Handbook for Sound Engineers

738 Chapter 21

with resistor material and construction. Note that only
the resistive portion of impedance generates noise—
pure inductors and capacitors do not generate thermal
noise. Therefore, in our Shure SM57 circuit model of
Fig. 21-1, thermal noise is generated by the 300: resis-
tance but not by the 6 mH inductance.
In a practical microphone preamp, we are usually
concerned with the signal-to-noise ratio at the output.
Although there may be many sources of internal noise
in the preamplifier and its gain may be varied over a
wide range, for simplicity noise is usually stated in
terms of EIN or equivalent input noise. This simplifica-
tion works because, in a good design, the dominant
noise source is the first amplification stage and subse-
quent stages contribute no significant noise.
As shown in Fig. 21-7, the EIN has three compo-
nents:

1. Et: the thermal noise of the source resistance.


2. En:the voltage noise of the amplifier.


3. In:the current noise of the amplifier.

When noise voltages are produced independently
and there is no relationship between their instantaneous
amplitudes or phases, they are said to be uncorrelated.
Total noise power is the sum of individual noise powers.
Therefore, the resultant voltage is the square root of the
sum of the squares of the individual voltages. For
example, adding two uncorrelated 1 V noises will result
in a 1.4 V noise because

(21-2)

When adding two noises, unless the second is a third
or more of the first (less than 10 dB difference), it will
have little effect on the total. For any amplifier,
minimum total noise is added when the source resis-
tance is such that IN flowing through the source creates
a noise voltage equal to EN. This source resistance is
called the optimum source resistance for that particular
amplifier. Perhaps the most useful function of an input
transformer in a microphone preamplifier is to convert,
as explained in Chapter 11 Audio Transformers, the
impedance of the microphone to this optimum value in
order to maximize SNR.
Measurement of noise is fertile ground for technical
misrepresentation. Some rather unbelievable EIN
numbers have appeared over the years. Most were based
on measurements taken with the preamplifier input
shorted, which ignores the noise contributions of both

RS (source resistance) and IN (amplifier current noise),

leaving only EN (amplifier voltage noise). Bias current

noise generates additional voltage noise when it flows

in the source impedance (not just resistance). In this

case the inductance of our SM57 model will indirectly

contribute real-world noise. To have any meaning at all,

EIN must specify the source impedance. With a 150:

source resistance,

for an ideal noiseless amplifier. If the preamplifier noise

is equal to that of the source, EIN will be 3 dB higher or

130.0 dBV =127.8 dBu. Noise figure, or NF, is a

measure of SNR degradation attributed to the ampli-

fier—in this case 3 dB. From an engineering point of

view there is little point in attempting to achieve NF

below 3 dB.^2

Note that the thermal noise Eq. 21-1 also includes a

term for bandwidth. Noise specifications such as EIN

frequently appear in data sheets without a specified

noise bandwidth. All other things equal, noise increases

as the square root of bandwidth. Therefore, there is

1.25 dB less noise in a 15 kHz bandwidth, and 3 dB less

noise in a 10 kHz bandwidth, than in a 20 kHz band-

width. Likewise, while measurements such as

A-weighted noise are both legitimate and useful, they

cannot be directly compared to unweighted measure-

ments. When comparing noise specifications, be sure

it’s an “apples to apples” comparison.

21.1.2.4 Bandwidth and Phase Distortion

Performance in the time domain, or waveform fidelity,

is critically important to accurate music reproduction.

Accurate time domain performance, sometimes called

transient response, requires low phase distortion. Pure

time delays exhibit a linear phase versus frequency

characteristic. True phase distortions are expressed as

DLP or deviations from this linear phase relationship.

Phase shift is not necessarily phase distortion.^3 In order

E 12 += 12

2=

1.414.=

Figure 21-7. Contributions to equivalent input noise.

RS

Et

Signal

Real mic Real preamp

EN

IN ZiZero noise

amp

EIN=223 nVrms

–= 133.0 dBV

–= 130.8 dBu