Digital Sound & Music: Concepts, Applications, & Science, Chapter 5, last updated 6/25/2013
Let’s look more closely at what’s going on when you set the input level. Any hardware
system has a maximum input voltage. When you set the input level for a recording session or
live performance, you’re actually adjusting the analog input amplifier for the purpose of ensuring
that the loudest sound you intend to capture does not generate a voltage higher than the
maximum allowed by the system. However, when you set input levels, there’s no guarantee that
the singer won’t sing louder than expected. Also, depending on the kind of sound you’re
capturing, you might have transients short loud bursts of sound like cymbal claps or drum
beats to account for. When setting the input level, you need to save some headroom for these
occasional loud sounds. Headroom is loosely defined as the distance between your “usual”
maximum amplitude and the amplitude of the loudest sound that can be captured without
clipping. Allowing for sufficient headroom obviously involves some guesswork. There’s no
guarantee that the singer won’t sing louder than expected, or some unexpectedly loud transients
may occur as you record, but you make your best estimate for the input level and adjust later if
necessary, though you might lose a good take to clipping if you’re not careful.
Let's consider the impact that the initial input level setting has on the dynamic range of a
recording. Recall from Section that the quietest sound you can capture is relative to the
loudest as a function of the bit depth. A 16-bit system provides a dynamic range of
approximately 96 dB. This implies that, in the absence of environment noise, the quietest sounds
that you can capture are about 96 dB below the loudest sounds you can capture. That 96 dB
value is also assuming that you're able to capture the loudest sound at the exact maximum input
amplitude without clipping, but as we know leaving some headroom is a good idea. The quietest
sounds that you can capture lie at what is called the noise floor. We could look at the noise floor
from two directions, defining it as either the minimum amplitude level that can be captured or
the maximum amplitude level of the noise in the system. With no environment or system noise,
the noise floor is determined entirely by the bit depth, the only noise being quantization error.
In the software interface shown in Figure 5.21, input levels are displayed in decibels-full-
scale (dBFS). The audio file shown is a sine wave that starts at maximum amplitude, 0 dBFS,
and fades all the way out. (The sine wave is
at a high enough frequency that the display of
the full timeline renders it as a solid shape,
but if you zoom in you see the sine wave
shape.) Recall that with dBFS, the maximum
amplitude of sound for the given system is 0
dBFS, and increasingly negative numbers refer to increasingly quiet sounds. If you zoom in on
this waveform and look at the last few samples (as shown in the bottom portion of Figure 5.23),
you can see that the lowest sample values the ones at the end of the fade are 90.3 dBFS.
This is the noise floor resulting from quantization error for a bit depth of 16 bits. A noise floor
of 90.3 dBFS implies that any sound sample that is more than 90.3 dB below the maximum
recordable amplitude is recorded as silence.
Why is the smallest value for a 16-bit sample
90.3 dB? Because
( )
Previous Page Next Page