System Gain Structure 1225
simple analog voltmeter can measure the rms value of a
sine wave. Complex waveforms require more sophisti-
cated instruments to yield their effective value.
The peak voltage of the waveform must pass through
the sound system component without being clipped. It is
the parameter of interest when establishing the gain struc-
ture of the system. The crest factor of the signal deter-
mines the energy content and therefore the power
produced by the amplifier and delivered to the loud-
speaker. It is of interest when considering the heat that the
loudspeaker must dissipate. Additionally, the rms voltage
is the signal parameter that relates most closely to the
loudness of the signal as perceived by a human listener.
Since the goal of an audio system is to reproduce
appropriate waveforms for a given application, these
waveform principles have universal application for all
parts of the sound system.
33.3 Gain Structure
The following are general terms of gain structure begin-
ning with how it applies to an individual piece of elec-
tronic equipment. It matters little whether the
equipment is a mixer, equalizer, amplifier, or other
active system component. By active we mean that the
device has a power supply for its internal active cir-
cuitry. This can be as simple as an internal battery or
two, or as complex as an internal or external ac
line-powered supply. The power supply voltages estab-
lish the maximum amplitude that a waveform can take
on as it passes through the component, Fig. 33-4. In
audio equipment, most power supplies form a bipolar
set of rails—a fixed positive and negative zero fre-
quency (dc) voltage that the waveform is developed
between. The value of the rail voltage determines the
peak amplitude that the waveform can take on. Exceed-
ing this peak value will cause deformation of the wave,
commonly known as clipping. We will proceed under
the assumption that the rails are fixed, and indeed they
are for most signal processing devices. Some power
amplifier topologies use multivalued or fluctuating rails.
The principles are the same, but we will not consider
such devices here.
Under a no input signal condition, all audio compo-
nents will still emit a residual output signal. Thermal
noise is generated at the molecular level and is present
at the output of all system components whether active
or passive. The level of the thermal noise determines the
noise floor of the component. In practice, other factors
can also make a contribution to the residual noise of an
electronic device. An undersized power transformer or
poor shielding can elevate some frequencies above the
broadband thermal noise floor, Fig. 33-5. Equipment
designers try to minimize thermal noise by component
selection and careful design, but it can never be elimi-
nated. We must accept the fact that it exists. Part of the
reason for establishing the proper gain structure of a
system is to render the effects of thermal noise insignifi-
cant. The thermal noise floor is affected by the settings
of the component’s level controls. While a low noise
floor can be achieved with all controls set at minimum,
this is not realistic, as we cannot operate it that way. The
controls should be set at a point appropriate for the
operation of the device. A good starting point is a
setting that produces the same voltage at the device
output that is present at the device input, often called
unity by audio practitioners. Level controls placed at
their 0 dB setting generally produce this condition, and
represent a good starting point for setting up a system.
Knowledge of the supply rails and noise floor estab-
lishes a dynamic range for the device—the difference in
level between the highest possible undistorted peak and
the lowest level that the signal can take on without
being buried in the noise. The dynamic range is what
can happen when a signal is passed through a device. It
Figure 33-4. Crest factors for sine waves and speech
waves.
0 VA
rms
Sine waveform Speech waveform
+ Rail
Crest Factor = 10log E^ Rail
peak /Erms
Peak
Figure 33-5. Thermal and spectral noise.