Upward compression in plain English
An ordinary compressor works from the top. It waits for a sound to get louder than its threshold and then turns it down. An upward compressor works from the bottom: it waits for a sound to get quieter than its threshold and turns it up. Both squeeze the dynamic range, the gap between the loudest and quietest moments, but they move different parts of the sound to do it.
That difference decides what you hear. When an ordinary compressor brings a snare hit down, the crack of the hit changes shape, and you win the level back with makeup gain. An upward compressor leaves the snare hit exactly where it was and lifts the ghost notes and the ring of the drum around it. The transients keep their height, and everything quieter moves up to meet them.
To keep the two apart, engineers call the ordinary kind downward compression. When a lesson elsewhere on this site says compression with no other word in front of it, it means downward; the compression lesson covers that from the start.
How does upward compression work?
Like any compressor, an upward compressor measures the level of the sound thousands of times a second and works out how much gain to apply from three settings: the threshold, the ratio and the range.
Say the threshold is −30 dB and the ratio is 3:1. A note at −48 dB is 18 dB under the threshold. At 3:1, those 18 dB come out as 6 dB, so the note leaves at −36 dB: it has been turned up by 12 dB. A note at −36 dB is 6 dB under, comes out 2 dB under at −32 dB, and gets 4 dB of lift. Anything at −30 dB or louder gets nothing, which is why the peaks don't move.
Follow that maths all the way down and silence would be turned up without limit, along with every scrap of hiss on the recording. The range control stops it. With 12 dB of range, nothing ever gets more than 12 dB of lift, and at 3:1 that limit arrives 18 dB under the threshold. Everything quieter than −48 dB gets exactly 12 dB more and keeps its distance from everything else down there.
The transfer curve in the lab draws the same thing. Input level runs along the bottom and output level up the side, and the dashed diagonal is where nothing changes. In upward mode the line rises above that diagonal under the threshold, peels further away until the range runs out, then runs parallel to it again.
Reading the lab
The big screen shows the whole loop. The pale shape is the original and the pink shape is what comes out, so wherever the pink is bigger than the pale, the sound is being turned up. The teal lane underneath shows the lift at each moment: it climbs in the gaps between hits and drops back to the floor on every hit. Hover over any point, or press and drag on a phone, for the exact figure.
The middle screen, the level spread, is the one to watch on this page. The lab cuts the loop into 30 ms slices and stacks them by level, so the pale bars show where the original spends its time and the pink bars show where the processed version does. Upward compression pushes the left side of the pile, the quiet slices, towards the threshold line. The right side stays put.
The panel on the right describes your current settings with numbers measured from the loop: how much the quietest tenth has come up, how far the loudest peak has moved and how big the gap between loud and quiet is now. Press A/B to switch the compressor off. Upward compression always makes a sound louder, so the lab turns the original up to the same loudness when you do, and the only thing left to compare is the shape.
The Downward card at the top switches the same controls to an ordinary compressor, which is the quickest way to hear the difference between the two. You can also upload your own file or record eight seconds with your microphone, and it loops through the lab in place of ours without leaving your device.
The controls
Upward compressors use the same names as ordinary ones, but threshold, attack and release each mean something slightly different when the compressor is turning sounds up. Range is the one ordinary compressors usually don't have.
Threshold
The threshold is the level the compressor lifts things towards. Anything quieter than it gets turned up; anything louder passes through untouched. Raise the threshold and more of the sound falls under it, so more of it gets lifted. Set it just under the level of the main hits or the main words and the compressor works on everything between them.
Ratio
The ratio sets how strongly the quiet parts get pulled towards the threshold. At 2:1, a sound 12 dB under comes out 6 dB under. At 4:1 the same sound comes out 3 dB under, so it has been lifted by 9 dB. A ratio of 1:1 does nothing at all.
Ratios of 1.5:1 to 2:1 are gentle enough for vocals and whole mixes. Past 4:1 the quiet parts end up very close to the threshold, which sounds dense and flat on acoustic sources but can be exactly what you want on a bass or a synth in electronic music.
Range
Range is the ceiling on the lift, in decibels. Below a certain level the quiet parts of a recording stop being music and start being hiss, hum, headphone bleed and the air conditioning, and range is what stops the compressor dragging those up as far as everything else. Six dB of range is subtle. Twelve is clear, and 20 dB or more turns the faintest decays into a wall of sound.
Some upward compressors don't have a range control at all. On those, the threshold and ratio between them decide how far the quietest sounds get lifted, so keep an ear on the noise as you raise either one.
Attack
Attack is how quickly the gain comes down when the level rises. On an upward compressor, that means how quickly the lift backs off when a loud hit arrives after a quiet passage. A fast attack of 1 or 2 ms lets go almost as the hit starts, so the hit comes through at its original level.
A slow attack keeps the lift on for the first part of the hit, which turns the front of the hit up by however much lift was there a moment before. In the lab that pushes the peak above the original. Sometimes that extra snap is welcome, but it can also push you into clipping on a track that was already close to 0 dBFS.
Release
Release is how quickly the gain goes back up once the level falls. On an upward compressor, that's how quickly the lift returns after each hit, which decides when the tail of a drum or the end of a word comes up behind it. A release of 50 to 150 ms brings the tail up almost as soon as the hit has gone.
A long release, 400 ms or more, lifts the tail slowly, so you hear the space after each loud hit swell up. It's the upward version of pumping, and on reverb and room mics it can sound like the room is breathing in after every hit.
Mix
The mix control blends the processed sound with the original. At 100% you hear only the upward compressor. At 50% the quiet parts come up by less, and the result sounds more like the original with some of the detail brought forward, which is a good way to use heavy settings without the flat, airless sound they have on their own.
Upward vs downward compression
On paper the two can give the same result. Take a vocal whose loud words peak at −6 dB and whose quiet words sit at −36 dB, a 30 dB gap. A downward compressor with a −30 dB threshold at 3:1 brings the loud words down to −22 dB, which leaves a 14 dB gap. An upward compressor with a −12 dB threshold at 3:1 and 20 dB of range lifts the quiet words to −20 dB, and the gap is 14 dB again.
The difference is in what moves. Downward compression changes the loud moments, which is where the transients are, so its attack and release reshape every hit. Upward compression changes the quiet moments, which on most recordings means the tails of notes and whatever noise sits under them.
| Downward | Upward | |
|---|---|---|
| Works on | Sound over the threshold | Sound under the threshold |
| Peaks | Turned down | Left where they were |
| Quiet parts | Left alone until you add makeup gain | Turned up |
| Transients | Reshaped by attack and release | Kept, unless the attack is slow |
| Noise floor | Raised by the makeup gain | Raised by up to the range |
| Overall level | Quieter, then made up | Louder |
The noise floor row catches people out. Downward compression raises the noise as well, as soon as you add makeup gain, because makeup turns everything up by the same amount. Upward compression raises the noise by up to the range, and only where the sound is quiet. The upward vs downward compression comparison goes through the two side by side.
Upward compression vs parallel compression
Parallel compression gets to almost the same place by another route. You crush a copy of the sound with a fast, heavy downward compressor, bring its level back up with makeup gain and mix it under the original. Where the sound is loud, the copy has been turned down hard, so it adds very little. In the quiet parts it's barely compressed and still carries all of that makeup gain, so it adds a lot. The blend lifts the quiet parts and leaves the peaks close to where they were, which is upward compression in effect.
The two don't sound identical. A parallel blend carries the character of whatever compressor crushed the copy. It has no range control either: the most it can lift anything is set by the copy's makeup gain and fader, and the noise floor gets that full lift. A real upward compressor gives you a more exact version of the same move. If your DAW's stock compressor only works downward, though, parallel compression is how you'll get this sound.
Plug-ins that do upward compression
Most single-band compressors, including the stock ones in most DAWs, only turn sounds down. Upward compression mostly turns up in multiband compressors and in a few dedicated tools.
Xfer's OTT is free and splits the sound into three bands, each with an upward and a downward compressor running at once; the Upwd knob sets how much of the upward half you get. OTT was based on a preset of the same name for Ableton Live's Multiband Dynamics, which can do upward compression on each of its bands too. The OTT lesson covers that sound, and the multiband compression lesson covers splitting a sound into bands.
Waves MV2 has a Low Level and a High Level control beside its output fader. Low Level is an upward compressor with the details hidden: push it up and the quiet parts follow.
Good uses for upward compression
Vocals and speech
A singer or speaker who drops their voice at the end of every line is a natural fit. An upward compressor lifts the quiet syllables without touching the loud ones, so the strong words keep their shape and you can still hear the soft ones. Watch the breaths: they can sit close to the level of the softest words, and then they come up with them.
Acoustic instruments
On acoustic guitar and piano, the decay of each note carries a lot of the tone. Lifting it gently makes a part sound fuller and longer without squashing the pick or hammer attack.
Drum rooms and reverb
Room mics and reverb returns are mostly quiet tails, so an upward compressor makes the room sound bigger and closer. Pick Drums in the lab and listen to what happens between the hits at 3:1.
Bass and sound design
In electronic music, upward compression on bass, leads and pads brings up the texture and the tail of every note. OTT on a synth or a bus is a common way to get the dense, bright sound heard in a lot of dubstep and future bass, and it's easy to overdo; a 30 to 50% mix is usually plenty.
Whole mixes
A gentle upward compressor across a mix, 1.5:1 with 2 to 4 dB of range, lifts the quiet passages of a song and leaves the loud ones alone, so you get more level in the verses without a downward compressor reacting to every chorus.
How to set up upward compression, step by step
- Start with the ratio at 1:1 and 12 dB of rangeThe compressor does nothing yet, which gives you a clean starting point to listen from.
- Set the threshold just under the loud momentsFind the level of the main hits or words and put the threshold a few decibels below it, so they pass through untouched.
- Raise the ratio until the quiet parts come forward2:1 is a gentle start. Stop as soon as the detail you want to hear is audible.
- Set the range by listening to the quietest soundsTurn it up until the tails and soft notes sit where you want them, and back it off if hiss or breaths start to arrive with them.
- Keep the attack fast unless you want extra snap1 to 5 ms keeps the peaks where they were. Slower values turn the front of each hit up.
- Set the release by listening to the tailsShort values bring them up evenly; long values make them swell.
- Compare at matched loudnessTurn the output down until the processed sound is as loud as the original, then switch the compressor in and out.
Starting settings for common sounds
These are places to begin, chosen so the compressor is clearly doing something. The right threshold depends on how loud your recording is, so it's given relative to the loud parts.
| Source | Threshold | Ratio | Range | Attack | Release |
|---|---|---|---|---|---|
| Lead vocal | 10 – 15 dB under the loud words | 2:1 | 6 – 9 dB | 1 – 5 ms | 100 – 200 ms |
| Acoustic guitar | 6 – 10 dB under the strums | 2:1 | 6 dB | 5 ms | 150 ms |
| Drum room | Just under the hits | 3:1 | 12 dB | 1 ms | 60 – 100 ms |
| Reverb return | Just under the loudest part of the tail | 2:1 – 3:1 | 6 – 12 dB | 2 ms | 200 ms |
| Synth bass | Just under the note starts | 3:1 – 4:1 | 12 – 18 dB | 1 – 2 ms | 80 – 120 ms |
| Mix bus | Just under the loud sections | 1.5:1 | 2 – 4 dB | 5 ms | 200 ms |
Common upward compression mistakes
Bringing up the noise with the detail
The quietest thing on a recording is usually noise, so it's the first thing an upward compressor finds. Keep the range as low as the job allows, and if a track has hiss or bleed, put a gate or an expander in front of the upward compressor to push the noise back down first.
Judging it by loudness
An upward compressor always makes a sound louder, and a louder version nearly always wins a quick comparison. Match the levels before you decide; in the lab, A/B does it for you.
Losing the quiet moments altogether
Pauses and the space between hits are part of how music moves. With a high ratio and a lot of range, all of that gets filled in, and a part that used to breathe ends up at one level all the way through. If the track sounds tiring after a minute, back the ratio or the range off.
A slow attack on a track near 0 dBFS
A slow attack turns the start of each hit up, so the peaks get louder. On a track that was already close to full scale that means clipping, and the limiter later in the chain will have more work to do.
Training your ears to hear upward compression
Upward compression is easiest to hear in the gaps, so that's where to listen. The quiz under the lab plays hidden settings at matched loudness and asks whether the upward compressor is on, then whether its range is small or large. Your best score is saved in your browser, so you can try it on a different sound and see how it compares.
Questions people ask
Is upward compression the same as parallel compression?
They do a similar job by different means. An upward compressor turns the quiet parts up directly. Parallel compression mixes a heavily compressed copy under the original, and because the copy is loudest where the original is quiet, the blend lifts the quiet parts much more than the peaks. On a meter the result looks a lot like upward compression.
Does upward compression make a track louder?
Yes. The quiet parts come up while the peaks stay where they were, so the average level rises without the peak level changing. That's also why it's easy to fool yourself: compare it with the original at matched loudness before you decide it helped.
Is OTT upward compression?
Partly. Xfer's OTT is a three-band compressor that does upward and downward compression at the same time, and its Upwd knob sets how much of the upward half you get. With Downwd turned to zero it's a multiband upward compressor.
Does upward compression bring up noise?
It does, because hiss and room noise are some of the quietest things in a recording. The range control limits how far anything can be lifted. If the noise still comes up, clean it with a gate or an expander before the upward compressor, or set the threshold lower.
What's the difference between upward compression and upward expansion?
Upward compression turns quiet sounds up, which shrinks the gap between loud and quiet. Upward expansion turns loud sounds up, which widens it. With downward compression and downward expansion that makes four combinations, and multiband plug-ins such as Ableton Live's Multiband Dynamics can do all four.
Can I do upward compression with a normal compressor?
Not directly, because a single-band compressor only turns sounds down. Put it on a send, or set its mix knob to around 30 to 50%, with a low threshold and plenty of makeup gain, and the parallel blend will behave much like an upward compressor.