Digital Sound & Music: Concepts, Applications, & Science, Chapter 5, last updated 6/25/2013
29
underuse your available dynamic range, you might run into problems when you try to run any
long fades or other processes affecting amplitude, as demonstrated in the practical exercise “Bit
Depth and Dynamic Range” linked in this section.
It should be clarified that increasing the input levels also increases any background
environmental noise levels captured by a microphone, but can still benefit the signal by boosting
it higher above any electronic or quantization noise that occurs downstream in the system. The
only way to get better dynamic range over your air conditioner hum is to turn it off or get the
microphone closer to the sound you want to record. This increases the level of the sound you
want without increasing the level of the background noise.
So let's say you're recording with a bit depth of 16 and you've set your
input level just about perfectly to use all of that dynamic range possible in your
recording. Will you actually be able to get the benefit of this dynamic range
when the sound is listened to, considering the range of human hearing, the range
of the sound you want to hear, and the background noise level in a likely
listening environment? Let's consider the dynamic range of human hearing first
and the dynamic range of the types of things we might want to listen to. Though
the human ear can technically handle a sound as loud as 120 dBSPL, such high
amplitudes certainly aren’t comfortable to listen to, and if you’re exposed to
sound at that level for more than a few minutes, you’ll damage your hearing.
The sound in a live concert or dance club rarely exceeds 100 dBSPL, which is pretty loud to
most ears. Generally, you can listen to a sound comfortably up to about 85 dBSPL. The quietest
thing the human ear can ear is just above 0 dBSPL. The dynamic range between 85 dBSPL and
0 dBSPL is 85 dB. Thus, the 96 dB dynamic range provided by 16-bit recording effectively
pushes the noise floor below the threshold of human hearing at a typical listening level.
We haven't yet considered the noise floor of the listening environment, which is defined
as the maximum amplitude of the unwanted background noise in the listening environment. The
average home listening environment has a noise floor of around 50 dBSPL. With the dishwasher
running and the air conditioner going, that noise floor could get up to 60 dBSPL. In a car
(perhaps the most hostile listening environment) you could have a noise floor of 70 dBSPL or
higher. Because of this high noise floor, car radio music doesn't require more than about 25 dB
of dynamic range. Does this imply that the recording bit depth should be dropped down to eight
bits or even less for music intended to be listened to on the car radio? No, not at all.
Here's the bottom line. You'll almost always want to do your recordings in 16 bit sample
sizes, and sometimes 24 bits or 32 bits are even better, even though there's no listening
environment on earth (other than maybe an anechoic chamber) that allows you the benefit of the
dynamic range these bit depths provide. The reason for the large bit depths has to do with the
processing you do on the audio before you put it into its final form. Unfortunately, you don’t
always know how loud things will be when you capture them. If you guess low when setting
your input level, you could easily get into a situation where most of the audio signal you care
about is at the extreme quiet end of your available dynamic range, and fadeouts don’t work well
because the signal too quickly fades below the noise floor. In most cases, a simple sound check
and a bit of pre-amp tweaking can get you lined up to a place where 16 bits are more than
enough. But if you don’t have the time, if you'll be doing lots of layering and audio processing,
or if you just can’t be bothered to figure things out ahead of time, you’ll probably want to use 24
bits. Just keep in mind that the higher the bit depth, the larger the audio files are on your
computer.
Practical
Exercise:
Bit Depth and
Dynamic
Range
Previous Page Next Page