Dynamic Range & Dithering Test
Hear what bit depth actually does to audio. A pulsing tone steps from full scale down to −90 dBFS, requantized live — plain, dithered, or noise-shaped. Mark the quietest step you can still hear in each: dither keeps the tone alive long after plain quantization dies.
Original skips quantization entirely. The other three round every sample to the selected bit depth — without dither, with dither, or with high-frequency shaped dither.
8-bit makes the effect unmistakable; 16-bit is the CD standard.
Steps the tone down 6 dB every few seconds, from 0 to −90 dBFS. Press the mark button for as long as you can still hear the pulses.
Records the current level for the current mode — your deepest mark per mode appears in the results below.
You will need to raise the volume to follow the ladder into the quiet steps. Turn it back down before playing anything else.
Run the ladder in each mode and mark the quietest level you can hear — the results will appear here.
Use headphones in a quiet room — laptop speakers and background noise swallow the quiet half of the ladder.
The deepest steps sit near the noise floor of a normal room. Closed headphones somewhere quiet buy you 20–30 dB.
Raising the volume as the ladder descends is part of the test — turn it back down before playing anything else.
The steady rhythm is what you pick out of the dither noise. Still feel the rhythm? The signal is still there.
Digital audio stores each sample as a whole number, and the bit depth sets how many values there are: 8 bits give 256 steps, 16 bits give 65,536. Rounding to the nearest step — quantization — hurts most where the signal is quietest, since a near-silent sound has only a few steps left to describe it: it turns grainy, then vanishes below the lowest step. Full scale down to that floor is the dynamic range, roughly 6 dB per bit — 48 dB at 8 bits, 96 dB at 16. Dithering is the counterintuitive fix: add a trace of noise before rounding, and a tone smaller than one step still shifts the odds of each sample rounding up or down, so your ear recovers it from the average. Noise shaping moves that noise into the high frequencies, where hearing is least sensitive.
How does the test work?
Everything is synthesized live with the Web Audio API — nothing is downloaded or uploaded. The widget plays a pulsing sine at the level you set and requantizes it sample by sample in four modes: Original leaves it untouched, Quantized rounds with no dither, Dithered adds noise before rounding, and Noise-shaped pushes that noise up the spectrum. The ladder button drops the level 6 dB at a time from 0 to −90 dBFS; mark every step where the pulses are still audible, then compare modes. The panel draws the processed waveform, normalized, so you can watch the sine collapse into a coarse staircase — and see the dither modes trade that staircase for a fuzz of noise.
What should I hear?
At 8 bits the differences are dramatic, which is why it is the classic demonstration depth. The plain quantized tone turns raspy below about −36 dBFS and cuts to silence past −48 dBFS, the theoretical 8-bit floor. Dithered, a steady hiss appears but the pulses stay audible to roughly −54 to −60 dBFS: the noise is not masking the tone, it is carrying it, toggling the lowest bit in step with a signal smaller than the step itself. Noise shaping thins that hiss and usually buys another 6 dB. The unquantized original is the control, audible all the way to −90 dBFS in a quiet room. This is why real recordings are dithered on the way down to 16 bits: the noise sits near −93 dBFS, too quiet to notice, while reverb tails and fades stay smooth instead of crumbling.
Bit depth and dynamic range
| Bit depth | Levels | Theoretical range | Where you meet it |
|---|---|---|---|
| 4-bit | 16 | ~24 dB | Demos like this one — the grit is audible even at full scale |
| 8-bit | 256 | ~48 dB | Early samplers, telephone-era audio, dithering demos |
| 12-bit | 4,096 | ~72 dB | Classic drum machines and vintage samplers, prized for the crunch |
| 16-bit | 65,536 | ~96 dB | CDs, streaming, downloads — the delivery standard |
| 24-bit | 16.7 million | ~144 dB | Studio recording and mixing, where editing headroom matters |
Does my listening gear change the result?
Yes — this test measures your whole chain, not just the math. A noisy room, leaky earbuds, or a device that cannot play quietly but cleanly will bury the deepest steps in every mode, so use closed headphones somewhere quiet if you want to reach the bottom. Your ears matter less than you might expect: at 440 Hz nearly everyone is sensitive, so a shallow result points at the room or the gear. And drop the volume back down before playing anything else — a level that made a −84 dBFS pulse audible turns a normal track into a blast.