LESSON 18 · SOURCES & DACS
What does 16-bit or 24-bit audio actually change?
Imagine taking little snapshots of a moving sound wave. Each snapshot needs a number to describe how high or low the wave is at that moment. Bit depth tells us how many different numbers we can choose from. More choices mean less rounding, not extra instruments, new notes or an automatically better song.
- Sample
- One snapshot of the sound wave’s level.
- Bit depth
- How precise the number stored for each snapshot can be.
- Sample rate
- How many snapshots are taken each second. That's a different setting.
Those values describe height, not how often snapshots are taken. First, experiment with a much smaller set of values so you can actually see why it matters.
START WITH ONE IDEA
More bits give the recorder more level choices.
Few available levels
A sound's level may fall between the choices, so the recorder must round.Many available levels
With more choices, the saved number can sit closer to the original level.Picture a measuring ruler. A ruler marked only at whole inches forces you to round. A ruler with smaller marks lets you record a more precise length. Bit depth works similarly, but for the height of an audio signal.
With 16 bits, a digital audio sample can use 65,536 different values. With 24 bits, it can use 16,777,216. That's 256 times as many possible values, giving 256 times finer level spacing on a comparable scale.
It doesn't mean 24-bit music has 256 times more detail you can hear. The original microphone, recording noise, mastering choices and your playback system still matter.
What do “PCM,” “signed” and “quantization” mean? Optional vocabulary
PCM is the name for saving a sound wave as a long list of numbered readings. Linear means the ruler has evenly spaced marks; signed means each reading can sit above or below the wave's centre line. Quantization is the unavoidable rounding to the nearest mark. These are recording terms, not four separate sound-enhancing features.
In signed 16-bit PCM, the stored codes run from −32,768 to +32,767. Their binary representation uses sixteen 0-or-1 positions. This is how the 65,536 possible patterns arise, not 65,536 musical notes.
VISUAL EXERCISE 1 · ROUNDING
Can you see what the recorder has to round?
Keep the same wave and the same 25 snapshots. Give the recorder more or fewer level choices. The gold line is the original mathematical signal. Blue dots are the numbers that get stored. Small connecting lines show the rounding gap.
Eight level choices create visible rounding gaps. Try 32 and 256 levels to see the blue dots move toward the gold curve.
This is a teaching close-up, not actual CD audio. The intentionally tiny 3-bit and 5-bit examples make the rounding visible. A real 16-bit or 24-bit sample has vastly more levels. The blue dots represent saved numbers, not the jagged analogue output of a DAC.
VISUAL EXERCISE 2 · REALLY QUIET AUDIO
What happens when the sound gets very quiet?
The earlier exercise used exaggerated settings. Now use real 8-bit, 16-bit and 24-bit formats. Turn the same tiny note down until it sits close to the available steps near silence. Notice which versions can still follow its shape.
This shows how far the quiet note's loudest point is below the digital maximum, called 0 dBFS. Move the control to see what happens when a sound is extremely quiet and the recording has fewer steps available.
At this extremely low level, 16-bit has only about one step from silence to the note's peak in this simplified no-dither model.
Important: Both wave shapes have been magnified by the same amount so that a nearly silent signal can be seen. This deliberately does not include dither, which helps real 16-bit recordings preserve very quiet information as low-level noise rather than simple rounding artefacts. Neither chart is a listening test or a DAC measurement.
THE BIG MISUNDERSTANDING
More available levels don't make the band more dramatic.
Bit depth describes the recording format's capacity. Musical dynamics describe how much the song changes between quiet and loud. A drummer leaving space before a heavy crash creates musical impact. A file marked “24-bit” does not create that impact if the recording or mastering already flattened it.
The same song at normal audio-editor zoom
Think of the waveform display in Audacity or a recording program. These are three digital versions of the exact same invented performance and the same master. Only the number of available stored levels changes.
8-bit
Coarser level rounding, especially in very quiet sound.
16-bit
Much finer level choices; CD-quality files can preserve this musical contrast.
24-bit
More numerical precision, especially useful while recording and mixing.
Notice something? At this zoom, the three waveforms look almost identical. Bit depth doesn't automatically make the chorus bigger, the drums hit harder or the band play more dramatically. The earlier magnified examples on this page show the much smaller rounding differences.
Original synthetic, music-like single-channel waveform. Every row uses the same time and height scales; the 8-, 16- and 24-bit examples round the same samples with no dither for illustration. An 8-bit file can still sound noticeably noisier or more distorted, particularly at low levels, even though its overall shape looks similar. This figure is not a recording or a listening test.
Compare this with mastering. A more dynamic master can sound quieter at the same stereo volume setting, yet have much bigger musical peaks when turned up for a fair comparison. A heavily limited master may look denser. Those differences come from processing the music, not whether the file says 16-bit or 24-bit. See three familiar, full-song waveforms in the mastering lesson.
ONE USEFUL PRODUCTION TRICK
Why would an engineer add a tiny amount of noise?
At very quiet levels, repeatedly rounding the same sound can produce an unwanted pattern. Engineers call this quantization error. The surprising solution is a tiny amount of random noise, called dither, added before reducing a finished recording to fewer bits.
Without suitable dither
Very quiet sound → repeated rounding → the error can follow the sound and become rough or tone-like.
With suitable dither
Very quiet sound → tiny random noise → the error behaves more like gentle background noise. Delicate fades can be represented more faithfully.
You do not need to add dither every time you listen. It is mainly a final-production step when changing to a lower integer bit depth, such as preparing a 16-bit CD master. It doesn't create extra bits or rescue a clipped recording.
Where do the familiar 96 dB and 144 dB figures come from? Optional maths
An ideal digital system gains roughly 6 dB of possible range for each extra bit. So 16 bits is roughly 96 dB and 24 bits roughly 144 dB by that shortcut. A more specific textbook calculation for a full-scale sine wave with ideal uniform rounding error and no dither gives 6.02 × bits + 1.76 dB, or about 98.1 and 146.2 dB.
Those are not measurements of a real DAC or how dynamic a song is. Analogue noise, recording conditions, dither and measurement bandwidth all matter. Read how DAC measurements are reported.
Can you actually hear 16-bit versus 24-bit music?
Often, the bigger audible difference comes from the recording and mastering, not the number on the file. A well-made, properly dithered 16-bit version can already reproduce music extremely well under typical listening conditions.
A 24-bit recording is valuable while you work: it gives you far more space for low-level information when leaving extra headroom for unexpected peaks. But a released 24-bit file does not automatically have wider soundstage, clearer instruments or more powerful drums than its 16-bit counterpart.
COMMON CLAIM
“24-bit always means more detail, deeper soundstage and bigger dynamics.”
WHAT THE SPEC ACTUALLY SUPPORTS
More possible numerical levels for each snapshot. The mix, mastering, source noise and playback chain determine whether that extra precision makes an audible difference.
Which bit depth should you choose?
- Listening to music? A well-mastered 16-bit recording is already excellent. Choose the master you prefer rather than automatically paying more for a 24-bit badge.
- Recording a voice, instrument or band? Start with 24-bit recording, typically at 44.1 or 48 kHz. You can leave room for loud unexpected notes without pushing the recording close to the top of the meter.
- Making a finished 16-bit file? Use suitable dither when reducing the integer bit depth, normally at the final export stage.
- Buying a DAC? Support for 24-bit files isn't the same as achieving 24 bits of noise-free analogue performance. Look at actual measurements and practical listening conditions.
Does 24-bit take more storage space? Optional detail
At the same sample rate and channel count, 24-bit uncompressed PCM uses 50% more audio data than 16-bit. For stereo 44.1 kHz audio, the raw rates are 1,411.2 kilobits/s and 2,116.8 kilobits/s. Formats such as FLAC compress losslessly, so their actual file sizes vary.
Sources and further reading
- Analog Devices · How bit depth relates to precision and possible dynamic range ↗
- iZotope · Why dither is used for final exports ↗
- iZotope · More technical dither and noise-shaping details ↗
About the visuals: These diagrams use original, invented signals and mathematically rounded sample values. They are not recordings, product measurements or listening-test results. The demonstrations intentionally leave dither out so rounding is easy to inspect.