What is gain reduction?

Gain reduction in plain English

A compressor only ever does one thing to the level: it turns it down. Gain reduction is the name for how far down, measured in decibels, at this exact moment. When the meter reads 0 dB the compressor is passing the sound through untouched, and when it reads 6 dB the sound coming out is 6 dB quieter than the sound going in.

It changes all the time. A compressor only turns down the parts that go over its threshold, so on a drum loop the reduction jumps up on every loud hit and falls back in the gaps. On a vocal it rises on the loud words and relaxes on the quiet ones. All that movement is why the gain reduction meter is the most useful readout a compressor has: the knobs say what you asked for, and the meter shows what the compressor is doing with this particular sound.

The rest of this page is about reading it, both the number and the way it moves. The compression lesson covers what each knob does; here the sliders in the lab are only there to give the gain reduction something worth looking at.

How to read a gain reduction meter

Every gain reduction meter starts at zero and moves as the compressor works, but they're drawn in a few different ways.

Bars that hang down from zero

Many plug-ins draw gain reduction as a bar that grows downwards from a 0 dB line at the top, the opposite way to a level meter, so a longer bar means more reduction. The amber lane along the bottom of the lab's Whole loop picture works like this, hanging down from its top edge. Ableton Live's Compressor and Logic's Compressor both have a gain reduction meter, and it's the first thing to find when you open either one.

Numbers, positive or negative

Some meters print the amount taken off as a positive number, so 4.0 means 4 dB down. Others print the change in level, so the same moment reads −4.0. They mean the same thing. The Turning down readout in the lab prints the amount taken off.

Needles

Hardware compressors, and plug-ins that copy them, often use a needle. On a Universal Audio 1176 or a Teletronix LA-2A you choose what the meter shows with a switch (a row of buttons on the 1176), and in its gain reduction position the needle rests at 0 and swings left as the compressor turns the sound down. A needle is slow. A standard VU meter is built to take about 300 milliseconds to reach its reading, so on a snare hit that's only loud for a few milliseconds the needle never swings as far as the real reduction.

Scrolling lanes

The most informative kind draws gain reduction over time, like a waveform. FabFilter Pro-C 2 scrolls the level and the gain reduction across its display, and the Whole loop picture in the lab does the same from start to finish: the pale shape is the original, the pink shape is the output, and the amber lane along the bottom is the gain reduction.

A lane shows what a bouncing meter can't, which is the shape of each dip. You can see how fast it falls, how deep it goes and how long it takes to come back, and those are the attack, the amount and the release.

How much gain reduction a setting gives

You can work out the most a compressor will take off a sound from the threshold and the ratio. With a hard knee, take how far the sound goes over the threshold and multiply it by one minus one over the ratio.

Say a snare peaks at −6 dBFS and the threshold is at −20 dB. The snare is 14 dB over. At 4:1 the compressor lets a quarter of that through, so the snare comes out 3.5 dB over the threshold, at −16.5 dB, and the meter reads 10.5 dB of gain reduction. At 2:1 the same hit gets 7 dB taken off, and at 10:1 it's 12.6 dB.

Those figures are a ceiling. They assume the sound stays over the threshold long enough for the compressor to clamp down fully, and a short drum hit often doesn't. With a slow attack the hit is already fading by the time the gain has dropped, so the meter shows less than the table says. In the lab, raising the attack from 10 ms to 30 ms takes the deepest reading on the drum loop from about 8 dB to under 5, with the threshold and ratio left exactly where they were.

The ratio also does less than you'd expect once the sound is only a few decibels over. In the top row, going from 2:1 to 10:1 adds just over a decibel. That's why lowering the threshold usually moves the meter much more than raising the ratio does, and the threshold and ratio lesson goes further into how the two trade off.

What 3, 6 and 10 dB of gain reduction sound like

Decibels are a ratio, so a few of them go a long way. Taking 3 dB off brings the signal down to about 71% of its height on a waveform. Taking 6 dB off halves it. At 10 dB it's down to about a third, which sounds roughly half as loud, and at 20 dB it's a tenth.

So 3 dB on the peaks of a vocal is easy to miss when you listen, and that's usually what you want. At 6 dB the loud words are clearly being held back, though someone who doesn't know the compressor is there won't pick it out. At 10 dB or more on every hit the compressor has changed the shape of the sound, and you'll hear the drums go flatter or the room and the tails of notes come up.

These are amounts on the loudest moments. A meter that peaks at 6 dB may spend most of its time at 1 or 2, which sounds very different from one sitting at 6 the whole time.

How much gain reduction should you use?

Judge it by ear with the levels matched. You still need somewhere to start, so here are amounts on the loudest moments that tend to do the job without drawing attention to themselves.

A parallel compressor is the exception. On a parallel bus the compressed copy is mixed in under the original, so 15 or 20 dB of gain reduction is normal there, and the parallel compression lesson explains why it can still sound natural.

Reading the movement as well as the number

Two compressors can both peak at 6 dB and sound nothing alike. What separates them is how the meter moves between peaks.

A dip on each hit, then back to zero

This is what most compression on drums and vocals should look like. The meter dips with the loud moments and returns to zero, or close to it, in the gaps. The compressor is working with the rhythm, so the quiet notes keep their level and only the loud ones are held back.

Never back to zero

If the meter hovers at 4 or 5 dB and never returns, either the threshold is so low that everything is over it or the release is longer than the gap between hits. The compressor is turning everything down all the time, and the difference between loud and quiet shrinks with it. In the lab, a 500 ms release on the drum loop keeps the amber lane between about 4 and 8 dB down for the whole loop.

Sometimes that's the goal. A slow compressor holding a vocal 2 or 3 dB down all the way through evens out the whole performance. On drums it usually flattens the groove, so shorten the release or raise the threshold until the meter starts coming back.

Snapping back after every hit

A very short release makes the meter spring back almost instantly. On drums that keeps the hits lively, but when the gain comes back up in time with the beat you can hear the sound swelling between hits, which is called pumping. On bass a release under about 20 ms can start following the waveform itself, and the meter chatters instead of moving smoothly; you'll hear that as a buzz.

A dip that starts late

On a lane display you can see how quickly the gain falls compared with the front of the hit. With a slow attack the gain comes down gradually, so the deepest part of the dip arrives after the transient and the first few milliseconds of each hit get through at nearly full level, which is how a compressor adds punch. A dip that starts right on the front of the hit means the attack is fast enough to catch the transient and soften it. The attack and release lesson covers how to choose between the two.

A meter that sits still

A meter parked at one reading, hardly moving, means the sound is over the threshold all the time and its level barely changes. A sustained pad or a heavily compressed sample can do this. The compressor has turned into a fader stuck a few decibels down, which turning the track down would do just as well, so raise the threshold until the meter starts reacting to the loud moments.

Peak and average gain reduction, and makeup gain

The biggest number the meter shows is the peak reduction, on the loudest hit. What you hear as a change in loudness depends more on the average, which is nearly always smaller. In the lab, at a −20 dB threshold and 4:1, the deepest dip on the drum loop is about 8.4 dB, yet the loop only gets about 4.6 dB quieter overall.

So setting makeup gain to the peak reading makes the compressed version too loud. Start well below it, then match by ear with the compressor switched on and off. In the lab, 4.5 dB of makeup matches the original almost exactly, even though the meter peaks at nearly twice that. Keep an eye on the output meter while you do it. With a 10 ms attack the front of each hit gets through before the gain comes down, so at matched loudness the loop's highest peak sits about 3 dB above the original's. That's fine as long as there's room for them, or a limiter later in the chain to catch them. The makeup gain lesson covers how to match levels in your DAW.

Gain reduction on limiters and in a chain

A limiter has a gain reduction meter too, and it reads the same way. The difference is that a limiter is set so nothing gets past its ceiling, so its meter only moves on the loudest peaks. On a finished master, a couple of decibels flickering on the biggest drum hits is a light touch. A limiter holding 5 or 6 dB down through a whole chorus is working hard enough that the drums lose their front. The limiting lesson explains how limiters use lookahead to catch peaks without distorting them.

When two compressors are in a row, their gain reduction adds up. One taking 3 dB off the loudest words of a vocal and a second taking another 3 dB gives 6 dB of control in total. Splitting it like that lets each compressor do a different job, with a fast one catching peaks and a slower one evening out the phrases, and it usually sounds less obvious than one compressor doing all 6. The parallel vs serial compression lesson compares the two approaches.

Why the same settings read differently

Put the same vocal through two different compressors at the same threshold, ratio, attack and release and the meters won't agree.

The detector is the part of the compressor that measures the level. Some respond to peaks, sample by sample, and others to a short average, the RMS level, which sits lower than the peaks. So at the same setting a peak detector shows more reduction. The compressor in this lab uses one, and the peak vs RMS compression lesson shows the difference.

Attack and release are shaped differently from one design to the next, so a 10 ms attack on one compressor won't match 10 ms on another. The meter has its own speed as well: a fast digital meter shows a 2 ms dip that a needle can't.

Then there's the sidechain. Many compressors let you filter what the detector hears, usually with a high-pass filter so heavy bass doesn't trigger it. Switch that filter on and the meter drops on bass-heavy material, because the detector no longer hears the low end that was pushing the level over the threshold. The sidechain compression lesson covers what else the sidechain can do.

So use the meter to compare settings within one compressor, and don't expect 5 dB on one plug-in to sound like 5 dB on another.

How to set a compressor by watching the meter

1. Loop the loudest partPlay the chorus of a vocal or the busiest bar of the drums, so the meter is reacting to the moments that need control.
2. Lower the threshold until the meter movesBring it down slowly. The first flickers show you where the threshold meets the loudest peaks.
3. Keep going until the peaks read the amount you wantFor a vocal or a drum, 3 to 6 dB on the loudest moments is a good start. The table above has amounts for other sources.
4. Watch it between the peaksIf the meter never gets back towards zero, shorten the release or raise the threshold. If it chatters, lengthen the release.
5. Match the level with makeup gainBring the level back until the compressor sounds equally loud on and off. That usually takes less than the peak reading.
6. Stop looking and listenSwitch the compressor off and on. If you can't hear what it improved, back the threshold off or take it out.

Common gain reduction mistakes

Aiming for a number

Reading that vocals get 6 dB and then setting every vocal to 6 dB ignores how the sound arrived. A singer who was already even needs less, and a sample that was compressed when it was made may need none.

Watching only the deepest dip

The peak reading on the loudest snare says little about the rest of the song. Watch the meter through a whole section, because what it does on an ordinary moment is what you'll hear most.

Comparing with the levels mismatched

Too much makeup gain makes the compressed version louder, and louder sounds better in a quick comparison. Match the level first, with the Original and Compressed switch in the lab or careful makeup gain in your DAW, and only then decide whether the gain reduction is helping.

Missing a meter that never comes back

A drum bus compressor with its meter parked at 4 dB is turning the whole bus down all the time, and the groove goes with it. It's easy to miss because nothing about the meter looks alarming. If it never touches zero, check the release and the threshold.

Trusting the meter over your ears

A needle meter under-reads fast hits, and a sidechain filter changes what the meter shows. If the meter says 2 dB and the snare sounds squashed, believe the snare.

Train your ears to hear gain reduction

The quiz under the lab covers the readouts and the sliders, matches the levels and asks you to name what you're hearing: a fast or slow attack, or a ratio of 2:1, 4:1 or 10:1. It's the same skill as reading the meter, done without it. Once you can hear a slow attack letting the crack of a hit through, you'll know what that slow, late dip on a lane display sounds like. Eight rounds take a few minutes, and your best score is saved in your browser.

Questions people ask

Keep going