What sample rate and bit depth actually mean

Every audio file carries two numbers, and almost every argument about "quality" mixes them up with each other and with the recording itself. They are simpler than the arguments suggest — and neither one says anything about whether a record sounds good.

Last updated 16 September 2026

Sample rate: how high the file can reach

Digital audio stores a measurement of the waveform taken many thousands of times a second. The sample rate is how many. CD uses 44,100 per second; a hi-res file often uses 96,000 or 192,000.

What that buys you is one specific thing: the highest frequency the file can represent is half the sample rate. 44.1kHz reaches 22.05kHz. 96kHz reaches 48kHz. That limit is not a design choice anyone can argue with — it falls out of the mathematics of sampling.

Human hearing tops out around 20kHz when young, and lower with age, which is why 44.1kHz was chosen in the first place: it covers the audible range with a little room for the filter. The practical argument for higher rates is not that you can hear 40kHz tones. It is that engineers get more room to place filters gently, well outside the range you can hear.

Bit depth: how far the file can reach down

Each of those measurements is stored as a number, and bit depth is how many bits that number gets. 16 bits for CD, 24 for most hi-res releases.

The common picture — more bits means a "smoother" waveform, fewer stair-steps — is wrong, and it is worth dropping, because it leads people to expect differences that are not there. What bit depth actually sets is dynamic range: the distance between the loudest sound the file can hold and the noise underneath everything. Roughly 6 dB per bit.

• 16-bit — about 96 dB of range. For reference, a quiet living room sits around 30 dB and a loud concert around 110 dB.
• 24-bit — about 144 dB in theory. In practice the analogue electronics in the recording chain run out long before that, so no real recording uses all of it.

That extra range earns its keep during recording and mixing, where you want headroom so that an unexpectedly loud moment does not clip. By the time a finished master reaches you, the practical benefit of 24 over 16 bits is much smaller than the file size suggests.

What neither number tells you

Neither one describes the microphones, the room, the performance, the mix or the mastering — and those decide how a record sounds. This is why a 1960s album transferred at 24-bit/192kHz can sound narrow and thin, while a modern 16-bit/44.1kHz master sounds enormous. The numbers describe the container's capacity. What was poured into it is a separate question.

What this costs in storage

Uncompressed stereo, per minute:

FLAC compresses this losslessly, typically to somewhere between 50% and 70% of the original depending on the music — quiet acoustic material compresses much further than dense loud material. Lossless means the decoded audio is bit-for-bit identical to what went in, which is the whole point: unlike MP3, nothing is decided to be inaudible and thrown away.

Why upsampling adds nothing

Taking a 44.1kHz master and resampling it to 96kHz produces a genuine 96kHz file. It will report 96kHz everywhere, and it is not lying. But there was never any content above 22.05kHz in the source, and resampling does not invent any — it describes the same audio with more numbers, and the file gets twice as large for no gain.

That is what "upscaled" means when a checker flags a file, and it is also why the flag is worth taking seriously even though nothing about the file is technically false.