Test your microphone’s background noise before a call

Set up a comparison you can trust

1. Open the test, enable microphone access and choose the input you actually intend to use. Check the visible microphone name rather than assuming the default is correct.
2. Put the microphone in its normal position and keep your system input volume fixed. Preview the level to check that the input is working.
3. Start the test. During the quiet step, avoid speaking, typing, handling the microphone or changing settings. Keep the page visible.
4. Speak normally during the voice step. Use a similar distance and speaking level when you repeat the test.
5. Listen back. Then choose Compare a change, alter one condition and repeat. Useful comparisons include a fan on versus off or a window open versus closed.

Keep the microphone, gain and position unchanged when you want to isolate a room change. If you intentionally change microphones or gain, treat that as a different setup rather than a controlled room comparison. The tool cannot verify your physical position or every operating-system setting.

dBFS describes the captured digital signal

The displayed levels use decibels relative to digital full scale, abbreviated dBFS. With the tool’s amplitude-1 reference, a more negative reading means a lower captured level. A change from −40 dBFS to −46 dBFS is a 6 dB reduction in this measurement. It is not a room sound-pressure reading or a percentage reduction in perceived loudness.

A single knock can make the loudest moment stand out without describing the whole recording. A steady fan can affect the average and baseline. Read the values together, then use playback to find out what you actually captured. The test does not identify a fan, traffic or speech as a sound source.

Work through a fan-on / fan-off example

Imagine that Test A captures an average of −40 dBFS with a desk fan running. Test B captures −46 dBFS after you switch the fan off, while the microphone position and gain stay fixed. The measured difference is −6 dB. These are illustrative numbers, not measurements from your room or a predicted result.

A · Fan on    −40 dBFS
B · Fan off   −46 dBFS

B − A = −46 − (−40) = −6 dB

Listen to both recordings at the same comfortable playback volume. Check whether the steady noise changed and whether any unexpected event explains the difference. If a vehicle passed during only one sample, repeat the pair before drawing a conclusion. A lower average is useful evidence of lower captured level; the recording tells you whether that change addresses the distraction you care about.

Account for processing between microphone and browser

The test requests echo cancellation, noise suppression and automatic gain control to be disabled, then reports the settings exposed by the browser. A request is not proof that the whole capture chain is unprocessed. The Media Capture specification distinguishes requested constraints from the current settings returned by getSettings(); hardware and operating-system processing can introduce further uncertainty.

Use the processing notes beside the results. Enabled or unknown processing qualifies what a comparison can establish. Your meeting application may apply its own processing, so this browser recording cannot guarantee what another application sends to the people on your call. After improving the setup, make a short recording or test call in the application you will actually use.

Reference: W3C Media Capture and Streams: constraints and current track settings

Use the voice assessment as a prompt to listen

The speaking step lets the tool estimate a voice-to-background margin and highlight quiet-step moments that may be distracting. Its labels use product heuristics, not a certified intelligibility or call-quality standard. The calculation cannot recognize whether the speaking step contains your voice or another loud sound, so follow the instructions and listen to the result.

• If the voice is not clearly above background, speak normally, check the selected input and try moving closer before retesting.
• If clipping is reported, reduce microphone input gain and repeat both recordings. A clipped signal is not a valid basis for the numeric comparison.
• If no usable signal is detected, check permission, mute controls and input selection. Digital silence does not establish a quiet room.
• If capture is interrupted or incomplete, repeat the measurement with the page visible rather than comparing partial samples.

Keep the useful result without uploading audio

Copy summary provides a readable result; Download JSON provides measurement definitions, results, processing flags and the level history. JSON excludes the audio, recording URLs, microphone names and device identifiers. Your test labels remain in the report, so choose labels you are comfortable sharing and review an export before sending it to someone else.

Recordings remain in page memory for playback and are cleared when you start again, refresh or leave. You can retain an exported comparison as a note for your own setup, but the test is not continuous monitoring or a saved recording library. Repeat it when your microphone, room or input settings change.

Technical references

These primary sources explain the underlying formats and browser APIs. Tool-specific behavior and limits above describe the current Buddy Code implementation.

• W3C Media Capture and Streams: constraints and current track settings