860 Chapter 25
Another good reason for not terminating with an
equal or any fairly low resistance is the effect on micro-
phone response and subjective quality. Having an induc-
tive characteristic, the dynamic microphone capsule has
an impedance that steadily rises with frequency,
becoming predominant at high audio frequencies where
the inductive reactance of the source is large with
respect to the coil-winding resistance. When terminated
with a relatively low resistance, the complex impedance
of the capsule and the termination resistor form a
single-order 6 dB/octave low-pass filter, gracefully
rolling off the high-frequency output of the microphone.
Not too useful.
With a fairly hefty cable capacitance, the system is
no longer graceful; the complete network now looks
like a rather rough second-order filter. There isn’t too
much to be done about that; regardless of termination
method, cable capacitance is here to stay and is always a
consideration unless the preamp is remoted to or close
to the microphone itself.
25.10.2 Optimizing Noise Performance
Amplifiers are not perfect. For noise criteria, the first
device that the signal hits in the amplifier is the key one,
since the noise it generates usually masks—by a large
margin—noise from all succeeding stages.
All practical amplifying devices are subject to a
variety of internal noise-generating mechanisms,
including thermal noise generation. When measured,
these give rise to some important values; namely, the
input noise voltage, the input noise current, and the ratio
between those two that is in effect the input noise
impedance. This becomes all important in a little while.
For the most part, bipolar transistors—either standard or
more usually large-geometry and sometimes multiparal-
leled—are used as front-end devices both in discrete
designs and op-amp IC packages in this application so
much of the following relates to both packages.
These noise voltages and currents alter in both indi-
vidual magnitude and ratio to each other with differing
electrical parameters, especially collector current.
Predictably, as this current decreases, so does the noise
current (most of the noise is due to minor random
discontinuities in device currents); the ratio between the
noise voltage and current—or noise impedance—may
be altered in this fashion.
Thermal noise generation is common to all resistive
elements. The amount is related to both the temperature
and the bandwidth across which it is measured; an
increase in either will increase proportionally the noise
power generated. Under identical circumstances, the
noise power that is generated by any values of resis-
tance is the same. Differing resistor values merely serve
to create differing ratios of noise voltage and noise
current; the product of the two always equals the same
noise power. This particular noise phenomenon, thermal
noise or Johnson noise, is totally unavoidable because
the nature of atomic structure is such that when things
get hot and bothered, they grind and shuffle about
randomly, creating electrical disturbances white in
spectra (i.e., equal energy per cycle bandwidth).
Even the real (resistive) part of the complex imped-
ance of a dynamic microphone generates thermal noise;
this ensures that there is a rigidly defined minimum
noise value that cannot be improved upon.25.10.3 Noise FigureThe difference between the noise floor defined by
thermal noise and the measured noise value of a prac-
tical system is known as the noise figure (NF) and is
measured in decibels (Noise Figure = System Noise
Theoretical Noise). The noise output from a resistor or
the real part of an impedance is calculable and predict-
able—Herr Boltzmann rules. A direct comparison of the
noise voltage measured at the output of an amplifier due
to a resistor applied to the amplifier input versus the
noise voltage expected of the resistor on its own is
possible just by simply subtracting the measured gain of
the amplifier. This is a measure of NF.
An interesting effect occurs when, with any given set
of electrical parameters set up for the amplifier front-end
device, the source resistance is steadily changed in
value. A distinct dip in the NF occurs, Fig. 25-38, and
the value of the resistor at which this dip occurs changes
as the device parameters are changed (collector current
primarily). For the usually predominant noise mecha-
nism (thermal noise), a minimum NF occurs with a tiny
amount of collector current (5–50μA) and a high source
resistance (50 k: up). Without diving into the mathe-
matics, the nulling is a balancing of interaction between
the external noise source and the internal voltage and
current noise generators.25.10.4 Inverse-Frequency NoiseThere is another major noise mechanism inherent to
semiconductors. It is the low-frequency (inverse level
with respect to frequency) noise—a burbly,
bumping-type noise caused by the semiconductor
surface generating and recombining sporadic
currents—most prevalent in dirty devices but present to
a degree in all. It is subjectively apparent and has to be