It is how many loudness steps each sound snapshot can be measured against, like rungs on a ladder.
Same wave, two ladders: few rungs round it into a jagged staircase (noise); millions of rungs trace it cleanly.
What it is
Bit depth is how many loudness rungs each digital sample can land on, between dead silence and full scale.
Key facts
Levels = 2 to the power of bits. 16-bit = 65,536 steps. 24-bit = 16,777,216 steps. 32-bit float = same ~16.7 million steps as 24-bit, but a floating exponent shifts them to give a vast usable range.
Rule of thumb: 6.02 dB of dynamic range per bit. 8-bit = 48 dB, 16-bit = 96 dB, 24-bit = 144 dB (theoretical).
Formula: Dynamic Range (dB) = 6.02 x N + 1.76, where N = number of bits.
Real ADC converters cap near 120-124 dB, so 24-bit's full 144 dB is theoretical, not achievable.
16-bit = CD / Red Book (44.1 kHz). 24-bit = studio & live standard. 32-bit float = clip-proof field recorders.
Bit depth controls DYNAMIC RANGE & noise floor; sample rate controls FREQUENCY range. They are separate.
Quantisation: each sample is rounded to the nearest rung. Fewer rungs = bigger error = audible distortion.
0 dBFS = decibels Full Scale = the loudest possible digital value = the hard clip ceiling.
Dither = tiny random noise added when reducing bit depth (e.g. 24 to 16) to mask quantisation distortion.
24-bit WAV is 50% larger than 16-bit at the same sample rate; 32-bit float can recover gain after clipping.
How it works
Mic signal becomes a continuous voltage wave.
Converter takes snapshots (samples) at the sample rate.
Each snapshot's loudness is measured and SNAPPED to the nearest available rung.
More bits = more rungs = finer detail and a lower noise floor.
Rounding to the nearest rung creates a tiny error called quantisation noise.
Dither noise is added on bit-depth reduction to keep quiet fades smooth.
Real examples
CD audio = 16-bit / 44.1 kHz (96 dB range).
Studio multitrack recording = 24-bit / 48 kHz.
Modern field recorders (Zoom, Sound Devices) use 32-bit float so you never clip a take.
Old 8-bit samplers and retro games = crunchy, hissy 48 dB range.
A 24-bit live recording lets you track 18 dB low and still capture a whispered line cleanly.
How it helps in live sound
Record your multitrack and virtual soundcheck at 24-bit, not 16-bit.
Aim peaks around -18 to -12 dBFS; 24-bit's headroom keeps low tracking clean.
Leave 18-20 dB headroom below 0 dBFS so a surprise shout never clips.
Only dither when bouncing a final mix DOWN to 16-bit (e.g. for CD).
Use 32-bit float recorders for unpredictable sources (crowd, speeches) to make clipping impossible.
Don't crank gain to 'use all the bits' on 24-bit; the noise floor is already inaudible.
Everyday analogy
It is like a ruler: 16-bit has thousands of tick marks, 24-bit has millions, so you can measure quiet detail far more finely.
Watch out
Myth: more bits = better sound quality. Reality: bit depth only sets dynamic range / noise floor and clip headroom, NOT frequency response or 'warmth'.
Fun fact
32-bit float files can be 'unclipped' after recording: a take that looks slammed past 0 dBFS can have its gain pulled back in software and come out perfectly clean.
Key takeaways
Bit depth = number of loudness rungs = 2 to the power of bits.
~6 dB of dynamic range per bit: 16-bit = 96 dB, 24-bit = 144 dB.
Bit depth = dynamic range & noise floor; sample rate = frequency range.
Record at 24-bit for safe headroom; dither only when dropping to 16-bit.
0 dBFS is the hard clip ceiling; track with 18 dB of headroom.