I am a bit confused on bit depth regarding analog to digital conversion...

During the ADC process, I understand that bit depth essentially represents a number of levels that sample points can be mapped to (ie 16bit has 2^16 possible levels for samples to snap (quantize) to and 24 bit has 2^24 possible levels).

What I am confused about is how these bit depth levels are spread out over the y axis for a given analog waveform graph during the analog to digital conversion process.

For example, if I have some circuitry that can create an analog waveform with maximum and minimum values between -5V and +5V, and a resulting waveform is passed to an ADC to undergo the conversion process, are the bit depth levels spread out evenly to span this complete theoretical 10V range, or are they only spread out across the actual highest and lowest voltage points for that very specific standalone waveform?

Furthermore, why is it not problematic if ADCs are converting analog signals with different ranges (given the same bit depth), wouldn't the digital output at that point be representing different things for different signals (ie level 1 (bit 1) might represent 0.001V for one converted signal whereas level 1 might represent 0.002V for a different converted signal with a smaller voltage range). Wouldn't this lead to all kinds of consistency problems in the digital realm?

1 Answer 1


This answer is about studio type audio digitization, creating the typical .wav file. In other specialty areas things may be different.

The AD converter has same step size all over its range. Or as you put it the bit levels are spread out evenly.

What you might think about is that in front of the ADC there often is a gain control. This may be set differently at different times. This is similar to how you have a volume control when listening to sound.

The recording engineer sets that gain to achieve two slightly contradicting goals: first to never go above the max signal the ADC can handle (this would create nasty distortion), secondly to keep a healthy distance between the wanted signal and the noise floor. Quantization introduces an error that the ear hears as a (more or less) white noise. In my experience this was difficult when recording at 16 bits, very much easier at 24 bits, and today with 32 bits some recording devices does not even have any gain adjustment because it is not needed.

In the later distribution of sound, say as old time CD, on youtube or whatever, a processing step is added in the production where the volume is digitally adjusted to a normal volume regardless of how it was recorded. This is done to allow the listener to keep the volume control at the same setting regardless when changing to another song (although the reason is slightly more involved than that). The actual sound volume when recording is not preserved, and whatever voltages the ADC saw is at that stage uninteresting.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.