๐ŸŽšAudio test

โœ“ Tested & verified Updated: How the test works

Audio test

  1. 1This pagebuilds the tone in your browser โ€” nothing is downloaded, nothing is uploaded
  2. 2Web Audio enginesample rate, channel count and latency are read here
  3. 3Your output devicepast this point the page is blind, so the last check is your own ears

Output sample rate

โ€”Hz

Press start and the browser reports what it is really sending to your sound card: the rate, how many channels it was given, and how long a sound waits in the buffer. A 440 Hz reference tone plays for three seconds so your ears can confirm the last step nothing can measure.

  • Channels acceptedโ€”
  • Output latencyโ€”
  • Highest tone it can carryโ€”
  • Reference tonesilent

One gear check a month

Monthly

One thing a month worth measuring before it fails properly โ€” a mic that reset itself, a stick that drifts, a switch sending two clicks. Two sentences and the test.

Free, no account. One click unsubscribes and deletes your address.

Try also

Is my audio working, and what is it running at?

The audio test reports two numbers instead of a yes: the sample rate your output is running at, in Hz, and the number of channels it will accept. Anything plugged in makes a noise, so silence is the rarest fault here.

The rate tells you which format the machine settled on. 44 100 Hz came from the CD; 48 000 Hz is the default nearly everywhere else and the rate all video uses. A rate down at 16 000 Hz means something has opened your microphone and pulled a wireless headset into its mono voice profile. The channel count tells you who is listening at the far end: two for an ordinary stereo pair, six for 5.1, eight for 7.1, one for that same hands-free profile.

Read together they explain the faults a green tick cannot: music gone suddenly thin, a surround receiver playing flat stereo, a headset that sounds like a phone call. Everything is generated in the page, and nothing is uploaded.

Both numbers are read from the output your browser is using right now, so they change when you change device โ€” plug in a headset and run the test again rather than trusting what was true a minute ago.

Reported rateWhere it comes fromWhat it tells you
8 000โ€“16 000 HzBluetooth hands-free profileA microphone is open somewhere; stereo is gone until it closes
32 000 HzOlder voice hardware and some virtual devicesUsable, but the top of the range is already cut
44 100 HzThe CD rate, still common on music-first setupsNormal; the ceiling is 22 050 Hz
48 000 HzThe default on most systems and all videoNormal; the ceiling is 24 000 Hz
96 000โ€“192 000 HzA USB interface or DAC set highNormal for that hardware, and inaudibly different from 48 000 Hz
Channels acceptedWhat it means
1Mono: a hands-free profile, a phone-style device, or a mono accessibility setting
2Stereo, including every virtual surround mode
6A 5.1 chain the browser can address
8A 7.1 chain the browser can address

This page is about the format leaving your machine, not about which speaker is silent. When one side plays and the other does not, that is the speaker test, and on a headset the headphone test adds the polarity check. When the count comes back at two on a system that should manage six, surround sound test works through the reasons. To hold one frequency at a level you choose rather than a fixed test tone, use the tone generator; to hear where the output rolls off at each end, run the frequency test. Input is the opposite direction entirely and belongs to the mic test.

Frequently asked questions

There is no sound at all on this computer. Where do I start?

Work outward from the application, because the chain has four places to fail and only one of them is hardware. Check the app is not muted in the volume mixer, then that the system output points at the device you are listening to rather than an idle HDMI monitor, then that the device is not disabled in the sound settings, and last that no program has taken exclusive control of it. If a tone plays here and nothing plays elsewhere, the fault is above the driver and the hardware is fine.

Is a YouTube test video as good as this?

It proves the same one thing and hides everything else. A video confirms sound arrives, but it travels through decoding, a player and often a downmix, so it cannot tell you the rate your output runs at or how many channels it accepts, and a slow connection can make a working system sound broken. This page generates the tone locally, so nothing between the browser and your speakers is being blamed for the network.

What is the Nyquist frequency, and does the sample rate limit what I can hear?

A digital system can only carry frequencies below half its sample rate, and that half is the Nyquist frequency. At 44 100 Hz the ceiling is 22 050 Hz and at 48 000 Hz it is 24 000 Hz, both comfortably above the top of human hearing, which is why higher rates make no audible difference to playback on their own. It matters when the rate collapses: a 16 000 Hz stream has an 8 000 Hz ceiling, and that is why voice profiles sound muffled.

Why does my headset show one channel and a low rate?

Because a microphone is open somewhere. The classic Bluetooth profiles cannot carry high-quality stereo and a microphone at the same time, so the moment any app claims the headset’s input, the link switches from the stereo profile to the mono hands-free one and the sound quality drops in the same instant. Close whatever holds the microphone, or select a separate input device, and the link returns to stereo without touching a single setting.

Does this measure how loud my speakers are?

No, and nothing in a browser can. Loudness at your ears depends on the amplifier, the drivers, the room and the volume knob, none of which the page can see; measuring it needs a calibrated microphone. What a page can state exactly is the digital level it sends, in dBFS, which is the half of the chain it controls. The other half is why the same setting is comfortable on one machine and painful on another.

How the test works

The rate comes from the audio context the browser opens for output, which follows the rate the operating system is running that device at rather than anything the page chooses; that is why it changes underneath you when a Bluetooth link switches profile. The channel count is the output’s own maximum, reported by the destination node before any sound is routed, and it describes what the device will accept rather than how many speakers are physically connected โ€” an eight-channel receiver with two speakers plugged in still reports eight. The tone itself is a sine wave generated in the page at a level well below full scale, faded in and out over a few milliseconds so that no click is louder than the tone. Nothing is downloaded, no file is decoded, and no request leaves while the test runs. What the page cannot do is hear the result: it reports what it sends and what the device accepts, and your ears confirm the rest.

Worked example

With sample rate the output reports 48 000 Hz, channels the device offers 2, this page works out a ceiling of 24 kHz. Your output is running at 48000 Hz, so the highest tone it can carry is 24 kHz โ€” anything above that is thrown away before it reaches the speaker. The device offers 2 channels, so this output is stereo. This is the rate the browserโ€™s audio engine runs at, not proof of what the hardware does: Windows and macOS can resample between the two without telling the page.

  • Output sample rate48 000 Hz
  • Highest tone it can carry24 kHz
  • Highest step this site will test23 kHz

Sources:

Why the rate and the channel count are worth knowing

The rate is a fault indicator, not a quality setting. Raising it does not improve playback, but watching it fall from 48 000 Hz to 16 000 Hz identifies a Bluetooth profile switch in one glance.

The channel count is what the device accepted, not what it promised. A receiver advertised as 7.1 that accepts two channels is telling you the connection or the system configuration is the limit, well before any speaker is at fault.

Neither number needs a permission prompt. Output tests read what the browser already knows, so there is no dialog to approve and nothing for a blocked permission to break.