Sample rate vs bit depth

At a glance

What's the difference between sample rate and bit depth?

Digital audio stores a sound as a long list of measurements of its waveform, one after another. The sample rate is how many of those measurements are taken each second, and the bit depth is how many binary digits each one is written down with. So one number is about time and the other is about level. On a graph of the wave, sample rate spaces the dots out along the horizontal axis and bit depth sets the grid they snap to on the vertical one.

Each setting has one job. The sample rate decides the highest frequency a file can hold and nothing else about the sound, while the bit depth decides how far below the loudest possible level the noise sits, which is the file's dynamic range. Neither one changes the loudest level a 16-bit or 24-bit file can hold, which is 0 dBFS either way.

The lab at the top of the page is a converter you can set too low on purpose. The first three steps play the same drum loop at CD quality, then with only the sample rate lowered, then with only the bit depth lowered. Watch which part of the spectrum moves each time.

The full background on each one, from the sampling theorem to 32-bit float, is in the module lesson. This page stays on how they differ and which one to worry about.

What does sample rate change?

It sets where the frequency range stops, and touches nothing below that. A file can hold frequencies up to half its sample rate, called the Nyquist frequency, so 44.1 kHz reaches 22.05 kHz and 48 kHz reaches 24 kHz. Everything below that line comes back out exactly as it went in, and everything above it has to be filtered off before sampling. If it isn't, it folds back down into the audible range as aliasing.

That's why a low sample rate sounds the way it does. At 11.025 kHz, nothing above 5.51 kHz survives, so cymbals, breath and the snap of a snare go missing while the kick and bass are untouched. Switch on Hear only what's lost in the lab and the part that goes is a thin, bright layer with no low end in it at all.

A higher sample rate doesn't make a 1 kHz tone smoother, because the converter's output filter draws the one smooth wave through the samples, and at any rate above 2 kHz that's the original tone. It doesn't place a drum hit more precisely in time, either. A properly filtered signal can put an event between two samples, so its timing isn't rounded to the nearest sample. Christopher Montgomery's video Digital Show & Tell, made for Xiph.Org, demonstrates both with real converters and an oscilloscope.

The last lab step makes the same point the other way round. A 110 Hz sine has nothing above 110 Hz, so lowering the sample rate to 8 kHz removes nothing from it, and the rate would have to fall below 220 Hz before the sine itself was in danger.

What does bit depth change?

It moves the noise floor up or down. Every sample has to be rounded to the nearest level the bit depth allows, and the rounding error is heard as a faint hiss under the music. Each extra bit doubles the number of levels and halves the size of the steps, which lowers that hiss by about 6 dB, so 16 bits gives about 96 dB of range and 24 bits about 144 dB. No interface gets near 144 dB in practice, because its analogue circuits add more noise than that, but a good one still measures well past the 96 dB of 16-bit.

Loud sounds hide the hiss. Quiet ones don't, so bit depth problems show up first in reverb tails, fade-outs, soft notes and anything recorded too low. In the lab at 6 bits the noise floor sits about 60 dB higher than at 16, and the grit is easiest to hear on the end of each hit as it fades. Hear only what's lost plays that error on its own, and unlike the sample rate's loss it covers every frequency, from the lows to the very top.

Bit depth decides how much headroom you can leave while recording. Every 6 dB your peaks sit below 0 dBFS leaves one bit unused at the top, and the noise floor stays where it was. On a 16-bit recorder, a vocal recorded 36 dB too low uses only about 10 bits, and the hiss comes up with it when you turn it up in the mix. The same take at 24 bits leaves the noise 48 dB lower, under anything you'd hear. Lowering or raising the sample rate on that take would change nothing about its noise.

With very few bits the rounding error stops sounding like hiss. On a steady tone it repeats with every cycle, so it turns into a buzz in tune with the note, which is what a bitcrusher does on purpose. Adding a little random noise before the rounding, called dither, turns that buzz back into an even hiss.

Which matters more, sample rate or bit depth?

For recording and mixing, bit depth, by a long way. The difference between 16 and 24 bits is 48 dB of room under the music, which decides whether a take recorded too low can be saved. The difference between 48 and 96 kHz is a band from 24 to 48 kHz that you can't hear, and that most microphones barely pick up.

A higher sample rate is worth it in a few specific cases. Music for video should be at 48 kHz because that's the standard for picture. Sound designers who plan to pitch recordings down an octave or more record at 96 or 192 kHz so there's content left to bring down into the audible range. Plug-ins that distort can alias less at higher rates, though turning on their own oversampling does the same job for less CPU, which the aliasing and oversampling lesson covers.

For a finished master, neither makes much difference. A 16-bit, 44.1 kHz file has more dynamic range than any mastered track uses and every frequency people can hear. When Spotify added lossless streaming to Premium in September 2025, it went up to 24-bit but kept 44.1 kHz.

How do sample rate and bit depth affect file size?

Uncompressed audio takes up the sample rate times the bit depth times the number of channels, every second. That figure is the bit rate. A CD is 44,100 × 16 × 2, or 1,411 kilobits a second, and dividing by eight gives bytes: about 10.6 MB for every minute of stereo.

Going from 16 to 24 bits makes every file half as big again. Doubling the sample rate doubles the size, and it doubles the processing as well, because every plug-in in the session has twice as many samples to work through each second.

Bit depth has no effect on processing. Your DAW turns every file into 32-bit or 64-bit floating point as it plays, so a 16-bit file and a 24-bit file cost the same CPU once they're in the session. Latency works the same way. It's the buffer size divided by the sample rate, so a 256-sample buffer is 5.8 ms at 44.1 kHz and 2.7 ms at 96 kHz, whatever the bit depth.

What happens when sample rate or bit depth don't match?

A sample rate mismatch changes speed and pitch. A file is only a list of numbers, so if a 44.1 kHz file is played back 48,000 numbers a second, it runs 8.8% fast and about 1.5 semitones sharp. Most DAWs convert mismatched files when you import them or as they play, but some leave it as a setting, and a sample that's slightly out of tune with everything else is the usual sign.

A bit depth mismatch does nothing you can hear. A 16-bit file in a 24-bit session plays exactly as it is; its bottom 8 bits are empty.

Converting each one down is a different job too. Lowering the sample rate, say from 48 to 44.1 kHz for a CD, needs a sample-rate converter, which low-passes everything above the new Nyquist frequency and works out new samples in between the old ones. Lowering the bit depth, say from 24 to 16 bits, needs dither, added once as the very last step. Going up in either direction adds nothing: an upsampled file has nothing above its old Nyquist frequency, and a 16-bit file saved as 24-bit has eight empty bits at the bottom.

How to choose sample rate and bit depth

1. Set the bit depth to 24Use it for every recording. It's the default in most DAWs, and it lets you leave plenty of headroom without the hiss coming up.
2. Pick 44.1 or 48 kHzUse 48 kHz if the music is going anywhere near video, and either one otherwise. Go higher only for one of the reasons above.
3. Keep the session at that rateChoose before you record. Changing it later means converting every file you've already got.
4. Bounce stems at the session rateUse 24-bit, or 32-bit float if a bounce might run hot, since a float file can't clip.
5. Convert only for the destinationLower the sample rate only when a format needs it, such as 44.1 kHz for a CD from a 48 kHz session, and dither only when the bit depth goes down.

Common mistakes

Raising the sample rate to fix hiss

Noise in a recording comes from the bit depth, the gear and the room, so a 96 kHz session records the same hiss as a 48 kHz one. Switch to 24-bit and record at a sensible level; if the hiss is still there, it's coming from the gear or the room.

Blaming the bit depth for a dull sound

A dull, muffled sample has lost its high frequencies, and bit depth can't do that. Look for a low sample rate somewhere in its history, or a low-pass filter, or a lossy file like a low bit rate MP3.

Treating hi-res numbers as a quality check

A 24/96 file made from a 16/44.1 source has the same content as the source, at more than three times the size. The numbers describe the file's container, and a sample pack or download is only as good as whatever was recorded into it.

Mixing up bit depth and bit rate

Bit depth is per sample. Bit rate is per second, and for an uncompressed file it's the sample rate and bit depth multiplied together with the channels. MP3 and AAC files are described by bit rate alone, such as 320 kbps, because they don't store a list of samples in the same way.

Training your ears to tell them apart

Each round of the quiz under the lab plays one converted version of the loop against the original, with the screens covered and the levels matched, and you say whether the conversion lowered the sample rate or the bit depth. Listen to the top of the hats for a lower sample rate, which makes them duller, and to the quiet ends of the hits for a lower bit depth, which adds a grainy hiss under them.

Questions people ask

Keep going