How do I test my microphone?
Click “Test microphone”, allow access and say a few sentences at your normal volume. The test shows the level in dBFS with a peak hold, plus a waveform and a frequency spectrum. It tells you whether your voice arrives too quiet, fine or too loud, warns about clipping and silence, and plays back a 5-second test recording. Everything runs in your browser.
What does dBFS mean?
dBFS stands for decibels relative to full scale. 0 dBFS is the largest value a digital signal can take, so every real level sits below it; each 6 dB less halves the amplitude. The test shows the average (RMS), which is closer to how loud something sounds, and the held peak, which decides whether the signal clips.
| Signal | Peak | Average (RMS) |
|---|---|---|
| Sine tone at full scale | 0 dBFS | −3.0 dBFS |
| Sine tone at half amplitude | −6.0 dBFS | −9.0 dBFS |
| Silence | −∞ | −∞ (shown as below −60) |
The AES17 measurement standard of the Audio Engineering Society defines the scale so that a full-scale sine tone also reads 0 dB FS as an average. Software following that convention therefore shows RMS values 3 dB higher than this test; the peaks match.
How loud should my microphone be?
While you speak, the peaks should land between −20 and −3 dBFS; the test marks that range in green. These are rules of thumb from voice recording, not a standard: below it, the voice has to be boosted a lot later and the noise comes up by the same amount. Above it, a single louder syllable is enough to hit the limit. Broadcasting works with similar headroom: the EBU R 68 recommendation puts the alignment level of digital studio equipment at −18 dBFS.
The rating uses the loudest moment of the last three seconds, so pauses between words do not count as “too quiet”. With automatic gain control on, the test measures audio that has already been adjusted, not your microphone's raw level.
How does the test detect clipping?
Digital audio stops at 0 dBFS: louder waveform tops are cut flat, which sounds harsh and cannot be repaired. In every analysis window the test looks for at least three consecutive samples at 99 % of full scale or more, and it only warns when that happens in three separate windows within two seconds. A tap on the microphone therefore does not trigger it.
Why does the test hear nothing?
If pure zeros arrive for three seconds, even the noise floor every real microphone has is missing: the signal is switched off somewhere. If every peak stays below −50 dBFS for five seconds, sound is arriving, just very little. Then check, in this order:
- the mute switch on the headset or cable, and the mute key on the keyboard
- the selected device, because a laptop with a headset and a webcam often has three inputs
- the input level: on Windows 11 under Settings › System › Sound › Input, on macOS under System Settings › Sound › Input
- the browser's permission: on Windows 11 under Privacy & security › Microphone, on macOS under Privacy & Security › Microphone
One requirement no setting can fix: the W3C specification Media Capture and Streams ties microphone access to a secure context, so over plain http the microphone stays blocked.
Echo cancellation and noise suppression: on or off?
The same specification describes three stages that browsers can put in front of the microphone: echo cancellation, noise suppression and automatic gain control. For calls over speakers, echo cancellation belongs on, otherwise the other person hears themselves. For singing, instruments or a recording you want to edit, all three tend to get in the way because they change the sound. The test toggles each stage separately and shows under “What the browser reports” whether the change was actually applied.
What happens to the test recording?
It lasts five seconds and stays in the memory of the open tab. The last two sit side by side, labelled with their processing settings: record once with and once without noise suppression and compare. The WAV download is an uncompressed copy of what the browser recorded in compressed form, not a lossless raw recording.
What the test cannot do
- Measure sound pressure: dBFS describes the digital signal, not how loud it is in the room.
- Rate singing or instruments: the “too quiet” and “too loud” limits are meant for speech.
- Keep running in the background: in a hidden tab the display and detection pause, and the microphone stays open until you stop the test.