Reverse Engineering AirPods Self-Calibration

ateliti99 · ateliti99 ·

9 min read Original article ↗

After you close the case, AirPods Pro can play a quiet burst of tones. Apple already confirms it as a self-check of the earbuds' speakers and microphones, added together with the hearing-health features (Hearing Test and Hearing Aid) in iOS 18.1. I took the sound apart and compared at how hearing devices are normally calibrated in a lab, to understand why AirPods do this trick.

Here the sound:

Since late 2024, owners of AirPods Pro 2 have been reporting a strange, faint "chime" coming from the closed charging case. It usually happens in a quiet room, a little while after the earbuds have been put away. One user on the Apple Community forum describes a "very quiet, high pitch tone" that happens "after the AirPods have been put back in their case and have been stationary for a while".

Apple answers the question in its own user guide:

"To help ensure that your AirPods microphones and speakers are operating at their best (for example, to help provide high-quality hearing test results), your AirPods may periodically play a quiet chime when they're in their charging case."
— Apple, Use hearing health features with AirPods Pro 2 or AirPods Pro 3

Note the wording! It is the AirPods that play the sound, not the case

The story was discussed it in several places:

So the sound is real and intentional. The interesting questions are what it is made of and why it is needed.

All sound pictures below are spectrograms. Read them in the following way:

  • left to right is time;

  • bottom to top is the played frequencies;

  • darker means louder.

How to read a spectrogram
A pure tone, like a whistle or a tuning fork, is a thin horizontal line. Noise ("shhh") is a fuzzy smear. A click is a thin vertical line.

A few units you will meet:

  • Hz (hertz): vibrations per second, i.e. pitch. 1 kHz = 1,000 Hz, roughly a high whistle.

  • Pure tone: a sound with a single frequency and nothing else.

  • dB (decibel): a scale for loudness differences. +6 dB is about twice the sound pressure, and +10 dB sounds is about twice as loud.

A short video shared on Reddit captures exactly this moment. Someone opens an AirPods Pro case, closes it, and leaves it alone on a desk. I extracted the soundtrack of that video and analysed it.

Frames from the video
Frames from a video shared on Reddit: (1) the lid is open, (2) the lid is closed and the green status light comes on, (3) the case sits untouched while the sound plays.

Here is the whole 27-second soundtrack as a spectrogram. I removed the steady background hum of the room so that anything new stands out:

Timeline of the recording
The lid closing is a loud click (a vertical line). About 16 seconds later, while the case sits still in a quiet room, a short staircase of tones appears (orange box).

Zooming in on the orange box:

The tone staircase
Six pure tones, one after the other: 1, 2, 3, 4, 6 and 8 kHz, each lasting 100 milliseconds.

What the measurements show:

  • Exact round numbers. Every tone lands within about 1 Hz of its round value.

  • One tone at a time. In each 100 ms slot only that slot's pitch rises above the background.

  • No gaps, no fades. Each tone switches straight to the next one in less than 5 ms. The whole sequence lasts 0.6 seconds.

  • Very quiet. In the video the tones are barely above the room's background noise. Due to the AirPods case.

  • Played once. In the 27 seconds of video there is only one sequence.

The pitches are not random either: 1, 2, 3, 4, 6 and 8 kHz are standard frequencies of a clinical hearing test. Pure-tone audiometry, which also uses 125, 250 and 500 Hz, that maybe are not easy to capture by the phone or to reproduce with the AirPods.

To see why AirPods bother, we need to know what a hearing test needs.

A hearing test plays tones at different pitches, quieter and quieter, until you stop hearing them. The quietest tone you can hear at each pitch is your hearing threshold, and the chart of those thresholds is your audiogram.

This only works if the device knows exactly how loud each tone is at your eardrum. Imagine the earphone has lost 5 dB at 4 kHz. Every 4 kHz tone then arrives 5 dB quieter than the test believes. You stop hearing it sooner, and the test concludes that your hearing at 4 kHz is 5 dB worse than it really is. A hearing aid tuned from that result would then amplify the wrong amount.

That's why, in clinics and factories, earphones and hearing aids are calibrated with an artificial ear:

The standard laboratory set-up
The standard set-up (schematic). The earphone plays known tones into a metal "artificial ear" whose cavity behaves like an average adult ear canal. A precision microphone sits where the eardrum would be. The measured levels are compared with the official reference values, and the device is adjusted at every pitch.

The main standards involved:

  • IEC 60318-4: the occluded-ear simulator, i.e. the artificial ear for earphones that sit in the ear canal.

  • ISO 389-2: the reference levels, i.e. how loud each test tone must be at the artificial ear for insert earphones.

  • ISO 8253-1: how hearing tests are performed, including routine checks of the equipment.

  • IEC 60118-0: how hearing-aid performance is measured.

In a clinic the equipment is not calibrated once and forgotten. ISO 8253-1 describes regular listening checks, periodic measurements and full calibrations, because earphones drift over time.

AirPods can't go back to the lab every week, or months, or year. But they carry something almost as good: their own case.

The case as a test chamber
With the lid closed, each earbud sits in a small, sealed air pocket that has the same shape every time. The earbud plays known tones, its own microphones listen, and the chip compares the result with the measurement taken at the factory (schematic, not to scale).

Every piece of lab equipment has a stand-in. The earbud's speaker replaces the lab earphone. The sealed pocket replaces the artificial ear. The earbud's own microphones replace the measuring microphone. The chip, comparing with values stored at the factory, replaces the analyzer and corrects the settings.

The test needs the right conditions, and they explain when you hear it:

  1. Lid closed. The pocket must be sealed so it has the same shape and volume every time.

  2. Not moving. Handling noise and bumps would spoil the measurement, and the lid must stay shut.

  3. Quiet surroundings. The tones are very soft. Background noise would mix into the measurement. A soft test sound is also less likely to bother you.

  4. A short wait. In the video it started about 16 seconds after the lid closed. Apple says it happens periodically, not every time the lid closes.

In my experience the sound shows up especially at night, and that fits the conditions above. At night the AirPods are put away, the case sits still, and the room is at its quietest, which is both when the test can run and when you are most likely to notice it.

What could the test find? Here is an example with random numbers. Imagine that earwax partly blocking the speaker and steals the high pitches.

What the case test can reveal
Illustrative numbers, not measured on a real AirPod. If the measurement in the case drops at high pitches compared with the factory value, the earbud knows how much to add back, or that something is wrong.

In practice, a check like this can catch:

  • a speaker or microphone that is getting less sensitive with age or humidity;

  • a grille blocked by dust or earwax;

  • a damaged speaker, which distorts the tones;

  • a mismatch between the microphones, which matters for noise cancellation, Transparency mode and calls.

Engineers at other companies have described the some similar ideas. A Bose patent plays test signals from the earpiece's driver into the cavity of the storage case and measures them with the earpiece's own microphones, then adjusts microphone and driver sensitivity. A Google patent describes "a tone of a certain frequency, a sweeping range of tones of various frequencies, or pink noise" played in a closed case to calibrate speaker, microphone and noise-cancellation gains. Research also shows why this matters: Guo et al. (2021) ran hearing tests with consumer true-wireless earbuds and had to calibrate them carefully to get reliable results.

Apple and regulators

News coverage

Forums, recordings and sound collections

Patents and research

Standards mentioned (not freely available online): IEC 60318-4 (occluded-ear simulator), ISO 389-2 (reference levels for insert earphones), ISO 8253-1 (audiometric test methods), IEC 60118-0 (hearing-aid measurements).

The video frames come from a video shared on Reddit. The spectrograms were made from that video's soundtrack, and the drawings are schematics, not to scale.

Discussion about this post

Ready for more?