How the tests work
Frequency Test uses browser audio standards and published hearing-screening guidance to build transparent, informal tools. It is not a clinical instrument and does not claim medical calibration.
How tones are generated
The site uses the browser’s Web Audio API. An OscillatorNode creates the waveform, a GainNode controls digital amplitude, and a StereoPannerNode routes the result left, right or center before it reaches the browser audio output.
Why hearing tests use sine waves
A sine wave contains one target frequency. Square and sawtooth waves include strong harmonics, so a listener might hear an upper harmonic even when the selected fundamental is not reproduced well. Those waveforms are available only in the general tone generator, not in hearing or response tests.
Frequency range and sample rate
Output is limited to the lower of 20,000 Hz or the browser sample rate divided by two, minus a 100 Hz margin. This respects the Nyquist limit. A browser reporting that it can synthesize 20 kHz does not prove that the DAC, wireless codec, amplifier or transducer reproduces it accurately.
Logarithmic frequency mapping
Manual sliders and default sweeps use a logarithmic scale. Equal distances represent equal frequency ratios, so 20→40 Hz occupies the same scale width as 10→20 kHz. This makes low, middle and high octaves practical to explore.
Start, stop and transition safety
The audio context is created or resumed only after a user action. Gain ramps from zero over 35 milliseconds and ramps back to zero before a node stops, reducing clicks and pops. High-frequency samples stop after three seconds, other individual tones after no more than ten seconds, and all audio stops when the page becomes hidden.
Adaptive hearing threshold
The full hearing test begins with coarse points from 8 to 20 kHz. When the first “not heard” response creates a boundary, the test subdivides that interval into smaller steps. Left and right ears are tested independently. The displayed value is the highest confirmed point in this browser session, not a dB HL threshold.
Silent control trials
Each ear sequence includes silent trials at unpredictable points. A “heard” response during silence reduces the reliability score; correctly identifying silence improves it. A silent trial does not advance the frequency sequence.
Reliability score
| Input | Why it matters |
|---|---|
| Device type | Built-in speakers and Bluetooth may limit extension or add processing. |
| Quiet confirmation | Background sound masks quiet test tones, especially at lower frequencies. |
| Channel verification | Confirms that isolated left/right trials reach the intended side. |
| Silent controls | Provides an internal check on guessing and expectation. |
| Replay count | Many replays may increase expectation bias and fatigue. |
| Left/right consistency | Large differences deserve a repeat before interpretation. |
The score describes the quality of this informal session. It is not an accuracy percentage and never makes the result clinical.
Local result storage
Completed test summaries are stored in browser localStorage on the current device. No account or server database is used. Audio is generated rather than recorded, and the site does not request microphone access.
References and standards
- W3C Web Audio API 1.1 — browser audio synthesis and processing.
- MDN autoplay guidance — user interaction requirements for audible playback.
- ASHA adult hearing screening guidance — context for how professional screening differs.
- WHO safe listening guidance — volume and exposure considerations.
Last technical review: August 8, 2026.