900 Chapter 25
across them is kept low (in this instance, 30 dBu, the
lower the better) it can be quite acceptable for a lot of
applications. Keeping the signal low necessitates
following gain to bring the signal back up again. That
means amplifier noise.
There is a wide variety of possible voltage-control
elements and Fig. 25-76E shows a smoother version of
the transistor limiter using a JFET. The principle is
much the same, only the side chain and VCA circuitry
has developed somewhat. FETs have such spread char-
acteristics that a preset is necessary to set their bias
points. In normal operation the FET needs to be biased
just nonconducting; that is, nonattenuating. This neces-
sary adjustment also provides a means of varying the
output level at which the limiter starts to bite, and inci-
dentally some control over the ratio of the gain reduc-
tion. (The greater the bias, the more control voltage
signal is needed to be generated before the FET turns on
and starts attenuating, where it does so in a tightly
controlled manner. A low bias results in a lower, or
indeed no, threshold and a far gentler gain-control ratio.
Carefully trading these two— a high threshold for a
hard ratio or low and “smushy” threshold and gentle
FET turn-on—against overall gain formed the basis of
FET-based compressors such as the famed Audio and
Design F760 and the UREI 1176LN.)
FETs have very high gate impedances, precluding
the need for a follower. The control voltage is summed
at the device gate with a sample of the input signal.
Automodulation is an effect of FETs where the
source-drain resistance (the resistance we’re depending
on as part of the input attenuator) varies with signal
voltage across it. This is attacked in two ways: first by
keeping the signal across the FET low, as in the tran-
sistor limiter, and second by supplying the gate with
some anti-wobble signal that does a fairly good job of
forcing the source-drain path to wobble against and
largely cancel the automodulation effect.
25.12.1.1 Side-chain Time Constants
Between the rectifier and the FET in Fig. 25-76E is a
simple resistor-capacitor network that determines how
the side chain works and its effect on the automatic gain
reduction. This is in contrast to the transistor circuit
where the reservoir capacitor discharges through the
transistors at one end and is charged rapidly through the
diode from the other. Here we can adjust the rate at
which the capacitor charges and at which it discharges.
The implications of these on how the circuit behaves
and sounds are crucial.
But why have time constants at all? If the idea is to
provide protection for overloads, why bother with how
they’re handled? Well that’s all diode clippers do, Figs.
25-76A and B. They have zero attack time, which
means there is no delay or run-up to them when dealing
with an overload. Similarly, they have a zero release
time, meaning that once the overload is dealt with it’s
instantly business as usual. The trouble is, they sound
horrible, as would either the transistor or the FET
limiter with infinitely short parameters.25.12.1.2 Attack TimeFig. 25-77 shows the first few cycles of a train of sine
waves that are in excess of the limiter threshold and the
effect as the limiter tries to reduce the output to the
prescribed level. Fig. 25-77A shows a zero attack time
and not unexpectedly looks very sawn off. Lengthening
the attack time somewhat, Fig. 25-77B, leaves a recog-
nizable but mutilated crest, while longer still is even
less bent, Fig. 25-77C. Unfortunately, the character of
distortion products generated by this effect are very
audible; they are loud (since it is a loud signal that is
subject to control) and of high order and unlikely to be
masked by the fundamental signal. Even at this stage it
is clear that a longer attack time takes less toll of the
input signal integrity; the less a waveform is modified,
the better it will sound. Expanding the time scale to
many cycles shows how lengthening attack time looks.
A long attack time, Fig. 25-77F, gradually reduces the
limiter gain until the signal is completely under control
while imposing less immediate distortion on it.
The tradeoff is apparent. An attack time long enough
to not mangle the program material also permits excess
output level for as long as the circuit takes to bring the
gain down sufficiently. Balancing this overshoot against
leading-edge distortion due to short attack times is a
subjective compromise.
Naturally the lower the frequency, the longer the
time period between cycle crests and the greater
distorting effect of attack times. An adequate
attack-time for high frequencies can easily be still way
too short for low frequencies, while an adequate attack
time for bass is unnecessarily long for highs. That’s life.
It is normal in a high-quality studio dynamics section
to use a full-wave rectifier for the side chain (as
opposed to the half-wave shown in these two exam-
ples). This gives twice as many opportunities per cycle
to sense and adjust the gain (one on the positive-going
peak, one on the negative) in addition to allowing for
the fact that few real-world signals are symmetrical;