Handbook for Sound Engineers

938 Chapter 25

signal and of the control voltage (CV); all that appears at
the output is the product of the applied signal and CV.
Product implies it is the result of multiplication, which it
indeed is. The circuit is the basis of a good if hardly
perfect analog multiplier. Better yet, since it uses good
old transistors with their exponential base-voltage to
collector-current response, the control law is for the large
part linear with respect to decibels of gain and attenua-
tion. Which is why all the bother and complication.

25.16.2.3 Log-Antilog Cell

A different approach, resulting in a different internal
topology is what could be called the log-anti-log
approach and is schematically described for simplicity
(although the actual integrated implementation is
nothing like this) in Fig. 25-108.
It again relies on the exponential relationship
between the base voltage and collector current of the
transistor. The first stage is a log convertor, converting
the (positive-going in this example) input signal into a
(negative-going) logarithmically representative voltage;
summed in with this is the control voltage, which, since
we’re in log domain, is linear voltage per dB; the
composite is then antilogged back into the real world.
The effect of adding or subtracting the control voltage is
to increase or decrease the linear end-to-end gain.

25.16.2.4 Commercial VCAs

Commercial IC VCAs typically use one or other of
these approaches; VCAs are almost always acquired
and used in IC form. One built out of discretes will
work, of course, but the inherently much closer
matching of active semiconductors on the same
substrate reduces much matching of parts and tweaking
out of various offsets, and the manufacturers have gone
to the bother of thermally compensating and biasing
everything up such that it “plays nice” with the real
world. Nevertheless, for optimum operation of this
arrangement or circuits based on it, preset adjustments
are the norm, even for integrated versions. Stabilization
of operation against temperature is a further complica-
tion, if perhaps less so for consoles that will spend their
lives in air-conditioned environments.
Beyond the basic and remarkably well-performing
circuit element, there are, of course, other issues that
come along with real live electronics. A prime consider-
ation is noise; the best operating points for the transis-
tors vary depending on the parameter that needs to be
optimal (see the discussion on microphone amplifiers

for collector-current versus noise); what may be right

for noise almost certainly isn’t right for adequate

large-signal handling. Attempts have been made to

provide for both by altering the bias point of the transis-

tors dynamically in accord with applied signals, such

that they’re closer to right for both the low-level region

(noise) and high-level operation. Another approach has

been to parallel up many of the IC VCAs in order to

improve the combined devices’ noise-voltage to

noise-current ratios to improve noise and better suit

operation at ordinary audio signal impedances and

levels (this goes hand-in-hand with paralleling or using

multiparallel input transistors to optimize OSI in trans-

formerless microphone amplifiers).

Input buffering and conditioning of the control

signals makes them easy to use; already quite linear, the

linearity can be extended over a greater control range

and can be arranged to be changed at so many dB per

volt of control signal (say, 20 dB/V) as to be conve-

nient with the A/D and D/A converters in an automation

system and the voltage swing off dc-driven faders. Typi-

cally, though, the control port sensitivity on integrated

VCAs can be much higher than this—a few mV per

dB—and needs to be treated with respect.

25.16.2.5 Control Voltage Noise

This discussion highlights a crucial design issue—that

of ensuring a very quiet control signal. This might seem

an odd concern until one realizes that at typical control

sensitivities, mere mV of undesired ripple or noise on

that control line will modulate the through audio notice-

ably. Note modulate. Since the balanced modulator that

is the VCA will not permit control voltage itself into the

output, a real audio signal has to be passing through the

VCA for this modulation to take place. This, more than

anything, is the underlying cause of VCAs’ largely

undeserved reputation for sounding dirty. Like all these

kinds of aspersions, there is a germ of a reality behind

them. And this one is CV noise. Any circuitry involved

in the CV should be handled with the care one would

apply to the “real” signal path; there is a very real temp-

Figure 25-108. Log-Antilog VCA Principle.

+

(^0)

• 0
  1
  Control
  voltage
  'Log
  domain'
  Logarithmic
  convertor
  Antilogarithmic
  convertor