How our tests work
What each tool actually does, what a browser can and cannot measure, and the claims we refuse to make.
The rule every tool follows
Test → Diagnose → Fix. A tool first checks whether the hardware misbehaves, then explains what the result means, then gives one next step. If the honest answer is “a sound cannot fix this”, the tool says so and points to repair.
Audio: the tones we play
All sound is generated live with the Web Audio API. Nothing is streamed or recorded, and levels are capped so tones do not clip.
| Tool | Signal | Why |
|---|---|---|
| Water Eject | Sine starting at 165 Hz, gliding to 250 Hz and back, five seconds each way. 30 s, 60 s, 90 s, or 30 s × 3 with two-second rests. | Each phone speaker resonates at a slightly different frequency. A slow glide spends time near each one instead of betting on a single tone; short bursts limit heat. |
| Speaker Cleaner | Dust mode steps through several frequencies and ends on a sweep; Clean holds a mid tone for 30 s; Vibrate sits lower. | Moving the diaphragm at several speeds gives loose particles more chances to shift when the grille faces down. |
| Ear Speaker Cleaner | Same modes with a lower level cap. | The earpiece driver is small and saturates quickly. |
| Speaker and Headphone Test | Tones that play only on the left or only on the right, plus a frequency sweep. | One-sided tones isolate one channel; the sweep exposes bass or treble dropouts a midrange beep would miss. |
| Bass Test | Steps from 200 Hz down to 20 Hz. | The point where the tone disappears is the practical bass limit of that speaker in that room. |
A tone can only move water or loose dust sitting in or on a speaker grille. It cannot dissolve residue, reach water inside the phone, or repair a torn diaphragm, corrosion or a loose connector.
Mouse and keyboard: what browser timing can measure
Input tests read the events your browser receives and their timestamps. Browsers round timers for privacy, typically to 0.1 ms in a normal page, so every result is limited by that precision.
- Double clicking is flagged when two presses of the same button arrive closer together than a person can click (around 80 ms), or a held button drops for a moment.
- Mouse polling rate uses raw or coalesced pointer events where the browser provides them (Chrome and Edge). If a browser groups movement into one event per frame, the test detects that and says so rather than reporting your refresh rate.
- Keyboard polling rate is inferred from how key timings line up on a grid. It resolves up to about 2000 Hz; 4000 and 8000 Hz keyboards cannot be confirmed in a browser.
- Keyboard latency covers only the time after a key press reaches the computer. The keyboard’s internal delay needs hardware to measure.
Sources and updates
Device and symptom guidance is checked against manufacturer support pages (Apple, Samsung, Google, Microsoft); browser behaviour is checked against MDN and standards documents. 27 of 29 tool pages list their sources at the bottom, and each shows the date it was last updated. When a manufacturer changes its advice, we update the page rather than adding a new one.
What we will not claim
- No success rates such as “98% of speakers fixed” — we have no way to measure that honestly.
- No “100% safe” promises. Every cleaning tool caps its level and tells you when to stop.
- No diagnosis we cannot observe. Where a browser cannot see something, the result says so.
Report a wrong result
If a tool gives a result that does not match your device, contact us with the tool, the device and the browser. Reports like these are how the limits above were found.