Bit depth is the number of bits used to represent the amplitude of each sample in a digital audio signal. In digital audio, it determines how finely the converter can resolve level differences and therefore sets the quantization noise floor and the theoretical dynamic range of a recording, at roughly 6 dB per bit.
Bit depth works together with sample rate, which sets the frequency range, to define a PCM audio format such as 16-bit / 44.1 kHz for CD or 24-bit / 48 kHz for video production. More bits do not make a signal brighter or more detailed in the musical sense; they push the noise floor lower and leave more room between quiet passages and the digital ceiling.
How bit depth defines amplitude resolution
During analog-to-digital conversion, each sample is rounded to the nearest of a fixed set of values, a step called quantization. With N bits there are 2 to the power of N possible values, so 16 bits give 65,536 steps and 24 bits give 16,777,216 steps between the most negative and most positive full-scale level.
The rounding introduces a small error on every sample. For complex signals this error behaves like random noise spread across the spectrum, and its level relative to a full-scale sine wave is approximately 6.02 times N plus 1.76 dB. Each added bit therefore halves the step size and lowers the noise by about 6 dB.
The table lists the bit depths that appear most often in audio work.
| Bit depth | Theoretical dynamic range | Typical use |
|---|---|---|
| 8-bit | About 50 dB | Early computer audio, lo-fi effects |
| 16-bit | About 98 dB | CD audio, most consumer distribution |
| 24-bit | About 146 dB | Recording, mixing, professional video |
| 32-bit floating point | Above 1,500 dB internally | DAW processing and some field recorders |
Fixed-point and floating-point bit depth
Integer formats such as 16-bit and 24-bit PCM are fixed-point: every value sits on a uniform grid ending at 0 dBFS, and anything louder clips. 32-bit floating point stores a 24-bit mantissa with an 8-bit exponent, so the precision moves with the signal level and values above full scale survive.
| Property | 24-bit fixed point | 32-bit floating point |
|---|---|---|
| Behavior above 0 dBFS | Clips permanently | Preserved, can be lowered later |
| Resolution | Uniform across levels | Relative to signal level |
| File size | 3 bytes per sample | 4 bytes per sample |
| Main role | Recording and delivery | Internal processing and safety recording |
Floating-point files protect against digital overs inside the software, but they cannot rescue a signal that already clipped in the analog front end or the converter. Recorders that advertise 32-bit float capture usually combine two or more converters to cover a wide input range.
Bit depth compared with sample rate and quantization
Bit depth is often confused with neighboring terms, yet each describes a different axis of a digital signal. Bit depth governs the vertical, amplitude axis, while sample rate governs the horizontal, time axis.
| Concept | What it controls | Effect when increased |
|---|---|---|
| Bit depth | Amplitude resolution | Lower noise floor, wider dynamic range |
| Sample rate | Time resolution | Higher Nyquist frequency, wider bandwidth |
| Quantization | The rounding process itself | Finer steps with more bits |
| Dither | Character of quantization error | Turns distortion into benign noise |
| Bit rate | Data per second | Rises with both bit depth and sample rate |
Uncompressed PCM bit rate equals bit depth times sample rate times channels. A stereo 24-bit / 48 kHz stream therefore needs about 2.3 Mbit/s, while a 16-bit / 44.1 kHz stereo stream needs about 1.4 Mbit/s.
Choosing a bit depth in practice
The right bit depth depends on the stage of the workflow. Recording calls for headroom, processing calls for precision, and delivery calls for compatibility and reasonable file size.
| Stage | Recommended bit depth | Reason |
|---|---|---|
| Tracking in the studio | 24-bit | Safe headroom with peaks around -12 dBFS |
| Unpredictable field recording | 32-bit float if available | Protects against unexpected peaks |
| Mixing and processing | 32-bit or 64-bit float internally | Avoids rounding errors in plugins |
| Music distribution | 16-bit or 24-bit | 16-bit exceeds most listening conditions |
| Broadcast and video delivery | 24-bit | Common requirement of delivery specifications |
When reducing bit depth, for example from a 24-bit master to a 16-bit file, dither should be added once at the final step. Without it, quiet fades and reverb tails suffer from quantization distortion that correlates with the signal and sounds gritty.
Misconceptions about bit depth
- More bits do not add musical detail: they lower the noise floor, and audible content above that floor is identical.
- Real converters do not reach 24-bit theory: analog noise limits practical dynamic range to roughly 110 to 125 dB.
- Recording too quietly wastes the benefit: very low levels raise the relative noise, even at 24 bits.
- Converting 16-bit to 24-bit restores nothing: the added bits only contain zeros below the original noise floor.
The short version
Digital audio measures the sound wave thousands of times per second. Bit depth decides how precise each measurement is. Think of a ruler. A ruler with only centimeter marks forces rough rounding. A ruler with fine millimeter marks rounds far less.
Rounding leaves a faint hiss under the music. More bits make that hiss quieter. At 16 bits, the hiss is already below what most people hear at home. At 24 bits, engineers get extra safety while recording.
Where it makes a difference
A classical engineer records a solo piano recital that ranges from near-silent pianissimo to loud fortissimo chords. To avoid any risk of clipping, the engineer sets the preamps so the loudest chords peak around -10 dBFS, and records at 24-bit / 96 kHz.
Because a 24-bit recording has a far lower quantization noise floor, the quietest passages sit well above the digital noise even with that cautious level. After editing and mastering in a 32-bit float session, the engineer creates the 16-bit release file with noise-shaped dither, so the long reverberant fade at the end of the final piece decays smoothly into silence instead of breaking into grainy distortion.