---
title: How Many Speakers Should a Car Have, and Where? — A Simulator That Follows Why Every Seat Hears Something Different
url: https://oosioo.com/en/p/%EC%B0%A8-%EC%95%88-%EC%8A%A4%ED%94%BC%EC%BB%A4%EB%8A%94-%EC%96%B4%EB%94%94%EC%97%90-%EB%AA%87-%EA%B0%9C%EB%A5%BC-%EB%8B%AC%EC%95%84%EC%95%BC-%ED%95%A0%EA%B9%8C
date: 2026-10-29T12:36:12+00:00
author: SYSOP
summary: Cars now carry anywhere from 8 to 36 speakers. Yet the driver, the passenger and the back seats hear different things in the same car, because the cabin is small, wrapped in glass, and nobody sits in the middle. I built a simple acoustic model that stacks 5,832 reflection paths inside a box-shaped cabin, and a solver that picks positions from 32 candidate spots (sedan) or 38 (3-row SUV) for a goal. Limited to 4 speakers for the driver, it kept A-pillar tweeters, a center speaker and a subwoofer; for all seats, the center speaker came first. The score stopped improving at about 13 speakers. Per-seat sound zones kept about 20 dB of contrast with headrest speakers even with the head moved 3 cm, but only 10 dB with door speakers.
---
# How Many Speakers Should a Car Have, and Where? — A Simulator That Follows Why Every Seat Hears Something Different

![A sedan cabin seen from above, dashboard on the left. The front doors have blue woofers and orange tweeters near the top and bottom. A blue circle spreading from the left door and an orange circle from the right door overlap near the driver's head. At each seat a red arrow shows where the vocal is heard and a green dashed line points to the dash center. On the right are per-seat frequency response curves and a table such as "Driver tone error 1.9 dB, vocal at R 8°"](/uploads/56b4ae21f6175a38.webp)

Put on music in a car and, from the driver's seat, the singer seems to stand in the left door. Move to the passenger seat and the singer moves to the right door. In the back, what sounded like it came from the front now comes from the door next to your ear. Same speakers, same song, different seat.

Cars now compete on speaker counts: 8, 14, 23, 36. Speakers go into the roof and into the headrests. So how many, and where, for sound that is good for everyone? And what changes if you only care about the driver, or want a different song in every seat?

This time I turned a car interior into a simple box, computed the sound in it, and added a solver that picks speaker positions for a given goal.

## What I set out to do

1. Collect, from public sources, how many speakers recent cars have and where.
2. Put numbers on how a car cabin differs from a living room or a concert hall, and how it changes the sound.
3. Build a simulator that takes speaker positions and types and returns, for each seat, a frequency response and "where the vocal is heard."
4. Add a solver that picks positions for a goal (driver, all seats, per-seat zones) and a speaker count.
5. Map real cars' layouts onto the model, and turn it into a web app.

I gave the research on cars and acoustics to an AI agent and had it separate what it verified by opening the source, what it saw only in search summaries, and what it could not verify. The simulator, solver, experiments and web app were also built and run by AI agents.

## How many speakers are in a car now

Here are the counts and positions manufacturers have published.

| Car | Audio | Speakers | Confirmed positions and features |
|---|---|---|---|
| Hyundai Ioniq 5 | Bose | 8 | 2 front door, 2 tweeters, 2 rear, 1 front center, 1 subwoofer |
| Hyundai Ioniq 9 | Bose | 14 | Upmixing that spreads stereo over many channels (Centerpoint 360), BassSync to time-align several woofers |
| Kia EV9 | Meridian | 14 | Upmixing (Horizon), EQ that follows speed and road (Intelli-Q) |
| Genesis G90 | Bang & Olufsen | 23 | Headrests, headliner, motorized pop-up tweeters |
| Tesla Model 3 (2024–) | In-house | 17 (base 9) | A-pillars, rear deck, trunk subwoofer |
| Volvo EX90 | Bowers & Wilkins | 25 | 2 per front headrest, 4 overhead, tweeter on top of the dash |
| BMW 7 Series / i7 | Bowers & Wilkins | 36 | 16 tweeters, 14 mids, 4 woofers, 2 subwoofers; 4 overhead, front headrests, seat exciters |
| Mercedes S-Class | Burmester | 31 + 8 exciters | 6 overhead 3D speakers, 4 front near-field speakers, 18.5 L subwoofer |
| Xiaomi SU7 | In-house | 25 | 7.1.4 Dolby Atmos, driver headrest speakers |
| Nio ET7 | In-house | 23 | A 3-way cluster beside each of the four main seats, 4 overhead |

It is not just the count. The positions changed. It used to be doors, dashboard and rear deck. Now there is the roof (sound from above, the height channels of Dolby Atmos) and the headrest (right behind your head).

Headrest speakers do two jobs. One is immersion: ambient sound close to the ears. The other is sound "for that seat only." Xiaomi says the SU7 sends navigation prompts and calls to the driver through the driver's headrest. Nissan's Bose systems in the Juke and Micra also put speakers in the driver and passenger headrests.

In Korea, Hyundai's NVH Research Lab announced "Separated Sound Zone" in 2018: raising and lowering several speakers so each seat hears something different. It had been in development since 2014, with production readiness targeted in one to two years. Hyundai did not publish speaker counts or effect sizes in dB, and I found no evidence it reached a production car. In 2020 Hyundai and Harman put road-noise cancellation (RANC), which cancels road noise with opposite-phase sound, into production for the first time in the GV80. Harman is the company Samsung Electronics bought in 2017. In 2025 Hyundai Mobis showed a "moving speaker" that rotates with a rollable display: toward the driver's face when driving, toward the windshield when parked.

Some go the other way and remove speakers. In 2020 Continental and Sennheiser showed a system that uses actuators to vibrate the pillars, door trim and headliner instead of speakers, saving 75 to 90% of the space.

## Three reasons a car is hard

### 1. The box is small

A car cabin is about 3 to 4 m³. The model here is a box 2.55 m long, 1.48 m wide and 1.18 m tall (4.45 m³, without subtracting what the seats occupy). In a small box, low sound is trapped by the walls and gets louder. The same woofer was on average 6 dB louder below 80 Hz than with no walls, and 12 dB louder at 40 Hz. Car audio people call this "cabin gain." An SAE paper (Rieder, 1995) says the cabin adds gain over the first three octaves of music.

Above that, the box rings. At 2.55 m long, the first lengthwise resonance is 67 Hz. The width and height resonances fall at 116 Hz and 145 Hz. Around them the response swings by more than 15 dB.

![Left: the frequency response of one lower front-door woofer at the driver's left ear. The gray line (no walls) is flat from 100 Hz to 2 kHz and falls at the low end. The blue line (in the cabin) is over 10 dB higher near 40 Hz, dips deeply near 100 Hz and peaks at 125 Hz, swinging widely. A dotted orange line marks the first resonance at 67 Hz. Right: bars showing how much the cabin adds per band for the woofer, the dash mid and the A-pillar tweeter. Below 80 Hz all three gain 6–8 dB; at 4–12 kHz only the woofer gains +5, the others -1](/uploads/aedd542fe8c50aed.webp)

### 2. Everything reflects

The windshield, side glass and dashboard plastic reflect almost everything. Carpet, headliner, seats and people absorb the highs. So reflections are short and many. A team led by Lee Sang-Kwon at Inha University measured reverberation in five classes of passenger cars and found "about 0.04 s in most cars," a tenth of a living room (0.4–0.6 s).

When reflections overlap the direct sound, certain frequencies cancel. EQ cannot fix those dips, because they fall in different places at different seats. Fill them for one seat and they stick out at another.

### 3. Nobody sits in the middle

This is the big one. At home you sit between the two speakers. In a car the driver sits against the left door. With a pair of front door tweeters, the left one arrives 1.5 ms before the right.

1.5 ms sounds short, but it is plenty for the ear. When the same sound arrives twice within 1 to 40 ms, we hear it coming from the first one. This is the precedence effect that Haas described in 1949. So at the driver's seat, a vocal that should be in the middle is pulled to the left door.

![Two cabin plans. The difference in arrival time between the left and right channels at ear height is colored: red means left arrives first, blue means right arrives first. Left (no delay): a dashed line splits the cabin down the middle, the upper half red and the lower half blue. Right (delayed for the driver): the dashed line bends to pass through the driver's head, and most of the cabin turns blue](/uploads/a55f0e493a57ab6a.webp)

The fix is to delay the nearer speakers. Delay the left tweeter by 2.0 ms and the left woofer by 1.5 ms and left and right meet at the driver's head at the same moment (the dashed line on the right). But that line moves away from the passenger. Tune for the driver and the passenger gets worse. The "driver" and "all seats" settings in many car audio systems are exactly this choice. Hyundai's navigation "Position" setting likewise lets you touch a point on a car picture to focus the sound there.

## How the simulator works

Precise acoustic analysis splits the cabin into hundreds of thousands of elements. Here I chose a simple model that runs in seconds instead.

- **Cabin**: a rectangular box, with different reflectivity per wall. High for glass and dashboard; for carpet and headliner, low at high frequencies.
- **Reflections**: "mirror speakers" reflected in the walls, up to four times along each axis. That is 5,832 paths from one speaker to one ear. This is called the image source method.
- **Speakers**: directivity that narrows at higher frequencies according to cone size. Tweeter 2.5 cm, mid 8 cm, woofer 14 cm, subwoofer 22 cm, headrest 5 cm. Each covers its own band.
- **Blocking**: when a seatback or a knee blocks the direct path, the highs are cut. This is how a low door woofer's treble is lost in the legs.
- **Seats**: 4 in the sedan, 6 in the 3-row SUV. Measured at both ears of every seat.

There are three measures.

- **Tone error**: assuming all seats share one EQ, the size of the ripple left after that EQ (1/6 octave, 50 Hz–12 kHz).
- **Vocal direction**: where a centered vocal is heard. It is pulled toward the earliest arrival, with later arrivals weighted down (about a third at 0.7 ms late). The target is the center of the dashboard. If left-only and right-only sounds spread less than 45°, the stage counts as too narrow and is penalized.
- **Zone contrast**: when only one seat should hear the sound, the energy at that seat's ears over the energy at the other seats' ears.

The candidate positions are the usual ones in real cars: dash center and ends, A-pillars, front doors upper and lower, rear doors, rear deck (sedan) or row-3 sides and cargo area (SUV), roof front and rear, under the front seats, a subwoofer, and a pair in every headrest. 32 spots in the sedan, 38 in the SUV.

## Same layout, three kinds of processing

First I turned on 16 spots, similar to a Tesla Model 3, and changed only the processing.

![Three sedan plans with the same 16 speakers and different processing. At each seat, a red arrow (where the vocal is heard) and a green dashed line (dash center). No processing: the driver's arrow points front-left and the passenger's front-right. Off-center angles: driver 59°, passenger 56°, row 2 left 11°, row 2 right 2°. Driver mode: the driver's arrow swings toward the green line, 16°; passenger 60°. All seats: driver 50°, passenger 40°, row 2 at 5° and 0°](/uploads/d1085aa4bd98d06b.webp)

| Processing | Driver | Passenger | Row 2 left | Row 2 right |
|---|---|---|---|---|
| None | 59° | 56° | 11° | 2° |
| Driver mode (delay and turn down the near side) | 16° | 60° | 14° | 5° |
| All seats (center upmix + rear delay) | 50° | 40° | 5° | 0° |

The numbers are how far the vocal is from the dash center.

- Driver mode brought the driver from 59° to 16°. The passenger stayed bad.
- "All seats" nearly fixed the back. It does two things: it moves the centered sound into the center speaker, and it delays the rear speakers by about 5 ms so the front speakers still arrive first in the back. The front seats improved only a little, to 50° and 40°, because the nearby door speaker arrives even before the center speaker.

I also tried where to aim the tweeters. Harman's (JBL/Infinity) installation guide says not to aim A-pillar tweeters at the listener but across the dash at each other, so the near tweeter is far off-axis, loses treble, and matches the far one. In the model at 6 kHz, the left-right level gap at the driver's ears shrank from 5.2 dB to 4.3 dB, about 1 dB. But the 1.5 ms time gap stayed. The same guide adds the condition "unless you have a processor with separate left/right levels and per-channel delay." Most cars now have one.

## Letting the solver choose

The solver is simple. It tries adding each candidate spot (as a left-right pair) and keeps whichever improves the score most. It stops at the count limit or when nothing improves, then refines by swapping chosen spots one at a time. With the transfer functions precomputed, scoring one layout takes a few ms, so it runs right in the browser.

![Top left: sedan score versus number of speakers the solver used (4–16). The blue line (driver goal) falls to 1.6 at 8 and then flattens, rising slightly at 16. The orange line (all-seat goal) falls from 6.3 at 4 to 3.2 at 13 and stops. Top right bars: mean contrast for the zone goal, 11 dB with 4, 21 dB with 8, 22 dB with 12, 22 dB with 15. The three plans below are the solver's choices. Driver goal, 8: A-pillar tweeters, dash center, under-seat woofers, rear subwoofer, roof rear. All seats: center, A-pillars, rear door tweeters, subwoofer, roof rear, under-seat woofers, 10 in all. Zone goal, 12: 8 headrest speakers plus roof rear and rear deck](/uploads/e25988c056a7daf3.webp)

- With the **driver goal** limited to 4, what remained was the A-pillar tweeter pair, the subwoofer and the dash center. A front-door mid pair went in first and was swapped for the A-pillar tweeters during refinement. Raising the tweeters close to ear height keeps them clear of legs and dash, and makes time and level easy to match. We hardly hear direction in bass, so one subwoofer does.
- For the **all-seat goal**, the dash center came first. That is what upmixing such as the Ioniq 9's Centerpoint 360 or the EV9's Horizon does: play the centered sound from a center speaker and it is more likely to sound centered from every seat.
- **Adding speakers stopped helping at about 13.** The sedan all-seat goal reached 3.20 at 13 and did not improve with more. The driver goal was already 1.60 at 8, and 16 was slightly worse at 1.68. The 3-row SUV all-seat goal stopped at 9. Judged only by what this model measures (a clear centered vocal, even across seats), 20 or 30 speakers are not needed.
- With a limit of 8, the **zone goal** kept only headrests (a pair per seat, 21 dB). After that it gained only about 1 dB.

So why 36? Because they do things this model does not measure: overhead height channels (Dolby Atmos), surround sound wrapping from behind the ears, prompts for one seat, headroom at high volume, and grouping several drivers at one spot to shape directivity. Many speakers means "more jobs" more than "more even."

## Real cars' layouts in the model

I mapped the published counts and position notes onto the model's candidate spots. Where positions were not public, I filled them with a common layout, and several drivers at one spot became one. So the BMW i7's 36 became 28 spots in the model, and the G90's 23 became 22.

![Plans of seven cars. Hyundai Ioniq 5 (8 spots: center, A-pillar tweeters, front door woofers, rear door woofers, subwoofer), Tesla Model 3 (16), Genesis G90 (22, including front headrests and roof), Xiaomi SU7 (22, driver headrest), BMW i7 (28, including under-seat woofers), Kia EV9 (12, 3-row SUV), Volvo EX90 (24, 3-row SUV, front headrests and roof). Under each: "driver-mode image error" and "zone contrast." Ioniq 5 21° and 10 dB, G90 17° and 18 dB, i7 17° and 18 dB, EV9 22° and 11 dB](/uploads/092a6dfbbaaec4e5.webp)

This is not a ranking. Real cars differ in every driver and every tuning, and the model knows none of that. What it shows is the character of each layout.

- **Driver-mode image error** was not very different between the 8-spot Ioniq 5 (21°) and the 28-spot i7 (17°). Tuning for one driver is mostly a few front speakers and delays.
- **Zone contrast** split between cars with headrest speakers (G90, SU7, i7, EX90: 17–18 dB) and without (Ioniq 5, EV9: 10–11 dB). For per-seat sound, headrests matter before count.

## A different sound for each seat: sound zones

Finally I tried making only the driver loud and the other seats quiet. The method is "acoustic contrast control," formulated in 2002 by Choi Jung-Woo and Kim Yang-Hann at KAIST. Each speaker's level and phase are chosen to maximize the energy at the bright seat's ears divided by the energy at the other seats' ears. Mathematically, it is finding the largest eigenvector of a pair of matrices.

![Left: contrast by frequency with the driver as the bright seat. Faint lines (head exactly in place) reach 40–55 dB above 1 kHz with headrests. Bold lines (head moved 3 cm): 8 headrest speakers average 20 dB, 8 headrest plus 4 door woofers also 20 dB, 8 door speakers 10 dB. A band at 12–15 dB is labeled "published in-car results." Right: SPL maps when only the driver should be loud, at 126 Hz and 1,008 Hz. At 126 Hz the whole cabin is evenly dark; at 1,008 Hz only the area around the driver's head is dark and the rest is patchy and light](/uploads/da763d2057a3a95c.webp)

- **With the head perfectly still, it reaches 40–50 dB.** The model's head never moves a millimeter, so on paper the cancellation is nearly perfect.
- **Move the head 3 cm and it drops to 20 dB.** The weights were set for the head in place and measured at three positions moved 3 cm forward, sideways and diagonally. 8 headrest speakers averaged 20 dB; 8 door speakers, 10 dB.
- **Real car experiments are lower.** A Southampton team (Cheer et al., 2013) used 4 regular speakers below 200 Hz and headrest arrays above it and got about 15 dB. A KAIST–Hyundai in-car experiment with 4 rear headrest speakers reached 12.4 dB over 88–4,525 Hz.
- **Bass does not split into zones.** At 126 Hz the whole cabin is evenly dark. The wavelength is 2.7 m, longer than the cabin. So real systems share the bass and split only mids and highs by seat.

20 dB means the next seat's music sounds "quieter," not "gone." That is enough to make navigation prompts louder for the driver, but not enough for four people to each hear their own music as if on headphones.

## Try it

```map /embed/cabin-speakers-2026-en.html 16:10
Car Speaker Placement Lab. A sedan and a 3-row SUV cabin seen from above. Dashed circles are candidate spots; click to turn them on or off. Pick a goal (driver, all seats, seat zones) and a count, then press "Run solver" to have it choose. You can load seven real cars' layouts and overlay an SPL map, the arrival-time gap or a zone map. "Wavefronts" shows the first sound from the left and right speakers spreading out, slowed to 0.6 seconds per millisecond.
```

On a narrow screen, **[open it in a new window](/embed/cabin-speakers-2026-en.html)**. At each seat, the red arrow is where a centered vocal is heard and the green dashed line is the dash center. "Tone error" in the table is the ripple left after the EQ all seats share. In the driver goal, each speaker shows its delay in ms.

![The Genesis G90 layout loaded with the seat-zone goal. At 800 Hz the map is darkest navy around the driver's head and lighter elsewhere. The chart on the right shows the driver's contrast by frequency: gray (head still) 30–50 dB, orange (head moved 3 cm) 15–25 dB. The table lists contrast by bright seat: driver 22.5 dB, passenger 22.5 dB, row 2 left 14.6 dB, row 2 right 14.3 dB](/uploads/8ea1e3c4787a9dac.webp)

![The solver run on the 3-row SUV with the all-seat goal and a limit of 14. A 141 Hz SPL map is underneath; 9 speakers are on: dash center, front and rear door tweeters, rear door woofers, row-3 side mids. Below: "Solver picked, in order: dash center → front door woofer (pair) → rear door tweeter (pair) → … ⇄ A-pillar → front door tweeter." The front door woofers were picked first and swapped out during refinement. In the table, row 2 is 0° off center, row 3 is 0–5°, the driver 52°](/uploads/d70a834158e1b700.webp)

## Where it went wrong

- **At first the solver chose a single center speaker.** In the first version, the driver-goal solver picked "dash center + rear door tweeters + roof." With only the center on the front stage, the vocal was always centered, but left and right were gone. I had measured "is the centered vocal centered" but not "how wide is the stage." I added a penalty when left-only and right-only sounds spread less than 45°.
- **Small speakers made bass.** The 9 cm center speaker alone appeared to produce 50 Hz, because cabin gain plus unlimited EQ boost lifted the bass. A small cone cannot move that far. I cut each speaker's low end with a steep 4th-order slope and limited EQ boost to +6 dB.
- **Headrest speakers ruined the image.** Feeding them the left and right channels directly made them arrive first and loudest, right behind the ear, so the vocal came from the back of the head. Real systems play low-level ambience in the headrests for music. I changed the model to do the same.
- **I measured the image only in the treble.** At first the image used a single 3 kHz band. Layouts where only tweeters played left and right and only the center played the mids then scored well. The body of a voice is around 1 kHz. I switched to the average of 1 kHz and 4 kHz.
- **Sound zones came out at 50 dB.** That was with a fixed head, far from the 12–15 dB of real experiments. I changed it to measure at three positions with the head moved 3 cm, and limited (regularized) how hard speakers may cancel each other. It is still 20 dB, about 5 dB above the experiments. Real cars have vibration, speaker variation and reflections off people's bodies that the model lacks.
- **The first computation took 28 seconds.** That is 32 sedan spots × 8 ears × 5,832 paths × 216 frequencies, unusable if done every time the page opens. I reduced the frequencies to 108 points at 1/12 octave and baked the precomputed transfer function table into the page (1.2 MB). Only the SPL map is computed on the spot, at a lower order.

## Why this works in one click

The principles are old. The image source method of stacking mirror speakers was written up in 1979, and the cone directivity formula is in textbooks. The precedence effect dates from 1949 and acoustic contrast control from a 2002 paper. Stitching them together and running them hundreds of times with different counts and goals is what machines are good at. Scoring one layout takes a few ms; a solver run, under a second.

What does not work in one click is the definition of "sounds good." This post counted four things: tone error, vocal direction, stage width and zone contrast. Car audio companies have dozens of people sit in the same car and listen and tune for weeks. What counts as "immersive," how much pull is acceptable, where to balance treble and bass, is set by ears more than by measurements. The model only points the direction for a layout.

## If you want to try this

- **For one driver, start with delay.** Delaying the near-side speakers by 1.5–2 ms and matching levels alone brought the image from 59° to 16°. This is what aftermarket DSP "time alignment" does.
- **For all seats, center plus rear delay.** Move the centered sound into a center speaker and make the rear speakers arrive after the front. The pull in the back nearly disappears. There is no way to fix both front seats at once.
- **Tweeters high, woofers low.** Treble is blocked by legs and dash and is directional. Bass hardly is.
- **For sound zones, start with headrests.** Door speakers give about 10 dB. And account for heads moving: numbers computed with a fixed head come out 20 dB or more too high.
- **Precompute the transfer functions.** Layout search is just sums of the same computations, so with each spot's transfer function to each ear, thousands of layouts take moments.

## What it can't do

- **The cabin is a box.** A real car has a raked windshield, seats and people taking up volume, and a trunk connected behind. Resonances and reflection directions differ. COMSOL's sedan example has a 3.35 m³ interior with its first resonance at 111 Hz. The model is 4.45 m³ and 67 Hz.
- **Wall reflectivity is assumed.** Not measured per material, just "high for glass, low for carpet."
- **Reflections stop at four per axis.** Resonance heights at very low frequencies are less accurate, and high-frequency scattering is absent.
- **The image is an estimate.** The precedence effect is a simple weighting, not fitted to listening tests.
- **It knows nothing of driver specs or tuning.** Real cars' layouts carry only counts and position notes. Do not use it to compare cars.
- **It does not measure engine, tire or wind noise, distortion at high volume, or the height effect of Dolby Atmos.**

Still, why a car sounds different in each seat becomes something you can hold. The main reason is not the number of speakers. It is that nobody sits in the middle. So much of car audio is less about adding speakers than about deciding whom to tune for. Next time you get in a car, move the sound position to the driver's seat in the audio settings. You will hear the singer walk from the left door to the middle of the dash.

### If you enjoyed this, please leave a comment.

## Sources

- Car audio: Genesis G90 [genesis.com](https://www.genesis.com/kw/en/models/luxury-sedan-genesis/g90/technology.html), Ioniq 5 [CarExpert](https://www.carexpert.co.nz/hyundai/ioniq-5/2025/technology), Ioniq 9 [Bose](https://automotive.bose.com/vehicles/hyundai/ioniq-9), Kia EV9 [Meridian](https://www.meridian-audio.com/media/news-events/new-kia-ev9-receives-ev-optimised-meridian-audio/), BMW 7 Series [What Hi-Fi](https://www.whathifi.com/reviews/bowers-and-wilkins-diamond-surround-sound-system-bmw-7-series-2024), Mercedes S-Class [WinFuture](https://winfuture.de/news,126092.html), Volvo EX90 [What Hi-Fi](https://www.whathifi.com/news/bowers-and-wilkins-brings-the-abbey-road-studios-sound-to-volvos-new-ex90-electric-suv), Tesla Model 3 [CarExpert](https://www.carexpert.co.nz/tesla/model-3/2025/technology) and [TeslaTap](https://teslatap.com/articles/audio-systems-for-the-model-3/), Xiaomi SU7 [Dolby](https://professional.dolby.com/music/dolby-atmos-for-cars/xiaomi/) and [RouteNote](https://routenote.com/radar/xiaomi-car-audio-how-good-are-the-su7-and-yu7-sound-systems/), Nio ET7 [NIO](https://www.nio.com/news/nio-et7-comes-standard-dolby-atmos)
- Seat setting: [Hyundai navigation manual (sound settings)](https://ownersmanual.hyundai.com/ivi/ccNC/AVNT/MID/English/Settings_sound.html)
- Sound zones: Hyundai SSZ [Hyundai press release, Aug 2018](https://www.hyundai.news/uk/articles/press-releases/hyundai-motor-company-showcases-next-generation-separated-sound-zone-technology.html), Choi & Kim 2002 [JASA 111(4)](https://dspace.kaist.ac.kr/handle/10203/6341?mode=full), Cheer, Elliott & Simón Gálvez 2013 [JAES 61(6)](https://eprints.soton.ac.uk/354606), KAIST–Hyundai in-car experiment [KAIST repository](https://koasas.kaist.ac.kr/handle/10203/204404), Bose headrest [Bose press release](https://www.bose.com/pressroom/bose-personal-sound-system-debuts-in-all-new-nissan-micra), Harman ISZ [What Hi-Fi](https://www.whathifi.com/news/ces-2015-harman-reveals-halosonic-noise-cancelling-and-isz-in-car-technologies)
- Korea: Hyundai–Harman RANC [Automotive World](https://www.automotiveworld.com/news-releases/hyundai-motor-company-and-harman-international-launch-the-worlds-first-road-noise-cancellation-system-into-production), Hyundai Mobis moving speaker [Hyundai Motor Group](https://www.hyundaimotorgroup.com/ko/news/CONT0000000000169292), Samsung's Harman acquisition [SamMobile](https://www.sammobile.com/2017/03/11/samsungs-harman-acquisition-complete-will-retain-workforce-and-brands/), speakerless audio [Continental](https://continental.com/en/products-and-innovation/innovation/infotainment-in-the-car/best-sound-without-speakers)
- Principles: cabin gain [Rieder 1995, SAE 950181](https://saemobilus.sae.org/papers/low-frequency-design-considerations-automotive-audio-systems-950181), passenger car reverberation [Lee et al. 2004, Transactions of the KSNVE](https://koreascience.or.kr/article/JAKO200411922205515.page?lang=en), cabin example [COMSOL Car Cabin Acoustics](https://doc.comsol.com/6.1/doc/com.comsol.help.models.aco.car_cabin_acoustics/car_cabin_acoustics.html), tweeter aiming [JBL/Infinity Improving Imaging](https://fr.jbl.com/on/demandware.static/-/Sites-masterCatalog_Harman/default/dwae55737b/pdfs/JBL_INF_Improving_imaging.pdf), image source method Allen & Berkley 1979 (JASA 65(4))
- The simulator, solver, web app and charts were all made for this post

## Image credits

- All figures and screenshots were made for this post

