compare that to a generic door station where the mic port is a bare hole in a flat faceplate pointed at the weather.
But hold the enthusiasm to what arrays actually do: beam-forming rejects acoustic sound arriving from off-axis directions. Wind buffeting is not off-axis sound — it is direct pressure on each element. A gust hits all the capsules with uncorrelated turbulence, and the array math gets nothing coherent to cancel. Arrays plus mechanical wind protection is a system; arrays instead of wind protection is a spec-sheet purchase.
Outdoor Audio Design Worksheet
Before mounting anything, walk the location and fill this in. It takes fifteen minutes and predicts almost every complaint you would otherwise get in month three.
| Factor | What to record | Red flag |
|---|---|---|
| Prevailing wind exposure | Direction and typical speed; is the faceplate facing into it? | Mic port facing prevailing wind, open fetch (parking lot, field) |
| Ambient SPL, worst hour | Measured dBA at rush hour / HVAC-on | > 70 dBA ambient at the talk position |
| Mounting substrate | Masonry, filled frame, or hollow metal? | Hollow mullion or sheet-metal panel (rain drum) |
| Overhead protection | Soffit, hood, or bare exposure? | No cover and no room to add a hood |
| Visitor position | Distance and height of the talker's mouth | > 0.6 m from mic, or drive-up window geometry |
| Speaker path back | What must the visitor hear over? | Idling truck at a gate: 75-85 dBA at the lane |
Speaker SPL vs Background Noise
Intelligibility runs both directions, and the return path is the one everyone forgets to engineer. For the visitor to understand the guard, the speaker output at the listening position needs to sit roughly 10-15 dB above ambient. At a truck gate idling at 80 dBA, that means delivering 90+ dBA of clean speech at the driver's ear — which is a real electroacoustic demand, not a checkbox. A typical compact door-station speaker driven by a 2 W amplifier produces on the order of 90-95 dB SPL at 1 meter; at 2-3 meters of drive-up geometry you have lost 6-10 dB to distance before weather adds anything. This is why gate stations that are perfectly adequate at a pedestrian door fail at vehicle lanes: the acoustic budget was never there. Spec the speaker and amplifier against measured worst-hour ambient at the actual talker position, and prefer units that publish output SPL rather than just amplifier watts.
Why Bell-and-Hood Mounting Matters
The cheapest 20 dB you will ever buy is mechanical. Recessed or hooded mounting breaks the direct wind path to the mic port — turbulence has to turn corners, and each corner strips energy from the low-frequency buffeting before it reaches the diaphragm. A simple weather hood or recessed bell also solves the rain-impact problem by keeping drops off the faceplate, and mounting into masonry or a foam-filled frame instead of hollow sheet metal kills the drum-head coupling. Orient the unit so the mic port faces across or away from prevailing wind, never into it; on exposed sites, a 90-degree turn of the station has fixed "audio problems" that two firmware updates could not. None of this appears on a datasheet, which is exactly why the same hardware produces a great system on one site and a callback generator on another.
Gain Staging: The Configuration That Actually Helps
If the mechanical design is right, there is one software layer genuinely worth tuning: gain structure. Automatic gain control (AGC) is the setting that turns a marginal outdoor station into an unusable one, because AGC chases the loudest thing at the mic — and outdoors, the loudest thing is the gust, not the visitor. During the lull the AGC cranks input gain hunting for signal; the next gust arrives amplified into clipping, and the guard gets alternating silence and distortion. On exposed stations I set input gain manually, calibrated against a live talker at the real visitor position, and either disable AGC or fence it to a narrow range where the hardware allows it. Same discipline on the output side: set speaker level against measured ambient, not against what sounds comfortable in a quiet commissioning hour, and leave headroom below clipping — a distorting speaker loses more intelligibility than a slightly quiet one.
Then verify with the test that matters: a two-way conversation at the actual visitor position, with air moving, scored on whether an unfamiliar sentence is understood on the first pass. Not a tone, not "can you hear me" — an unfamiliar sentence. Familiar phrases get reconstructed by the listener's brain and mask a marginal channel; a sentence the guard has never heard is the honest metric. Log the gain settings in the commissioning record so the next tech does not "fix" them back to defaults after a board swap.
Audio Codec Tradeoffs Outdoors
The network leg matters less than people think, but it is not free. Narrowband G.711 caps the audio at 3.4 kHz — serviceable for clean speech, but it discards the consonant energy above 3 kHz that lets a listener separate speech from residual weather noise, so a marginal outdoor signal gets harder to understand after encoding. Wideband codecs preserve content to 7 kHz and buy noticeable intelligibility on SIP paths that support them end to end. The trap is packet loss: outdoor stations on marginal wireless or congested links suffer more from 2-3% loss and jitter than from codec choice, and heavily compressed low-bitrate modes fall apart when the input is already noisy. Order of operations: fix the acoustics, keep the transport clean, then spend attention on codec selection.
Designing Intercoms for Real Weather
Pulling it together, the design sequence that survives weather looks like this: survey wind and ambient SPL at the worst hour, not the site-visit hour. Choose hardware engineered for outdoor acoustic duty — shielded or recessed mic, published SPL output, IP66-rated housing so the acoustics survive the second winter, wideband codec support. Mount it hooded or recessed, into solid substrate, mic port out of the prevailing wind. Set DSP conservatively and let the mechanical design do the heavy lifting. Then commission on a bad day: if you can only test on a calm afternoon, bring a leaf blower — 10 seconds of moving air across the faceplate at commissioning has saved me multiple month-three callbacks. Browse the Intercom & Paging catalog for the range of outdoor-rated stations, and look at the Code Blue emergency stations as the benchmark for what fully weather-engineered audio hardware looks like when the use case cannot tolerate failure.
Deployment takeaway: Wind noise is direct pressure on the diaphragm, not ambient sound — DSP cannot subtract it because there is nothing coherent to subtract. Budget your intelligibility mechanically: hood or recess the station, mount into solid substrate, face the mic port away from prevailing wind, and spec speaker SPL 10-15 dB above measured worst-hour ambient at the visitor position. Treat noise reduction and beam-forming as the last 6-10 dB, never the first 30. And commission with moving air — a leaf blower across the faceplate for ten seconds tells you more than any calm-day test call. If an existing site tanks in weather, re-aim and hood the station before you replace it; the fix is usually a bracket, not a board.
Where This Fits in a Deployment Program
Outdoor audio deserves the same site-survey rigor the industry already gives camera fields of view and reader mounting heights. When I review intercom specs for gates, campuses, and parking structures, the acoustic survey — wind exposure, worst-hour ambient, substrate, visitor geometry — is the first document I ask for, and its absence predicts the callback pattern almost perfectly. Purpose-built outdoor stations cost more than repurposed indoor door stations precisely because the housing is doing acoustic work; on exposed sites that delta buys years of not rolling trucks. Review the options across all Code Blue products for hardened emergency and areas-of-refuge duty, or the wider Intercom & Paging catalog for door and gate stations by environment. If you have a site where the audio dies every time the weather turns, send over the mounting photos, the wind exposure, and your worst-hour ambient reading — we can help you spec a station and mounting approach designed for the days your customer actually calls about.