Why does a recording not use 16 or 24 bits per sample instead of 10? · Lobeworks/17

Because once the step of the converter is finer than the noise of the amplifier in front, more bits only write the noise with more digits.

Why does a recording not use 16 or 24 bits per sample instead of 10?

Because once the step of the converter is finer than the noise of the amplifier in front, more bits only write the noise with more digits. Neuropixels 1.0 has about 6 µV of input noise in its spike band, and with its gain 10 bits already put the step below that. It is a bathroom scale that wobbles by a few grams: a display with seven decimals shows the wobble, never more weight.