Sample Rate and Bit Depth Explained

Headphones and DACs advertise "24-bit/192kHz" the way cameras advertise megapixels: a bigger number looks better, and the box rarely says what either figure actually gets you. Sample rate and bit depth aren't one big "quality" dial for audio. They answer two separate, narrower questions: which pitches a digital recording can hold, and how quiet its background hiss can be. Neither one is asking "how good does this sound?"

What sampling captures (hint: not a staircase)

A converter turning sound into digital measures the incoming signal at fixed intervals (the sample rate) and rounds each measurement to the nearest value it can store (the bit depth). What a sample rate can capture is set by the Nyquist–Shannon sampling theorem, laid out formally in engineering course material such as an introductory Signals and Systems chapter on the LibreTexts engineering library. In plain terms: a signal with nothing above a certain pitch is completely captured by sampling at twice that pitch, and it's rebuilt exactly by smooth curves around each sample, not by drawing straight lines between them.

The picture most people have of what happens between samples, a jagged staircase or dots joined by straight lines that only smooth out as the sample rate goes up, is a myth, not how the maths works. Xiph.org engineer Monty Montgomery's demonstration of digital audio fundamentals, hosted at people.xiph.org, takes on that picture directly: "all signals with content entirely below the Nyquist frequency...are captured perfectly and completely by sampling," and "the analog signal can be reconstructed losslessly, smoothly, and with the exact timing of the original analog signal." A properly working converter doesn't join the dots. It produces the one smooth waveform that fits every sample, which is the original. A cheap or badly designed converter can produce something closer to a staircase, but that's a flaw in that converter, not in sampling itself.

Why 44.1 kHz, of all numbers?

44.1 kHz didn't come from the science of hearing. It's a leftover from how digital audio was stored before hard drives could do the job. In the late 1970s, the only affordable way to get digital audio from a studio to a CD factory was to disguise it as a video signal and record it onto a video tape. A sample-rate history explainer published by forasoft.com walks through the maths: European PAL video gives 294 usable lines at 50 fields per second, and packing 3 audio samples into each line gives 294 × 50 × 3 = 44,100 samples per second exactly, while the American NTSC format lands within about 50 samples per second of the same number. The PAL figure won because it was the round one, and Sony's PCM-1600 recorder, released in 1979, brought it into professional use. A separate account on Wikipedia's entry for 44,100 Hz confirms the PCM-1600 date and the 1980 Red Book standard that then put 44.1 kHz on every commercial CD. None of that is about sound quality. 44.1 kHz is still with us because the CD locked it in.

The professional side of the industry never settled on that number out of habit. The AES3 digital-audio-interface standard, sold by the Audio Engineering Society and read here in a copy hosted by McGill University's engineering department, expects audio at one of the rates listed in AES5, the Society's separate recommendation on preferred sample rates. It has exactly three choices built in (32 kHz, 44.1 kHz and 48 kHz, defaulting to 48 kHz if none is given), and it limits each sample to 24 bits, in ranges of up to 20 or up to 24 bits. That 24-bit ceiling is where the next section comes in.

What bit depth really controls

Bit depth isn't a second quality dial on top of sample rate. Sample rate decides which pitches a recording can hold; bit depth decides how quiet the digital silence under the sound can be. That's its dynamic range: the gap between the loudest sound the format allows and the faint hiss left over from rounding every measurement. Each extra bit doubles the number of levels available, and each doubling adds about 6 dB (20 × log₁₀ 2). MIT's digital-audio course notes give the same rule as ΔL ≈ (6 × N) dB. So 16-bit audio has about 96 dB of range above that hiss in theory, and 24-bit has about 144 dB.

That hiss isn't automatically distortion, and whether it stays harmless depends on dither: a tiny bit of noise added before rounding, which the same MIT course notes describe as swapping distortion that follows the music for a steady background hiss that doesn't. Montgomery's Xiph.org piece describes it the same way: "proper dither...renders quantization noise independent of the input signal," so "dithered quantization introduces no distortion, just uncorrelated noise." A properly dithered 16-bit recording isn't a rougher version of the 24-bit one. It's the identical waveform with about 48 dB more hiss underneath, far below the background noise of almost any real playback system. Bit depth doesn't decide how finely the shape of a wave is traced over time (that's sample rate's job), and the two numbers don't trade off against each other. That dither hiss is even across all pitches, the same sense in which our noise generator's white noise is "flat". White, pink and brown noise: what the colours mean explains what "flat" means and why it sounds the way it does.

Where bigger numbers stop mattering, and where they still do

For a finished recording you're listening to, both numbers top out well below what gets marketed. Montgomery's Xiph.org analysis concludes that "16 bits is enough to store all we can hear, and will be enough forever," and an independent write-up on the science of hearing at tonestack.net gets to the same answer by measurement, putting properly dithered 16-bit/44.1kHz audio's noise-free range at about 118 dB above the quietest sound we can hear. That's already more than a real room or real speakers can use. Sample rate works the same way above 48 kHz: what we can hear stops around 20 kHz, and any rate more than double that already covers it with room to spare.

One part of this isn't as settled as the rest. Montgomery goes further and argues that 192 kHz playback isn't just pointless but slightly harmful, because, in his account, "inaudible ultrasonics contribute to intermodulation distortion in the audible range" once they reach an amplifier or speaker never designed to handle sounds that high. That's one expert's argument, not an accepted fact. Physicist David Griesinger, who presented on this same question at an AES convention in Banff, writes on his own page that what he first suspected was audible ultrasonic distortion in his own setup turned out to be ordinary distortion in one speaker driver. Another system might point to a different cause, or none. Whether it happens on any particular setup is an open question that depends on the hardware, so we're reporting it as Montgomery's view, not our conclusion.

The bigger numbers aren't useless, though. They do a different job, before the music reaches you. Recording and mixing usually run at 24 bits because a project piles up tiny rounding errors across thousands of effects, volume changes and combined tracks, and the extra bits give room to absorb them. You'd never hear the difference directly. The same analysis notes that "an engineer...requires more than 16 bits during mixing and mastering" for exactly that reason, and the extra bits are dropped once the mix is finished and released. Sample rates above 48 kHz help in a similar way during production: the extra space between 20 kHz and the upper limit lets the recording's filters roll off gently instead of steeply. That's a real benefit while recording, and it disappears once the finished file is converted down for release.

A browser can't check any of this for your own equipment, and we don't claim it can. The sample rate and bit depth your system actually uses depend on your operating system's audio driver and your audio interface, and, for anything recorded through a microphone, on the recording side covered in why isn't my microphone working. If you came here wondering what a codec does to these numbers on the way to your headphones, Bluetooth audio latency by codec covers the compression on top of them. And if your real question is how cleanly your equipment plays a waveform back, rather than how it's stored, total harmonic distortion covers the measurement made for that.