Handbook for Sound Engineers

Consoles 861

considered. Measured alone, low-frequency noise has

its own set of collector current and source resistance

nulls, usually far higher in current and lower in resis-

tance than for thermal noise.

Commonly known as 1/F noise—implying its

predominance at very low frequencies—it is often spec-

ified by way of the frequency at which it is contributing

the same amount of noise as the device’s effective

thermal noise. Below this knee frequency 1/F noise

predominates. A good clean device will have a knee

frequency below 10 Hz; judicious filtering along the

signal path can render 1/F noise unimportant within a

system, but it remains a serious consideration within

each individual amplifier stage.

25.10.5 Optimum Source Impedance (OSI)

A compromise has to be struck. To make a generaliza-
tion, a 100μA collector current and a 10 k: source
impedance for a typical low-noise pnp transistor seem
about right. (Pnp transistors are commonly used in this
area due to slightly better low-frequency performance
figures over npn types.) The source resistance value is
that at which the device is optimally quiet for audio
purposes and is known as the optimum source impedance
(OSI). Incidentally, this impedance has absolutely
nothing to do with the kind of circuit configuration in
which the device may be. Whether it be in a
common-base amplifier with an input impedance of 50:
or in a totem-pole front end with bootstrapping and a
consequential input impedance of over 10 M: , it doesn’t
matter. The source impedance for optimum noise perfor-
mance stays at 10 k: , or whatever, provided that the
collector current is the same in all cases. Optimum source
impedance has nothing to do with input impedance.

This optimum impedance varies depending on the

type of input device used. For an FET, the noise figure

typically obtainable drops to an amazingly low value

but, unfortunately, at a substantially useless impedance

of several dozen megohms. Even supposing it were

practical to provide a source impedance of that magni-

tude, the whole arrangement would be so sensitive to

any electromagnetic fields (such as RF) that even tiny

amounts present would obliterate the noise advantage.

The design and construction of capacitor microphones

using FET front-ends highlight the hazards; the end

results often show such capacitor microphones can be

several dB noisier than a dynamic micro-

phone/front-end combination.

Good bipolar transistors have OSIs in the region of

5k:to 15 k:, whether discrete or as part of an IC

amplifier package. Fortunately, these values closely

coincide with the source resistance value that provides

for optimum flatness of device transfer characteristics.

This helps a long way toward best frequency versus

phase linearity, which translates to enhanced stability in

a typical high negative-feedback amplifier configuration.

Fig. 25-39 shows the effect of altering the source

impedance into such an amplifier (using a conventional

bipolar transistor input device) on output frequency

response. The droop is due to the excessively high

source impedance reacting against the device

base-emitter, board, and wiring capacitances to form a

low-pass filter. The high-frequency kink is a practical

effect of the curious mechanism; when a bipolar tran-

sistor is fed from an impedance approaching zero, its

high-frequency gain-bandwidth characteristic extends

dramatically, radically altering the phase margin and,

consequently, the stability of an amp designed and

compensated for more ordinary operating circum-

stances. The kink is a resonance within the amplifier

loop caused by erosion of phase margin resulting from

this mechanism. It is but an uncomfortably short step

from oscillation.

As can be seen from the graph in Fig. 25-39, the

response is maximally flat at a source resistance of

around 10 k:, about the same value as the OSI for

optimum noise performance for the same configuration.

A problem to reconcile is that our practical source

impedance is nominally 200: for a dynamic micro-

phone, whereas the OSI for the best conventional input

devices is around 10 k:. How do we make the two fit?

25.10.6 Microphone Transformers

There are some horrible stories about how bad trans-

formers are. Properly designed and applied, however,

Figure 25-38. Bipolar noise curves (noise figure curves for a
good pnp front-end transistor for collector current versus
source resistance).

8 7 6 5 4 3 2 1

Source resistance sketch graph

100 71 k 7 10 k 7 100 k7

1 mA

Collector current

100 MA

10 MA

Frequency - 1 kHz, NF–dB