Bit depth is the value that shows how many bits are assigned to each sample of digital audio. It is written as 16-bit, 24-bit, and so on, and the larger the value, the more amplitude steps a single sample can express. It is a separate element from sample rate, which decides the resolution in time.
Relation to dynamic range
The greater the bit depth, the lower the quantization noise in theory, and the wider the range you can record from the clipping ceiling down to very small signals. Recording at 24-bit makes it easier to leave headroom while keeping enough precision, without pushing the input right up to the limit.
16-bit and 24-bit
16-bit has long been used for CDs and general distribution files, while 24-bit is common for recording, editing, and exchanging mix material. The basic approach is to keep a generous bit depth while you work and to reduce it on export to suit the final use. Check the delivery requirements for supported formats.
How 32-bit float differs
The 32-bit float used inside DAWs and in some recorded files represents numbers differently from fixed-point 24-bit. Internal calculations can hold values above 0 dBFS, but that does not mean clipping that occurred at the A/D stage of the mic input can be repaired afterward.
Not the same as bit rate
Bit depth is the amount of information used for one sample of PCM audio. The bit rate shown in kbps for MP3 and AAC is the amount of data used per second. The names are similar, but they are not the same setting.
When you reduce bit depth
When you convert from a higher bit depth to 16-bit or similar, dither is sometimes added to make the quantization error less noticeable. Dither is normally applied only once, at the final stage of production. Where to set it in Logic Pro is shown in How to change bit depth and sample rate (Japanese).
