Sample rate: how many snapshots per second

The sample rate is how often the analog signal is measured. The Nyquist theorem says you can reconstruct frequencies up to half the rate, so the two common rates play out like this:

• 44,100 Hz (44.1 kHz) — the CD standard; captures up to ~22 kHz.
• 48,000 Hz (48 kHz) — the video/broadcast standard; captures up to 24 kHz and is what most USB mics and browsers natively run.

Human hearing tops out around 20 kHz (optimistically), and speech lives almost entirely under 8 kHz. Both rates cover that with huge margin, which is why the tester reports whichever the browser negotiated and you don’t need to change it. The only sample-rate advice that matters: keep the whole chain on one rate. A 48 kHz mic feeding a 44.1 kHz recording project gets resampled — resampling is fine when it happens once, sloppy when it happens three times.

Bit depth: how precise each snapshot is

If sample rate is the frame rate, bit depth is the resolution of each frame — how many voltage levels the converter can distinguish. It sets the noise floor:

• 16-bit — ~96 dB of dynamic range. The CD standard; entirely adequate for speech.
• 24-bit — ~144 dB theoretical, ~120 dB in real converters. Its practical gift is headroom: you can record quiet (peaks near −18 dB) and still be far above the noise floor, then boost later.
• 32-bit float — what the browser pipeline delivers internally, and what some new recorders write. Float audio effectively can’t clip at the converter stage, though the analog side (mic capsule, preamp) still can.

The tester shows “32-bit float” because that’s the format Web Audio hands the page — your hardware’s actual ADC depth isn’t exposed to browsers and honestly doesn’t matter for what this tool checks.

Channels: one mouth, one channel

Mono (1 channel) is the natural format for a single talker — built-in laptop mics, headset boom mics, and most USB condensers are mono. A device reporting 2 channels is either a genuine stereo pair (rare for speech work) or a mono capsule wrapped in a stereo container (common — the L/R meters in the tester will move identically). For calls, streaming and podcasts, mono is what you want; it keeps the voice dead-center for every listener and halves the file.

What actually matters more than the numbers

For real-world intelligibility, three unglamorous factors beat any spec on the box:

1. Distance. Halving the distance to the mic roughly doubles the level — far more than any ADC upgrade buys. A hand’s width is the baseline.
2. Room noise. Fans, echo and keyboard clatter fight the noise-suppression algorithm; the meters show it as a floor that never rests.
3. Processing flags. Echo cancellation, noise suppression and auto-gain — the flags visible in the tester’s info grid — shape call audio more than bit depth ever will. They’re right for meetings and wrong for music; knowing they’re on explains most “why do I sound underwater” mysteries.

If you’re shopping on specs, USB vs headset vs built-in covers what the hardware tiers actually change. And when a working mic still sounds bad, start with the mic not working checklist — most “quality” problems are the mute/gain layer, not the capsule.

Frequently asked questions

Is 48000 Hz better than 44100 Hz for a microphone?

Neither is 'better' for speech — both capture the full audible range with margin (Nyquist caps them at 24 and 22.05 kHz). 48 kHz is the video/broadcast standard and what most USB mics and browsers negotiate natively; 44.1 kHz is the CD standard. The audible difference between them is zero; pick whatever your chain uses natively and avoid resampling.

Do I need 24-bit recording for voice?

For calls and meetings: no — the far end hears a compressed codec anyway. For recording you might edit: 24-bit is cheap insurance, because its extra dynamic range lets you record conservatively (peaks around −18 dB) without losing quality, then normalize later. 16-bit has ~96 dB of dynamic range — still more than a quiet room needs — but leaves less headroom for mistakes.

Why does the tester say '32-bit float' when my mic is 16-bit?

Because two different layers are being measured. Your hardware's ADC might genuinely be 16 or 24-bit, but the moment audio enters the browser's Web Audio pipeline it's carried as 32-bit float samples — that's the only precision the page can observe. The float format preserves everything the ADC captured; it doesn't claim the ADC is better than it is.

Should I use a stereo microphone for calls or podcasts?

Almost never. Speech is a point source — your mouth — so mono is correct and keeps voice centered for every listener. Stereo mics exist for ambience, ASMR and music, and a stereo pair pointed at one mouth mostly adds phase problems. If your device reports 2 channels, it's usually just the interface format, not two meaningful capsules.

What is a 'channel count' of 1 vs 2 actually telling me?

How many discrete audio signals the device delivers. One channel = mono (built-in laptop mics, most USB mics, every headset mic). Two channels could be a real stereo pair or — common with USB interfaces — a stereo wrapper around one capsule. The tester's L/R meters tell you which: identical meters mean one source wrapped in two channels.

Why does my expensive mic still sound bad on calls?

Because the bottleneck moved. Call apps compress to ~16–32 kbps Opus and stack echo cancellation, noise suppression and auto-gain on top — the negotiated flags you can see in the tester's info grid. A $300 condenser and a $30 headset land within a hair of each other after that pipeline. Spend effort on distance and room noise instead; that's where call quality is actually won.