Mic Quality Test: 3 Numbers That Show Whether It's Your Mic or Your Room

The mic test you just ran in Chrome was graded on a curve. As of November 2026, Chrome by default switches on echo cancellation, noise suppression and automatic gain control for microphone streams unless a page asks otherwise. That clean playback mostly tells you how good Google's filters are. Your podcast recorder gets the raw signal.

Read three numbers off a free waveform view instead. It takes about two minutes and shows what your mic and your room really sound like.

Our take: most people who hate their sound buy a new mic when the room is the real problem. As of November 2026, ANSI/ASA S12.60, the US standard for classroom acoustics, sets a limit of about 0.6 seconds of reverberation for small unoccupied classrooms, a room designed for speech. A bare home office with hard floors can ring longer than that. A $150 USB mic won't fix a room. The clap test tells you which problem you have before you spend a dollar.

How do you test your microphone quality with real numbers instead of just listening back?

Listening back fails twice over. Within seconds your ears tune out the hum of your own room, and the laptop speakers or earbuds you play the take on hide hiss and boom.

Get the setup right first:

• Turn processing off (the browser section below shows where).
• Set input gain so loud speech peaks between -12 and -6 dBFS and never touches 0.
• Put the mic where it will really sit, and close every other app that might grab it.

Record about a minute in Audacity, which is free on Windows and Mac, and save it as an uncompressed WAV. A phone voice memo compresses the file and smears the quiet parts. The run: sit still for 10 seconds, read for 20 to 30 seconds in your normal call or show voice, then give one sharp clap and keep quiet for 3 seconds.

Now read your three numbers:

1. Noise floor: the average level of the silent stretch, in dBFS. Select it and run Analyze > Contrast in Audacity to get an RMS reading.
2. Speech-to-noise: your typical speech peak minus the noise floor, in dB. Ignore the odd spike.
3. RT60: how long the clap takes to die away, in seconds (method below).

Say your floor reads -56 dBFS and your speech peaks at -8 dBFS. The gap is 48 dB, which passes for a podcast. The floor is only borderline, so the hiss needs fixing, not your voice level. Turning yourself up won't help.

That raises the real question: where do the lines sit, and who decided?

What readings pass for a call, and what does a podcast demand?

Every line below is tied to a published spec or to how the audio actually gets heard. None of it is a hunch.

Noise floor. As of November 2026, ACX, Audible's audiobook production platform, asks for a noise floor no higher than about -60 dB RMS. No public spoken-word spec is stricter, so that's where the podcast pass line sits. A call can pass at -50 because Zoom's noise suppression and its voice codec cover up steady hiss.

Speech-to-noise. As of November 2026, ANSI/ASA S12.60 targets a speech level about 15 dB above background noise in classrooms. Speech peaks run roughly 10 dB above the average level of speech. That makes a gap of about 25 dB between peaks and floor a rough minimum for being understood on a call, as of November 2026. Sounding clean takes far more. As of November 2026, podcasts are commonly normalized to about -16 LUFS and often played on earbuds, so any hiss gets turned up along with your voice. Hence the 45 dB podcast line.

Echo. As of November 2026, the roughly 0.6-second classroom limit works for a live listener, whose two ears and eyes help filter the echo out. A recording strips those cues away, so treated voice rooms aim for about 0.3 seconds.

The most common mistake is turning gain down to get a better floor. The floor drops, but your voice drops with it, so the gap that matters stays put. Fix the source instead:

• Floor fails: shut off the noise source (fan, fridge, a laptop next to the mic) or switch to a dynamic mic used up close.
• Gap fails: move the mic closer before you touch gain. Each time you halve the distance, your voice gains about 6 dB over the room.
• Echo fails: add soft surfaces (a rug, curtains, a closet full of clothes) before you buy anything.

The results table below sets the lines for each use, and the scorecard under it turns your three numbers into a grade. The echo number is the one people skip, and it needs nothing more than your hands.

How do you measure room echo with a hand clap, and what RT60 should you aim for?

RT60 is the time a sound takes to fall 60 dB after it stops, and ISO 3382-2 is the standard for measuring it in ordinary rooms. The catch: a hand clap at home almost never gets 60 dB above the noise floor. So you time the first 20 dB of decay, called T20, and multiply by 3. The standard extrapolates the same way.

As of November 2026, ISO 3382-2 defines T20 as evaluated from 5 dB to 25 dB below the starting level. Timing from the clap's peak down to -20 dB is a fair shortcut at home and runs slightly short.

Say the clap peaks at 1.200 s and the tail is 20 dB lower at 1.340 s. T20 is 0.14 s, so your RT60 is about 0.42 s. That's borderline for a podcast and a pass on Zoom.

Clap in two different spots and average the results. If the tail rattles and repeats, you'll see spikes in the decay. That's flutter echo bouncing between bare parallel walls. A bookshelf or a hung blanket breaks it up, and it's worth doing even when the number passes.

Why does your mic sound fine in a browser test but bad in a recording?

As of November 2026, the W3C Media Capture and Streams specification lets a web page request settings such as echoCancellation and noiseSuppression for a microphone, with autoGainControl alongside them. As of November 2026, Chrome, Edge and other Chromium browsers turn them on by default.

Each one bends your test result:

• Noise suppression removes the hiss your recording will still have.
• Auto gain pushes quiet speech up and pulls loud speech down, which hides bad gain settings.
• Echo cancellation can clip the ends of words and makes the room sound drier than it is.

The same filters explain why a mic can sound fine on Zoom and awful in a podcast recorder that captures it raw. Neither one is wrong. They're measuring different things. Chrome has no simple setting to turn these off, so run the test in a local app like Audacity. The table below shows where to switch each filter off in Zoom and in your operating system.

Why does your Bluetooth headset mic sound so bad on calls?

It's the protocol, not the brand. As soon as an app opens the mic, the headset drops out of A2DP, the music profile, and into the Bluetooth SIG's Hands-Free Profile (HFP). HFP carries audio both ways over a much thinner link.

Classic HFP uses the CVSD codec at an 8 kHz sample rate, which as of 2026 limits voice to roughly 4 kHz. That's landline-phone sound. HFP 1.6 added the mSBC codec at 16 kHz, known as wideband, which as of 2026 works only when both devices support it. Your music and the other callers drop to call quality too, and people blame that on the mic as well.

To check, record the same passage through the headset mic and through the laptop's built-in mic. If the headset take sounds muffled and its noise floor looks suspiciously good, you're hearing narrowband compression, not a quiet mic.

If your laptop has a mic, the fix is free. Use a wired mic or the laptop's built-in one for input and keep the headset for listening only, so it stays in the music profile.

What loudness should a podcast hit, and where does -16 LUFS come from?

LUFS measures loudness the way ears hear it, averaged over the whole episode. Gain, by contrast, is set by momentary peaks. As of November 2026, Apple Podcasts recommends about -16 LKFS (the same scale as LUFS), plus or minus 1 dB, with true peaks no higher than -1 dBFS. The Audio Engineering Society's TD1004, published in 2015 for streaming, recommends roughly -16 to -20 LUFS with true peak below -1 dBTP, as of November 2026. That's why most podcasters settle at -16, the loud end of the range.

It's also why the test comes first. Loudness is set in editing, and it does nothing for a bad floor. Normalize a recording with a -50 dBFS floor up to -16 LUFS and the hiss rises right along with your voice. Measure with Audacity's Loudness Normalization effect or the free Youlean Loudness Meter, then export at -16 LUFS integrated and -1 dBTP.