6 The quiet that wasn’t
Decarbonisation arrived wearing the promise of silence. The electric city would be a quiet city; the combustion engine’s roar would simply subside, and the acoustic dividend would be collected alongside the carbon one. What has happened instead is stranger and more instructive. The electric vehicle proved quiet enough at low speed to be dangerous, and is now required by law to make artificial sound. The heat pump replaces a near-silent indoor boiler with a continuous exterior hum a few metres from the neighbour’s bedroom. The wind turbine trades the power station’s remoteness for a rhythmic swish over rural dwellings. This is the distributed transition — felt at the doorstep, one installation at a time. Its counterpart is the concentrated transition examined next, in which the same energy system pools its sound into single, vast, humming sites.
The paradox flattens easily, and in two opposite directions. The false pessimistic version says the transition is making things louder; taken as a whole, it is not, and nothing here should be read as a case against decarbonisation. The false optimistic version says electrification is an acoustic policy; it is not that either. The accurate version is narrower and more interesting: the energy transition redistributes sound across space, time and frequency. It withdraws a source that was loud, intermittent and concentrated along roads, and introduces sources that are quieter, continuous and scattered across kerbsides, gardens and horizons. Whether any particular window ends up better or worse off depends on where it sits in that redistribution.
Three properties recur across the three sources examined here, and between them they are what the redistribution consists of. Each new source is quieter than the one it displaces, which is the reason the promise of silence was plausible. Each is more continuous, which means the relief of the interval — the gap between one lorry and the next — is what the transition actually removes. And each sits closer to the dwelling, or is tonal, or is modulated, which is to say that each has more of its character in the parts of the signal an annual A-weighted average is least equipped to record. The metrics were tuned to the source that is leaving. That mismatch was set out in its cleanest form in Chapter 2 and reaches its fullest expression at data-centre scale; here it is the condition the argument works in.
6.1 The vehicle’s two silences
The electric car has two silences, and only one of them was ever real. Below roughly 20 km/h it is markedly quieter than its combustion predecessor — quietly enough that the difference became a pedestrian-safety problem, and the law moved to fill the gap. Above that speed the silence is largely a fiction, because what a light vehicle emits at urban cruising speed comes mostly from the contact between its tyres and the road, a source indifferent to what turns the wheels. The first silence had to be legislated away; the second needs no legislation, because it never existed. Between them they define both the size of the acoustic dividend electrification actually pays and the narrow band of conditions under which it is paid at all.
Take the first silence. For a century, vehicle sound regulation ran in one direction only: downward, capping the maximum a machine might emit. UNECE Regulation No 138, adopted in 2016, runs the other way. It requires pure electric and hybrid vehicles capable of moving without their combustion engine to carry an Acoustic Vehicle Alerting System (a loudspeaker, in effect), generating sound automatically from start-up to approximately 20 km/h and when reversing, because the vehicles’ reduced audibility endangers pedestrians, cyclists and above all visually impaired road users (United Nations Economic Commission for Europe, 2017). The regulation prescribes minimum overall levels that rise with speed, so that a listener can judge not merely presence but behaviour. Those minima are 50 dB(A) at 10 km/h, 56 dB(A) at 20 km/h and 47 dB(A) in reverse. It adds a mandatory frequency shift as the vehicle accelerates, and a ceiling of 75 dB(A) to keep the cure from reproducing the disease (United Nations Economic Commission for Europe, 2017). In the European Union the requirement entered through Regulation (EU) No 540/2014, whose Annex VIII was rewritten to incorporate the UNECE provisions: mandatory for new vehicle types from July 2019 and for all new electric and hybrid vehicles from July 2021 (European Commission, 2017; European Parliament and Council of the European Union, 2014).
The EU text adds a stipulation of genuine acoustic-cultural interest. The generated sound should be continuous, should indicate the vehicle’s behaviour, and should be “similar to the sound of a vehicle” of the same category fitted with an internal combustion engine (European Commission, 2017). The machine designed to retire the engine is required to impersonate it. This is not nostalgia written into law but an admission about perception: a century of exposure has trained urban ears to parse one particular spectrum as approaching car, and the trained response is a safety asset too valuable to discard along with the technology that produced it. The engine is being decommissioned as a power source and retained as a signal.
Read as safety engineering, AVAS is uncontroversial and evidently sensible. Read as measurement politics, it is a quiet landmark: the first occasion on which the regulatory apparatus, having spent decades defining noise as an excess to be capped, found itself legislating a minimum — conceding, in legal text, that sound carries information whose absence is itself a hazard. Floor and ceiling now sit in the same annex, and the distance between them, 56 to 75 dB(A) at 20 km/h, is the entire band within which the electric vehicle is permitted to exist acoustically. Noise regulation counts those decibels as unwanted sound; safety regulation counts the same decibels as a signal that must be present, and the emission is entered twice with opposite signs.
The second silence dissolves under measurement, and the reason was met at the close of Chapter 5: above modest urban speeds the tyre, not the engine, is what a light vehicle is heard by. The measurement campaign of Pallas and colleagues, undertaken to extend the CNOSSOS-EU emission model to electric vehicles, put numbers to what that dominance implies for electrification: at 30 km/h the electric car was on average 2.6 dB quieter than its combustion counterpart, and above 40 km/h the difference had collapsed to between 0.4 and 1.6 dB (Pallas et al., 2016). That is within touching distance of nothing, and well inside the tolerance the Directive itself allows on emission inputs. A companion study of tyres representative of electric-vehicle fitment found rolling noise to be the main source even at 20 km/h (Czuka et al., 2016). Seen against these figures, the R138 threshold is not an arbitrary administrative round number but a codified acoustic judgement: 20 km/h is the speed above which the regulator assumes the tyre will announce the vehicle unaided (United Nations Economic Commission for Europe, 2017). The two silences meet exactly at the point where the law stops speaking.
The diagnostic consequence is easily stated and widely underestimated. Electrifying the fleet does not quieten the street wherever traffic moves faster than roughly 30 km/h. What it changes is the spectrum, and with it the composition of the residual problem — the low-frequency bands where the engine lived subside, the mid-frequency broadband of rolling noise remains, and what is left is a matter of pavements, speed and tyre design rather than powertrains. It is also, incidentally, an argument already won by the 30 km/h urban limit: the only regime in which the electric fleet’s acoustic dividend is actually paid out is one in which vehicles spend most of their time at speeds where propulsion still matters. The quiet electric city is available, but it has to be driven slowly into being.
6.2 The heat pump on the party wall
If the electric vehicle rearranges the street, the air-source heat pump rearranges the garden fence. Acoustically it is a fan and a compressor running for long stretches of the day and, unlike the gas boiler it replaces, mounted outdoors. Siting is often within a few metres of a neighbour’s bedroom window — frequently against a wall, or in the corner formed by two, with each reflecting surface adding to the level at the receiver. What it emits is not loud by any traffic standard. It is continuous, it carries tonal components, its energy sits low in the spectrum, and it sits exactly where the domestic soundscape used to be quietest, at the hours when the surrounding street is quietest too.
Each of those four properties works against the instrument used to assess it. A continuous source has no events to count and no gaps to lose; averaged over a day it can be reported as modest while never once affording relief. A tonal component is what the ear locks onto and the broadband total dilutes. Low-frequency energy is what A-weighting is built to discount, by some 16 dB at 125 Hz and 26 dB at 63 Hz, as recalculated in Chapter 2. And a source that matters most when everything else falls silent is a source whose significance is a matter of contrast rather than level — which no absolute threshold, however carefully negotiated, is constructed to see. This is in miniature the profile that returns at industrial scale in Chapter 7: quiet by the numbers, intolerable in the room.
The complaint statistics have begun to register it. In its 2026 data briefing, covering October 2024 to September 2025, the Chartered Institute of Environmental Health recorded noise complaints about air-source heat pumps as a distinct source for the first time: 103 complaints across England and Wales, concentrated in the South West, against a total of just under 307,000 noise complaints of all kinds (Chartered Institute of Environmental Health, 2026). One hundred and three is a small number — a rounding error in the national ledger, three complaints in ten thousand. Its significance lies in the fact of appearing at all. Categories in complaint statistics behave rather like species in a census: the year in which one is first recorded separately matters independently of its initial abundance, because it marks the moment at which an administrative apparatus judged the thing distinct enough, and likely enough to grow, to deserve a line of its own. The count is small because the installed base is small; the installed base is the variable that policy is actively trying to enlarge.
The governance response has been unusually quick, and its form is instructive. In Britain, a domestic heat pump may be installed without planning permission provided a standardised sound calculation (distance, reflecting surfaces, barriers) shows the level at the nearest neighbour’s window to fall below a fixed threshold. In March 2025 the standard governing that calculation, MCS 020, was split in two and revised, and the threshold tightened from 42 to 37 dB(A) (The MCS Service Company, 2025). Five decibels is a substantial correction: a cut of roughly two-thirds in the sound energy a compliant installation may deliver to the neighbour. It was conceded not because heat pumps had grown louder but because the earlier limit, set when installations were few, proved too permissive once they multiplied.
The revision is a small, complete specimen of something that recurs throughout these chapters. A threshold looks like a scientific constant and behaves like a negotiated settlement — one that holds while the population of sources is sparse and is reopened when that population grows. Note also what the tightening does not address: tonality, and the character of the sound rather than its level. The number moved; the quantity being counted did not. The interactive later in this chapter implements the underlying arithmetic and lets both thresholds be tested against any unit, distance and siting.
6.3 The turning blade
The third new source stands furthest from the window and generates the longest arguments. Wind-turbine sound is aerodynamic in origin, dominated by the interaction of the blade with the air, and its most distinctive feature is amplitude modulation: a periodic rise and fall of the level at the blade-passing rhythm, the swish that distinguishes a turbine from any steady industrial source (C. Hansen & Hansen, 2020). Modulation matters because hearing is built to notice change rather than constancy. A fluctuating sound is more salient than a steady one of identical average level — and average level is exactly the quantity the assessment records. The modulation depth can be several decibels while the mean moves not at all.
The field evidence points where the mechanism predicts. A year-long study at dwellings near two Australian wind farms found that residents’ annoyance reports clustered overwhelmingly at night and in the early morning, and that in at least half of the recorded episodes the sound was described as swish or swoosh (K. L. Hansen et al., 2021). The complaints track a descriptor (a quality of the sound) rather than a level; and they concentrate in the hours when atmospheric conditions favour modulation and background sound is lowest, which is to say in the period the assessment framework covers least well.
The guidance registers this source with unusual, and revealing, hesitancy. The WHO’s 2018 environmental noise guidelines supply the road-traffic recommendations that anchor Chapter 3. For wind turbines they issue only a conditional recommendation, that average exposure be kept below 45 dB \(L_\text{den}\), the conditionality reflecting the limited body of high-quality evidence (World Health Organization Regional Office for Europe, 2018). For night-time exposure no recommendation is offered at all: the evidence was graded too low to support one. And the guidelines volunteer a caveat that repays slow reading. Describing this source through \(L_\text{den}\) or \(L_\text{night}\), they concede, “may be a poor characterisation of wind turbine noise” and may limit the very ability to observe associations between it and health outcomes (World Health Organization Regional Office for Europe, 2018).
The concession is made in the instrument’s own founding document, by the instrument’s custodians, and its logic runs in a circle. The indicator averages away the modulation; the modulation is what the evidence suggests does the annoying; studies built on the indicator therefore struggle to find the association; the association, unfound, is graded as weak evidence; the weak grading yields a merely conditional recommendation — and the indicator is retained, because it is the indicator, and because a conditional recommendation is not the sort of finding that displaces a measurement standard. The 58 Hz tone met earlier vanished into the A-weighted total by the same arithmetic. No further comment is offered here; the interest on this deposit falls due in Part III.
What the indicator averages away can still be made audible. Below, a synthesised turbine-like sound holds its mean level fixed while its amplitude modulation is raised or lowered at will — so that the modulated and steady versions share an identical \(L_\text{eq}\), the very quantity the assessment records. One of them is far harder to ignore than the other.
Source · Synthetic auralisation generated in the browser; not a field recording. Modulation phenomenology and the typical audible depth follow C. Hansen & Hansen (2020); peak-to-trough depth definition after the Institute of Acoustics reference method (Institute of Acoustics, 2016). The modulated and steady signals are normalised to an identical \(L_\text{eq}\); the arithmetic was verified in Node.js.
The source is no abstraction in Andalusia. At the close of 2025 the region counted 158 wind farms totalling some 3,712 MW of installed capacity, and the zones approved under Spain’s maritime spatial plans for future offshore wind lie principally off the coasts of Málaga, Granada and Almería (Agencia Andaluza de la Energía, 2026). The turbines of the energy transition are not someone else’s horizon.
6.4 Composing the net-zero street
The argument can now be handed to the reader, because it is an argument about combination and combination is what arithmetic does better than prose. A decarbonised street is assembled in the simulator below from the components verified in the preceding sections, and set against its own combustion baseline. The traffic module implements the CNOSSOS-EU category-1 emission model with the coefficients as amended in 2021, treating electric vehicles as rolling noise plus the R138 minimum below 20 km/h (European Commission, 2021; Pallas et al., 2016; United Nations Economic Commission for Europe, 2017). The heat-pump module implements the distance-and-reflection arithmetic of the MCS calculation and tests the result against both the former 42 dB and the current 37 dB thresholds (The MCS Service Company, 2025). The turbine module propagates a declared sound power across open ground against the WHO’s conditional 45 dB guideline (World Health Organization Regional Office for Europe, 2018). Three modules of input; one window as output.
The simulator is worth provoking deliberately. Set the traffic to 50 km/h and slide the electric share from nothing to everything: the façade level barely moves, and the spectrum below shows why — the electric bars fall away at 63 and 125 Hz and hold their ground at 1 and 2 kHz. Drop the speed to 20 km/h and the dividend appears at last, some four and a half decibels of it. Then quieten the street altogether, and the heat pump and the turbine surface as the loudest things at the window — sources that would once have drowned unnoticed in traffic.
Source · Traffic emission: CNOSSOS-EU category 1, Commission Directive (EU) 2015/996 as amended by Commission Delegated Directive (EU) 2021/1226 (coefficients cross-checked against the NoiseModelling reference implementation, Université Gustave Eiffel) (European Commission, 2021; Kephalopoulos et al., 2014). AVAS minima: UNECE Regulation No 138 (United Nations Economic Commission for Europe, 2017). Electric-vehicle treatment follows Pallas et al. (2016). Heat-pump thresholds: MCS 020 / MCS 020 a) (The MCS Service Company, 2025). Wind-turbine guideline: WHO 2018 (World Health Organization Regional Office for Europe, 2018). Propagation uses standard geometric divergence; assumptions are stated in full beneath the simulator. All arithmetic verified in Node.js. Accessed 2026-07-15.
That last behaviour is the whole argument in operational form. The new sources do not become audible because they grow. They become audible because the thing that masked them is withdrawn — which means the transition manufactures its own complainants, and does so most reliably where it succeeds best. A street that electrifies its traffic and warms its houses with compressors has not exchanged a loud soundscape for a quiet one. It has exchanged an intermittent, familiar, broadband, publicly owned sound for a continuous, novel, tonal, privately owned one, and it has moved the source from the road to the boundary wall, taking responsibility with it from the council to the neighbour.
6.5 Designing the transition’s sound
Nothing about the soundscape of the net-zero street is accidental. It is being decided at the time of writing — in type-approval annexes, in permitted-development thresholds, in setback distances, in the March revision of a calculation standard most people will never read. The window in which such defaults are set is brief, and it stands open now.
Every source examined here is more governable than the one it replaces, not less. The AVAS spectrum is specified by regulation, which means it could in principle be specified well: nothing but convention requires a synthetic warning to imitate a combustion engine rather than to be designed, from first principles, as a signal that is maximally detectable and minimally annoying. The heat pump’s siting is a published formula with distance, reflection and screening as its terms — three levers, all available at the survey stage and all cheaper before installation than after. The turbine’s modulation is an active field of research with engineering levers attached (C. Hansen & Hansen, 2020). The combustion engine offered none of this: its sound was a by-product of thermodynamics, and could be muffled but not composed. The sources of the transition are, for the first time, authored. The design turn of the final chapter follows from that, as the completion of the diagnosis.
One boundary must be drawn honestly before moving on. Everything examined here is small, near and numerous: a loudspeaker per bumper, a compressor per garden, a turbine per ridge. Each is quiet on its own terms; the burden is the sum, and the sum accumulates one planning decision at a time. The same transition has another face — singular, vast and continuous — in which the electricity for the vehicles and the models alike is consumed in windowless halls that hum at frequencies the A-weighted map barely registers. The distributed transition has been audited. The concentrated one is next.