908 Chapter 25
25.12.5.1 rms Detection
Hitherto, detection of signal levels in dynamics had
been either peak or average. These were actually
achieved by broadly similar circuitry with the difference
dictated by the attack time applied after signal rectifica-
tion; short attack times allowed the reservoir capacitor
to charge immediately to the highest signal level
applied, while longer attack times tended to smooth out
the peaks, settling on an average value of the applied
rectified waveform. And a sort of mushy continuum
existed between the two.
Rms (root mean square) detection has the intent of
providing a measure of the energy in an applied wave-
form, the actual power. The reasoning is that a power
measurement could be considered more equivalent to
loudness. Rms is achieved by first squaring the applied
signal (i.e., multiplying it by itself, not turning it into a
square wave), finding an average value of that squared
source, and then determining the square root of that
average (unsquaring it). Seems like a lot of bother to go
to, doesn’t it? Well, no one would have bothered if there
was a reasonably straightforward method. This comes
from an application of a log anticell.
A precision-rectified (meaning accurate down to
very low levels) input signal is logged, and then its
output is doubled (added to itself); doubling a log value
squares the number it represents. This signal is then
integrated with a time constant long enough to allow
reasonable averaging of the lowest frequency under
consideration; this incidentally defines the minimum
attack time of the processor. This log value average is
then halved (division of a log value by two is the same
as finding the square root), delivers a log-world rms-
detected output. (In this circumstance a subsequent
antilog conversion is unnecessary. Actually, the square
rooting is ignored at this point, too, since it can be
achieved in a later scaling exercise.) The good news is
that all that can be done with a handful of transistor
junctions. Release time can be extended with a
following buffered capacitor, but often the imbued time
constant of the rms detection serves as symmetrical
attack and release. This somewhat leisurely time
response (necessary to permit good rms detection at low
frequencies, devoid of distortion-creating ripple) in and
of itself ensures that the behavior of such a dynamics
section can’t get too wild and interesting, but by corol-
lary such processors do afford probably the least intru-
sive method of automatic volume control, which is a
highly prized attribute on occasion.
25.12.5.2 ThresholdingThe rms-detected control signal is then masked in a
threshold determining circuit; typically this is a preci-
sion rectifier with its reference point determined by a
threshold control voltage—the purpose of this is to
ignore all variation of the detected voltage until it
exceeds (in the case of a compressor) the threshold
point, beyond which its output follows the rms detector
output. Any control signal escaping the thresholder still
has a dB/V characteristic, being still logged, following
the input signal. Another (linear-think) way of looking
at this is that a division takes place; the detected control
signal is divided by the threshold, but with any result
less than 1 masked out at 1, only greater than unity
results being passed.
If the thresholder is designed to pass only changes
below the threshold, then the low-level effects of expan-
sion and gating are possible, signals above threshold
being ignored. Separate thresholders and following
conditioners are necessary for each desired function of
the dynamics section.25.12.5.3 RatioIf this thresholded control voltage were applied directly
to the (level-adjusted) control port on the VCA, some-
thing odd would happen: nothing. More precisely, above
the threshold the control signal would rise in accord with
a rising applied audio signal to the precise extent that the
gain reduction resulting from it would be exactly the
same as the increase in signal. The VCA output would
remain at a fixed level for any applied signal level above
the threshold. In other words, it is a compressor with an
infinity:1 ratio, meaning that above the threshold, any
amount of signal level variation will have no effect on
the output. In yet other words, it would be a limiter
(albeit with slow dynamic response).
Introducing a variable attenuator in the feed to the
VCA control port from the thresholder affords altering
the amount of dynamic gain reduction; less control
signal variation, less gain reduction. A nice feature of
this very simple approach in log world is that this atten-
uation (equivalent to solving for variable roots in linear)
results in precise applied signal level to dB gain reduc-
tion ratios; for a given setting, if the input signal were to
rise 6 dB, the output would rise only 3 dB; this ratio,
2:1, would obtain linearly for any applied signals above
the threshold.
Astute readers may wonder what would happen if
the control signal, rather than being attenuated, was