Bit Depth: What 16-Bit and 24-Bit Audio Actually Give You

What it actually controls

Digital audio takes an amplitude measurement at each sample. Bit depth is how many binary digits are used to record that measurement, and therefore how many distinct levels are available.

16 bits gives 65,536 levels. 24 bits gives 16,777,216.

What those extra levels buy is range, not resolution in the everyday sense. This is the single most common misunderstanding about the subject: bit depth does not make the sound more detailed, it lowers the floor beneath it. The signal you can hear is described with the same accuracy either way; what changes is how far down the quiet parts can go before they run into noise.

The 6 dB rule

Each bit doubles the number of available levels, and doubling amplitude is about 6 dB. So:

theoretical dynamic range = bit depth x 6.02 dB

Which gives:

• 16-bit: about 96 dB
• 24-bit: about 144 dB
• 32-bit float: an enormous range, and used for a different reason (see below)

For context, the difference between a quiet listening room and a level that will damage your hearing is roughly 80 to 90 dB. 16-bit already covers the usable range of human hearing in a domestic space, which is why the format has held up for four decades.

Why quantisation error is not the problem people think

Rounding each sample to the nearest available level produces an error. If that error correlates with the signal it sounds like distortion, which is genuinely unpleasant.

Dither solves this. Adding a tiny amount of carefully shaped noise before quantising decorrelates the error from the signal, converting it from distortion into a low, constant, benign noise floor. That is a good trade, and it is standard practice in every competent mastering chain.

Properly dithered 16-bit audio has a noise floor around 96 dB down and no signal-correlated quantisation distortion at all. This is the reason the "16-bit sounds harsh" claim does not survive contact with a level-matched blind comparison.

When 24-bit genuinely matters

It matters most in production, and that is not a hedge:

• Recording headroom. You cannot know a peak before it arrives. Extra range lets an engineer set levels conservatively without pushing quiet material into the noise floor.
• Processing. Every EQ, compression and gain stage introduces rounding. Working at higher depth keeps accumulated error far below anything audible through a long chain.
• Very wide dynamic material. Some orchestral and film work legitimately uses more range than 16-bit comfortably provides.

For playback of a finished master, the case is much weaker. A 24-bit delivery file lowers a noise floor that was already below the noise floor of your room, your amplifier, and in most cases the recording itself.

32-bit float is a production format specifically: it makes clipping recoverable, because the value is stored with an exponent rather than a fixed ceiling. It is not a playback improvement.

What bit depth does not do

• It does not set frequency response. That is sample rate.
• It does not add detail to a signal above the noise floor.
• It does not improve a recording that was made poorly.
• It does not survive being upconverted. Padding 16-bit audio to 24 bits adds zeros, not information.

The practical summary

Buy the master, not the number. A well recorded, well mastered 16-bit release beats a poorly mastered 24-bit one every time, and the difference between those two masters will be vastly larger than the difference bit depth could ever make. Where the same master is offered in both, take the higher depth if it is free and do not pay a premium for it.