Microphone Quality Test — SNR Score

This tool profiles a microphone chain the way a broadcast engineer profiles a studio: it samples three seconds of room silence to find the noise floor, then four seconds of your voice to measure signal-to-noise ratio (SNR) and clipping headroom, and combines them into a 0–100 quality score. Every calculation runs locally in your browser through the Web Audio API; nothing is recorded or uploaded.

How to use it

1. Pick the exact input you will record with. USB interface, XLR preamp output, webcam mic, headset: choose it in the device list before measuring, because each source has its own noise character.
2. Disable all automatic processing. Turn off OS noise suppression, “audio enhancements”, and automatic gain control. macOS puts a Voice Isolation toggle in the Control Center while a mic is active; Windows hides per-device effects under Settings > System > Sound > device properties. Variable gain between runs makes scores incomparable.
3. Set up your real recording position. Headphones on, mouth 15–20 cm from the capsule, mic slightly off-axis from your breath. Measuring from a distance you will never record at produces a score you cannot use.
4. Run stage one: three seconds of silence. Do not type, breathe into the mic, or touch the desk. Decide before you start whether the room fan stays on; the measurement captures whatever is really there.
5. Run stage two: four seconds of steady speech. Count from twenty to thirty or read a paragraph at your actual performing loudness. Watch the live meter and aim for peaks landing between roughly −12 and −6 dBFS.
6. Read the four cards, fix the weakest metric, and re-run. The composite score weights the noise floor at 35%, SNR at 40%, and headroom at 25%, with any clip event penalized directly.

What each metric actually measures

The noise floor, in dBFS

The floor is the sum of three things: acoustic room noise arriving at the capsule, the microphone’s own electrical self-noise, and the preamp’s noise added on top. Quality condenser capsules list self-noise around 7–20 dB(A) measured per the IEC 60268-4 microphone specification method, while budget laptop MEMS capsules commonly sit nearer 30–40 dB(A), before you add any room or preamp contribution.

One distinction matters more than any other here: dBFS is a digital ruler, not a sound pressure reading. The same room measured with the input gain two steps higher will show a “worse” noise floor in dBFS even though the acoustic environment has not changed. The meter is honest only when you hold gain fixed, which is also the only way two test sessions are comparable. No browser tool converts these numbers into calibrated dB SPL; that requires a measurement microphone with a published sensitivity and a calibrator.

Signal-to-noise ratio

SNR is your voice’s peak level minus the noise floor, both in dBFS. A voice peaking at −12 dBFS over a floor of −62 dBFS gives 50 dB of SNR, plenty for any platform. Two levers dominate it: distance and silence. Moving from 40 cm to 20 cm from the mic roughly doubles the direct signal (about +6 dB by the inverse-square relationship) while the room noise contribution stays flat, buying you the entire difference in SNR. Killing a fan buys the rest.

Headroom and clipping

Speech has a crest factor of 12–18 dB between its loud peaks and its average level. Peaks that reach the 0 dBFS digital rail truncate: the waveform loses its shape permanently, and no noise gate or EQ restores a clipped transient. Steady peaks below −6 dBFS leave 6 dB of insurance for the laugh, the shout, and the doorbell you did not plan for.

Score bands and what they predict

Score	Grade	Typical noise floor	Typical SNR	Real-world outcome
90–100	Studio grade	below −60 dBFS	above 35 dB	Voiceover and post-production clean; >35 dB SNR with zero clipping
75–89	Broadcast / podcast	−60 to −50 dBFS	25–35 dB	Streams, YouTube, interviews pass without processing
55–74	Workable	−50 to −40 dBFS	15–25 dB	Needs a gate or de-hummer; hiss audible on headphones
< 55	Marginal	above −40 dBFS	under 15 dB	Fan rumble and whine dominate; fix the room or interface first

For reference, a 16-bit converter’s theoretical ceiling is about 96 dB of dynamic range (6.02 dB per bit plus 1.76), and 24-bit sits around 144 dB; real-world numbers land 10–20 dB lower because of analog noise, so a −60 dBFS floor with a −10 dBFS voice peak is close to what consumer hardware can genuinely deliver.

Accuracy, limits and troubleshooting

Gain is the hidden variable. The score measures your whole configured chain, not the microphone alone. When comparing two mics, set each one’s input gain so your voice peaks at the same level, otherwise you are ranking preamp settings.

A high floor usually means the room, not the gear. Close the window, silence the HVAC, and re-measure before buying anything. Traffic rumble around 40–100 Hz and 50/60 Hz mains hum from bad grounds are the two most common fingerprints; watch the spectrum analyzer page to tell them apart.

A low SNR with an acceptable floor means distance or output level. Get closer with a pop filter 5–10 cm off-axis, or raise the interface gain two steps.

Built-in 3.5 mm motherboard jacks couple noise from the machine itself (GPU whine, ground loops). A USB interface with a proper preamp typically improves the score by 10–20 points on the same microphone.

Two microphones in one source? Separate them at least three times the distance between them (the classic 3:1 rule) to limit comb filtering from the doubled signal.

Scores drift with the building. At night, offices read 5–10 dB quieter than at noon. Note the time of day when you baseline a setup.

Related tools

• Mic Test — The 30-second functional check: does the right device capture a clean signal at all.
• Decibel Meter — Measure the room’s acoustic environment in dB SPL alongside this tool’s dBFS view.
• Room Acoustic Test — Find the standing waves and flutter echo that raise your noise floor.
• Mic Loopback — Hear exactly what your microphone is transmitting, in real time.