Recording - November 2015 - 60

have gains which vary a lot when the bias changes, making them
highly prized for use in compressors. (And I do mean prized; clean
new-old-stock 6386 tubes seem to start at around $100 on Ebay.)
The early tube compressors used a lot of parts, generated a
lot of heat, and required a lot of finicky adjustments. Classic
compressors using tubes as gain control elements, along with
the Fairchilds, include units from Federal and Altec and the
Universal Audio 176, plus the more recent Manley Vari-Mu.
Wanting to avoid tube compressors' fiddly adjustments (not to
mention patents), manufacturers devised circuits that compressed using opto-couplers. These included a light source (originally an incandescent bulb, later an electroluminescent panel
like those found in night lights); when the control voltage got
higher, the light got brighter.
The designers placed a photocell next to the light source.
When light shone on the photocell, its electrical resistance went
down-the brighter the light, the lower the resistance. The circuit
was arranged so that when the photocell's resistance went
down, the volume of the signal was lowered. So you had a
chain of cause and effect: Higher input signal > Higher control
voltage > Brighter light > Lower photocell resistance > Lower signal gain. Classic opto compressors included the Teletronix LA2A (which used tubes for the signal chain, but not for the gain
control element) and the Universal Audio LA-3A, which replaced
the LA-2A's tubes with transistors.
The next compressor circuit developed used a field effect transistor, or FET, for the gain control element. It worked similarly to
the opto-coupler: the circuit was arranged so that a higher control
voltage on the gate of the FET made its resistance go down, and
the lowered resistance decreased the signal gain. The classic FET
compressor was the Universal Audio 1176; it required extensive
tweaking at the factory to keep the distortion low, which increased
the cost, but its sound (especially on vocals and drums) was so
seductive that it rapidly became a fixture in studios.
Most recently, designers have incorporated voltage-controlled
amplifiers, or VCAs, into compressors; VCA-based compressors
include designs from dbx, Drawmer and PreSonus, the Really
Nice Compressor from FMR Audio, and the compressors included in the channel strips of SSL consoles.
Why such a variety of designs? Because all of these compressors sound different, and recording engineers like each for
different reasons. It's worth noting that most of these circuit types
are available as models in software, which is extra-useful.
Putting, say, an actual-factual 1176 on six separate sources
would cost a bundle of money, but in software you can open
multiple instances and use them to your heart's content.
The threshold problem
In the early days of compressors, it was difficult to adjust the
threshold directly. So, working on the principle of "Don't raise
the bridge-lower the river", designers had users adjust the
input level until it exceeded the (unchangeable) threshold of the
compression circuits, and produced the degree of compression
the user wanted-the user would use a combination of gain
reduction metering and ears to decide this. Then, once the relationship of the signal level and the threshold was producing the
desired degree of compression, the user would set the output
control to produce the desired level. This could be called the
Very Inconvenient Method of Threshold Adjustment.
In later designs, as amplification circuits became less expensive,
an actual threshold control could be incorporated; this was one of
the massages applied to the control voltage in the sidechain.
Modern compressors mostly use direct threshold control, and it's
60

RECORDING November 2015

standard in software compressors, except for those which directly
model classics like the LA-2A and 1176 part for part.
Why does anyone use the Very Inconvenient Method anymore? Because, as it happens, some of the best sounding compressors ever made (like the LA-2A and 1176) used it, and people want to get That Sound, even at the cost of inconvenient
operation.
Incidentally, part of That Sound produced when signals go
through classic compressors doesn't come from compression at
all. These circuits incorporate multiple transformers, and the
early ones used multiple tubes as well. The tubes and transformers added their own color to the signal, and engineers like
Geoff Emerick would run signals through, say, a Fairchild compressor with the level set to produce no compression at all, just
because it sounded cool.
One at a time
Compressing an entire mixed signal is important, but it's a discussion that really belongs under the heading of mastering.
These days, compression is used as much on individual instruments and voices as it is on complete mixes.
Remember a few months ago, when I talked about the signal
flow inside a console-how the signal goes through the mic preamp, then the channel fader (perhaps with an equalizer along the
way)? Well, it's possible to interrupt the flow just before the signal
gets to the fader, and put in an insert jack (or a pair of them). The
insert jack can be used to patch in all sorts of external signal processing equipment, but what's relevant to us right now is that this
is the ideal place to put a compressor into the signal chain; it gets
to work with line-level signals (which it likes), but the signal feeding the compressor won't be affected by the fader's setting;
instead the fader controls the level of the compressed signal going
to the mix or the channels of the multitrack recorder.
Why compress individual channels? Well, a little compression
helps a vocal sit nicely in a dense instrumental mix, without the
singer's softest syllables getting buried. It's the exact same problem that the engineers in prehistoric times faced on acoustical
recordings, except that the voice has to ride over music rather
than noise-and it's a lot easier to let the little engineer inside the
compressor do the job than to push the singer back and forth.
Cute compressor tricks
Look again at Figure 2. If you interrupt the sidechain of a feedforward compressor before it gets to the detector and insert
other circuits there, cool things can happen. If you put in a highpass filter (removing signals below, say, 150 Hz-next month
we'll chat about filters when we discuss EQ), then those low-frequency signals (bass notes, kick drum hits) won't make the compressor compress, but the rest of the signal will. This allows the
engineer to make a recording which is nicely compressed, without having the whole thing "pump" in response to those bass
notes and kick drums. API puts exactly this option into its compressor circuits; they call it Thrust®.
What if you put a bandpass filter into the sidechain, and set
it to pass the frequencies around 7.5 kHz and reject everything
else? Well, those frequencies are where vocal sibilants live, and
by including them (and only them) in the signal feeding the
sidechain, you ensure that the compressor will only start compressing when a signal around 7.5 kHz arrives at the input. This
means that the compressor will squeeze down the level of a
nasty sibilant, but will otherwise leave the signal alone, which is
useful if your mic/singer combination has a tendency toward
sibilance. The combination of a compressor and a bandpass filter in the sidechain is sometimes sold as a de-esser.
Thinking about the attack of a compressor, it would sometimes
be nice if it could actually start compressing before a loud sound
arrived. This sounds like science fiction, and it's hard to do in a
hardware compressor, but it's actually rather easy to do in software, particularly if the compressor's being used while already-

Table of Contents for the Digital Edition of Recording - November 2015