High-Frequency Hearing Test — 8–22 kHz

A high-frequency hearing test on this page plays discrete sine tones from 8 kHz to 22 kHz and records which ones you notice, building a simple heard map. Tones are synthesized locally with the Web Audio API, nothing is uploaded, and the result is informational screening rather than a medical or age verdict.

How to use it

1. Use wired headphones and keep the volume low and fixed. Wireless codecs and small speakers can remove the top octave before your ear gets a chance, and volume changes between presets ruin comparability.
2. Start with 8 kHz. Each preset plays a continuous 1.5 s sine with a 20 ms attack and an 80 ms release at a fixed gain of 0.18 (about -15 dBFS); the 8 kHz preset is your reference for what heard sounds like.
3. Press Play tone, then answer I heard it or I did not hear it. The answer fills the matching chip in the map, and you can replay a preset to change your answer.
4. Step upward through 10, 12, 14, 16, 18, 20 and 22 kHz. The readout counts heard presets out of eight; untested presets stay blank until you answer.
5. Replay the boundary pair. If 16 kHz is heard and 18 kHz is not, replay both; the transition between heard and not heard is the interesting data.
6. Repeat the map on a second device before interpreting it. A missing tone can come from the speaker, codec or sample rate as easily as from your ear; the next section lays out the separation procedure.
7. Keep the map or clear it. The last map restores from this browser only (key hf-hearing-test:last), and Clear saved result erases it.

What changes above 8 kHz

Hearing above 8 kHz is the least standardized part of the audiogram: thresholds swing widely between people of the same age and depend on ear canal, transducer and setup.

ISO 7029:2017 models threshold distributions for otologically normal listeners from 125 Hz to 8 kHz between ages 18 and 80, with informative medians extending to 9–12.5 kHz. Those statistics show thresholds drifting upward with age, and the drift is largest at the top of that range. The population-level onset of the measurable shift sits roughly in the 30s and 40s above 8 kHz. Two honest caveats: the spread between individuals is much larger than the median shift, so a clean 16 kHz tone in your 50s and a missing one in your 30s are both ordinary, and the page cannot separate age from noise history or earwax. Clinical audiometry usually stops at 8 kHz because that is where diagnostic weight concentrates (ISO 8253-1), so nothing here is an audiogram.

Device and codec limits

A missing 18 kHz tone is as likely to be a playback-chain limit as a hearing limit, because nearly every consumer audio path is filtered somewhere above 16 kHz.

Lossy encoders band-limit to save bits. Fraunhofer’s MP3 and AAC design literature recommends limiting an encoder’s frequency response to about 16 kHz where clean full-band coding is not possible, which is why a 128 kbps stream can carry no 18 kHz content at all. Bluetooth paths re-encode and resample, so usable bandwidth depends on the codec and link, not the source file. Small speakers roll off early and beam: at 17.4 kHz the wavelength is about 19.7 mm (λ = 343 / f at 20 °C), so a tone can be clear on axis and nearly absent a few centimetres to the side. Sample rates add another ceiling: 44.1 kHz reaches a 22.05 kHz Nyquist limit, while a 32 kHz path stops at 16 kHz.

Playback chain	Where it commonly stops	Why it stops
128 kbps MP3 or AAC stream	Around 16 kHz	Lossy encoders band-limit the top octave to save bits
Bluetooth audio	Codec and link dependent	The stream is re-encoded and resampled before it reaches the headphones
Phone or laptop speaker	Frequently well below 20 kHz, and directional	Tiny drivers plus beaming at short wavelengths (17.4 kHz is 19.7 mm)
Wired headphones, lossless source	Usually to the listener’s own limit	No codec stage; only driver and seal remain

Device limit or hearing limit: a five-minute check

Separating a device ceiling from your own threshold means changing one variable at a time and watching whether the missing tone moves with the device or stays with you.

1. Run the map on two independent chains. Use wired headphones on a computer, then a different phone or player. A tone that appears on one chain was never a hearing result.
2. Switch from Bluetooth to wired for the comparison. The wired path removes the codec and the negotiated link bitrate from the equation.
3. Verify the tone exists in the source. Play a lossless local file at the same frequency and check it on a spectrum analyzer or oscilloscope via an interface loopback or a measurement microphone rated above 20 kHz. A peak at 17.4 kHz means the file is fine and the question moves to the output device.
4. Change one thing at a time. Reseat the headphones, swap channels, and play a known-good 10 kHz control tone before and after; beaming and seal bite hardest at the top.
5. Read the pattern, not the single tone. A tone missing on one device and present on another is a chain limit. A tone missing on two independent wired chains is worth retesting another day and, if new or one-sided, mentioning to an audiologist. That is a screening signal, never a diagnosis.

Volume, safety and the age question

High-frequency tones sound sharp at levels that feel modest, and the natural reaction to a missing tone is to raise the volume, which is the wrong move.

Presets are fixed near -15 dBFS, but system volume and headphone sensitivity decide the real acoustic level, so start low. Chasing a tone upward pushes high-frequency energy into the driver and gives your ear nothing it needs. Keep pets out of the room for the 20 kHz and 22 kHz presets, since dogs and cats hear far above the human range. Results describe this session on this gear: a missing preset is not evidence of hearing damage, and a heard one is not proof of youth. This page is educational screening, not a medical test; if you are concerned about your hearing, or notice one-sided or sudden changes, see an audiologist or clinician.

Accuracy, limits and troubleshooting

Everything downstream of the oscillator changes the result more than the code does.

• dBFS only. Preset gain is fixed near -15 dBFS; there is no calibrated SPL or dB HL, so the page cannot say how loud anything was.
• Frequency is exact; everything downstream is not. The oscillator is sample-accurate; the DAC filter, amplifier, driver and seal decide what survives to your ear.
• OS processing can filter the top end. Windows audio enhancements, spatial sound and voice-isolation style processing can band-limit or duck high frequencies; disable them before comparing devices.
• Sample rate matters here. 44.1 kHz gives a 22.05 kHz Nyquist limit and 48 kHz gives 24 kHz, but a 32 kHz path stops at 16 kHz.
• Bluetooth is a wildcard. Codec and link quality change the top octave; treat wireless results as provisional and confirm wired.
• Small speakers are directional. At 17–22 kHz the wavelength is roughly 16–20 mm, so a centimetre of angle or distance changes level dramatically; keep your position consistent.
• In-app browsers can block audio. If Play does nothing, click once in the page; if the state reads Unavailable, use a normal browser tab.
• What this page cannot do. It cannot measure thresholds in dB HL, mask the opposite ear, test bone conduction or separate age from noise exposure; the map is informational only.

Related tools

• Hearing Age Test — The same high-frequency idea framed as a listening-age walk with preset categories.
• Online Hearing Test — Screen the 125 Hz to 8 kHz range per ear with a threshold staircase and SVG audiogram.
• Low-Frequency Hearing Test — The opposite end of the spectrum, 10 Hz to 100 Hz, with speaker-limit warnings.

Aliases covered on this page: high frequency hearing test, high frequency test, 16khz test, and upper hearing range check. Informational screening only.