Gates, De-essers and Sidechains

The compression lesson describes a device that turns a signal down when it gets too loud. Everything in this lesson is that same device with one thing changed: what it is listening to.

That is the whole idea, and it is worth holding onto before any of the controls make sense. A compressor has two halves that are easy to mistake for one. There is the part that processes the audio, and there is the part that watches a signal and decides when the processing should happen — the detector, or sidechain. Feed the detector something other than the audio itself, or invert what it does when the threshold is crossed, and you have a gate, a de-esser, or a ducker. Three different boxes on the shelf, three different problems solved, one mechanism underneath.

Gates: Turning the Signal Down When It Is Too Quiet

A gate inverts the compressor's decision. Above the threshold, the signal passes untouched; below it, the gate turns the signal down. It exists because microphones do not only hear what you pointed them at.

The classic case is a drum kit. The tom microphones spend most of a song hearing cymbals, snare and the room, and the toms themselves for perhaps four bars of it. Every open channel adds that bleed to the mix, and the bleed arrives at slightly different times from slightly different distances, so it also smears the kit's transients. Gate the toms and they contribute only when they are hit.

Threshold is where the gate decides the source has arrived, and the thing to understand is that you are not setting it by the level of the tom. You are setting it in the gap between the tom and the loudest bleed — a gap that may be narrow, and that a hard-hitting drummer with quiet toms may close altogether. When the gap is not there, the gate cannot find it, and the answer is a microphone problem rather than a gate problem. Which pattern the mic has and where it points decides how much bleed exists in the first place, as the microphones lesson sets out.

Range, sometimes called depth, sets how far down the signal goes when the gate is closed. The instinct is to close it completely, and it is nearly always the wrong instinct. A tom that vanishes into silence and then snaps into existence draws attention to the processing; a tom attenuated by 15 or 20dB when idle solves the bleed problem and still sounds like a drum in a room. Reach for full attenuation only when total silence is the point.

Attack has to be fast or the gate eats the transient it was opened by, and the transient is most of what a drum is. Hold is the minimum time the gate stays open once triggered, and it exists for exactly one reason: without it, a signal that hovers near the threshold makes the gate open and shut repeatedly, producing a stutter that is far more distracting than the bleed. Release governs how it closes, and setting it too fast truncates the decay so the tom stops dead rather than ringing out.

Most gates also apply hysteresis — they open at the threshold you set but do not close again until the level falls somewhat below it. That asymmetry is the other half of the anti-chatter mechanism, and it is why a gate that seems to have no chatter setting usually still behaves sensibly.

Expanders: the Same Job With the Volume Knob Turned Down

An expander does what a gate does, gently. Rather than dropping the signal by a fixed amount once it falls below the threshold, it applies a downward ratio — every decibel below the threshold becomes two, or three, or four. Quiet gets quieter, and it does so proportionally instead of all at once.

The practical difference is audibility. A gate makes a decision, and decisions can be heard. An expander makes a slope, and slopes are much harder to notice. For anything sustained — a vocal, an acoustic guitar, a room mic — an expander with a modest ratio removes most of the between-phrase noise while leaving no seam where the processing starts. Most modern gates are expanders with the ratio set high, which is why the two controls so often live in the same window.

The general rule: gates for percussive sources with real silence between hits, expanders for everything that decays.

De-essers: Turning It Down Only When It Is Bright

Sibilance is the "s", "sh" and "t" energy in a vocal, and it is the same performance you liked, so you cannot remove it without cost. It lives roughly between 2 and 12kHz, and the part that actually offends usually sits between 5 and 8kHz — nearer 4 to 5 on a darker male voice, nearer 7 to 8 on a bright voice through a large-diaphragm condenser. Find yours by ear before you set anything, because a de-esser aimed at the wrong band either does nothing or gives you a lisp.

A static EQ cut is the wrong tool and it is worth being clear about why. The frequencies that carry sibilance also carry the vocal's air and articulation, and the singer is only sibilant for a few dozen milliseconds at a time. Dip 6kHz for the whole song, as the EQ lesson would let you, and you have paid for four seconds of harshness with three minutes of dullness.

A de-esser is a compressor whose detector has been filtered. It listens only to the sibilant band, so it does nothing at all while the singer is on a vowel, and reduces gain the instant an "s" crosses the threshold. The trade is entirely in what it reduces:

A split-band de-esser attenuates only the sibilant band and leaves everything below it alone. Precise, and the source of the classic de-essing artifact — pull that band hard and the consonant does not get quieter so much as it turns into a lisp, because you have removed the part of the sound that makes an "s" an "s" rather than a "th".

A wideband de-esser turns the whole signal down for the duration of the consonant, the way a person riding a fader would. Less surgical, but it preserves the internal balance of the consonant, so what you hear is a quieter "s" rather than a damaged one. It is the more forgiving default, and a common working approach is a gentle wideband pass to take the edges off, with a narrow split-band stage afterwards for the few syllables that still stand out.

Position matters. Compression raises quiet material towards the loud material, and sibilance is a peak that most compressors are too slow to catch — so a compressor reliably makes a vocal more sibilant than it was. De-essing after the compressor addresses what you actually have; de-essing before it addresses what you started with and then hands the result to the thing that undoes your work.

Sidechain: the Detector Fed From Somewhere Else

Everything above uses an internal sidechain — the detector listens to the same signal being processed, possibly filtered. Route a different signal into the detector and the same device becomes something else again: the level of one track now controls the gain of another.

The plainest use is ducking. A voice into the detector, a music bed into the audio path, and the music steps out of the way whenever someone speaks and comes back when they stop. Radio has done this for a century; it is also how a podcast intro works, and how a club record makes room for a vocal.

The version most people meet first is kick and bass. Both occupy the bottom of the spectrum and both want to be the loudest thing there, so feeding the kick into a compressor on the bass makes the bass step aside for each kick and return between them. Set gently it is a clarity tool and nobody notices it. Set hard and fast it becomes the audible pumping that defines several dance genres, and at that point it is not a fix, it is the sound.

The third use is the least visible and possibly the most useful: filtering the detector of an ordinary compressor. A bus or mix compressor sees the whole spectrum, and low frequencies carry most of the energy, so a kick drum can pull the entire mix down on every beat while nothing else has changed. High-pass the detector at 80 or 100Hz and the compressor stops reacting to the bass while still processing it. The audio path is untouched; only the decision changed. This is the same distinction every device here turns on, and it explains why a mix bus compressor with a sidechain filter sounds "less squashed" than one without at the same gain reduction.

What Changes on Stage

Gates earn their place live, on drums, for the same reason they do in the studio and more urgently: every open channel on a loud stage is also a path to feedback, and each doubling of open microphones costs you 3dB of available gain, as the feedback lesson explains. Gating the toms genuinely buys headroom.

Vocal gates live are a different matter, and the usual advice is to be careful. A stage is loud, the gap between a quiet phrase and the wedge behind the singer is small, and a gate that clips the front of a word is worse than the bleed it was fixing. An expander with a modest ratio is the safer version of the same idea.

De-essers appear on live vocal channels reasonably often and rarely need to work as hard, because a stage vocal is not being scrutinised in a quiet room. Sidechain ducking barely appears at all outside of theatre and corporate work, where a presenter's microphone ducking playback is standard.

What This Means for You

• Every device here is a compressor with a different detector. Ask what it is listening to, and what it does when the threshold is crossed, and the controls stop being arbitrary.
• Set a gate's threshold in the gap between source and bleed, not by the source. If there is no gap, the microphone choice and position created the problem and only they can solve it.
• Use a range of 15 to 20dB rather than closing the gate completely. You want the bleed gone, not the instrument teleporting in and out.
• Reach for an expander on anything that sustains. A slope is much harder to hear than a decision, and most sources are not percussive.
• Find the sibilant band by ear before setting a de-esser. It moves between about 4 and 8kHz depending on the voice and the microphone.
• De-ess after the compressor, not before. Compression makes sibilance worse, so treating it first means treating the wrong signal.
• Try wideband before split-band. Attenuating the whole consonant sounds like a quieter "s"; attenuating only its top sounds like a lisp.
• High-pass the sidechain of any bus compressor. Otherwise the kick drum decides the gain of your entire mix eight times a bar.
• Live, gate the toms and go gently on the vocals. Fewer open microphones is real headroom; a gate that bites the front of a word is not worth it.