how compressors work: attack, release, ratio, threshold
how the four core compressor controls actually work, how they interact, and what happens when you get them wrong. a technical guide for music producers.
the four knobs that control everything
you open a compressor plugin and see four controls: threshold, ratio, attack, and release. every compressor in existence, from a 1960s Teletronix LA-2A to a modern spectral processor, is built on these four parameters. attack is how fast you brake. release is how fast you get back to cruising speed. learn what they do and how they interact, and you can operate any compressor on the market.
the controls split into two categories. threshold and ratio determine how much compression happens. attack and release determine when it happens. getting the “how much” right with the wrong “when” produces results that sound mechanical or unmusical. getting the “when” right with the wrong “how much” either does nothing or crushes the signal. both categories matter equally.
key takeaway
threshold and ratio decide the amount. attack and release decide the character. most compression mistakes come from setting the amount first and hoping the character sorts itself out. start with the timing, then dial in the depth.
threshold: where compression begins
the threshold is the level in decibels where compression starts acting. signals below the threshold pass through untouched. signals above the threshold get reduced.
set the threshold at -18 dBFS. a vocal phrase that peaks at -12 dBFS exceeds the threshold by 6 dB. that 6 dB overshoot is what the compressor acts on. a quieter phrase that peaks at -20 dBFS is below the threshold and passes through unchanged.
the threshold is the single most important control for determining how much of the signal gets compressed. a high threshold (close to 0 dBFS) means only the loudest peaks trigger compression. a low threshold (far below 0 dBFS) means almost the entire signal is being compressed continuously.
practical threshold settings:
- light compression (peaks only): set the threshold so gain reduction happens only on the loudest moments. 1 to 3 dB of gain reduction on peaks. the compressor is mostly idle
- moderate compression: set the threshold so the compressor engages on most phrases. 3 to 6 dB of gain reduction on the loudest moments, occasional gain reduction on medium moments
- heavy compression: set the threshold low enough that the compressor is working on nearly every note. 6 to 12 dB of gain reduction on peaks. the compressor is rarely idle
threshold in hardware vs software
in hardware compressors, the threshold is sometimes labeled differently. the LA-2A has a “peak reduction” knob instead of a threshold. the 1176 has an “input” knob that drives the signal harder into a fixed threshold. the concept is the same: you are controlling how much of the signal exceeds the compression onset point. software compressors typically use an explicit threshold control in dBFS.
ratio: how much compression
the ratio determines how aggressively the compressor reduces the signal above the threshold. it is expressed as input-to-output: at 4:1, every 4 dB of signal above the threshold becomes 1 dB at the output. the compressor has reduced 3 dB of the 4 dB overshoot.
at 1:1, no compression happens. the input equals the output. at infinity:1, any signal above the threshold is clamped to the threshold level. this is a limiter.
the math is simple. for a signal that exceeds the threshold by X dB at a ratio of R:1, the output above the threshold is X / R dB. the gain reduction is X - (X / R) dB.[^1]
examples with a -18 dBFS threshold and a -12 dBFS peak (6 dB above threshold):
- 2:1 ratio: output is 6/2 = 3 dB above threshold = -15 dBFS. gain reduction: 3 dB
- 4:1 ratio: output is 6/4 = 1.5 dB above threshold = -16.5 dBFS. gain reduction: 4.5 dB
- 8:1 ratio: output is 6/8 = 0.75 dB above threshold = -17.25 dBFS. gain reduction: 5.25 dB
- 20:1 ratio: output is 6/20 = 0.3 dB above threshold = -17.7 dBFS. gain reduction: 5.7 dB
notice how the returns diminish. the jump from 2:1 to 4:1 adds 1.5 dB of gain reduction. the jump from 8:1 to 20:1 adds only 0.45 dB. above 10:1, increasing the ratio barely changes the result. this is why ratios above 10:1 are considered limiting: the behavior is effectively the same as infinity:1.
the knee
the knee controls the transition between uncompressed and compressed signal around the threshold. a hard knee means the compression ratio applies abruptly the instant the signal crosses the threshold. a soft knee means the compression gradually increases as the signal approaches and exceeds the threshold.
a soft knee typically starts applying gentle compression 3 to 6 dB below the threshold and reaches the full ratio 3 to 6 dB above it. the result sounds more natural because the transition into compression is gradual rather than sudden.
hard knee compression sounds more precise and controlled. soft knee compression sounds more transparent and musical. for most mixing applications, soft knee is the better default. hard knee is useful when you want obviously aggressive compression or when using a compressor as a limiter.
key takeaway
the ratio determines how much the signal above the threshold is reduced. the knee determines how abruptly that reduction begins. a 4:1 ratio with a soft knee sounds gentler than a 4:1 ratio with a hard knee, even though the gain reduction at high levels is the same.
attack: how fast the compressor responds
the attack time controls how quickly the compressor reaches its full gain reduction after the signal exceeds the threshold. it is measured in milliseconds.
a fast attack (0.1 to 1 ms) means the compressor clamps down almost instantly. the transient, the initial spike of a drum hit or vocal consonant, gets compressed along with the sustain. the result is a smoother, more controlled sound, but with less punch.
a slow attack (10 to 30 ms) means the compressor lets the initial transient pass through before engaging. the first few milliseconds of a drum hit or plucked string arrive at full volume. then the compressor kicks in and controls the sustain. the result is a punchier sound with preserved transient impact.
the attack time is the most important control for shaping the character of your compression. two compressors with identical threshold, ratio, and release settings will sound completely different if one has a 0.5 ms attack and the other has a 20 ms attack.
practical attack settings:
- very fast (0.1 to 1 ms): catches everything. reduces transient impact. good for controlling harsh peaks, de-essing, limiting
- medium (1 to 10 ms): catches most of the transient but lets the very first click through. the most versatile range for vocals and instruments
- slow (10 to 30 ms): lets the full transient through. controls only the sustain. the classic setting for punchy drums and bass
- very slow (30+ ms): the compressor barely catches the transient. mostly smoothing the overall dynamics. useful for mix bus and mastering
attack time and low frequencies
fast attack times can cause distortion on low-frequency signals. a 100 Hz bass note has a period of 10 ms. an attack time of 1 ms means the compressor changes its gain several times within a single wave cycle, effectively altering the shape of the waveform. this creates harmonic distortion. it is not always a problem (you can use this deliberately for bass character), but if your bass sounds gritty or fuzzy with fast compression, try a slower attack.[^2]
release: how fast the compressor lets go
the release time controls how quickly the compressor returns to zero gain reduction after the signal drops below the threshold. it is also measured in milliseconds, though typically in a range 10 to 100 times longer than attack times.
a fast release (10 to 50 ms) means the compressor recovers quickly. it grabs the peak, reduces it, and lets go before the next peak arrives. this preserves the natural dynamic contour of the performance but can create audible “pumping” if the gain reduction is heavy: you hear the compressor gripping and releasing on every beat.
a slow release (100 to 500 ms) means the compressor holds its gain reduction longer. it smooths out the dynamic contour, creating a more consistent level. the risk is that the compressor is still holding gain reduction when a quiet passage arrives, reducing the level of content that did not need compression.
the release time determines how the compressor breathes with the music. get it right and the compression disappears into the performance. get it wrong and you hear the compressor as a separate element, pumping and releasing in a way that fights the natural rhythm.
program-dependent release
many modern compressors offer “auto” or “program-dependent” release. instead of a fixed release time, the compressor adjusts its release based on the incoming signal. short, sharp transients get a fast release (to recover before the next hit). long, sustained signals get a slow release (to avoid pumping).
this adaptive behavior models what hardware compressors do naturally. the optical element in an LA-2A has an inherently program-dependent response: it reacts quickly to fast signals and slowly to sustained ones. program-dependent release in a plugin is a mathematical approximation of this behavior.[^3]
how the controls interact
the four controls do not operate in isolation. they form a system where each setting influences how the others behave.
threshold + ratio
together, these define the compression curve. a low threshold with a gentle ratio (2:1) might produce the same average gain reduction as a high threshold with an aggressive ratio (8:1). but the character is different. the low threshold approach compresses everything gently. the high threshold approach leaves most of the signal alone and clamps hard on the peaks.
for transparent compression, use a lower threshold with a lower ratio. for aggressive peak control, use a higher threshold with a higher ratio.
attack + release
together, these define the envelope. the combination determines whether the compressor grabs transients, controls sustain, pumps rhythmically, or smooths everything to a flat line.
common combinations:
- fast attack + fast release: tight, aggressive. grabs every transient and lets go quickly. good for parallel compression where the compressed signal is blended with the dry
- fast attack + slow release: smooth, controlled. catches transients and holds them down. the classic vocal compression setting
- slow attack + fast release: punchy. lets transients through and recovers before the next one. the classic drum compression setting
- slow attack + slow release: gentle, transparent. smooths the overall dynamics without acting on individual transients. the classic mix bus setting
all four together
the interaction of all four controls is what makes compression an art. the same 4 dB of gain reduction can sound invisible or destructive depending on how you arrive at it.
tip
here is a reliable starting workflow. set the attack to medium (5 to 10 ms). set the release to medium (50 to 100 ms). lower the threshold until you see 3 to 4 dB of gain reduction on the loudest moments. now adjust the attack: faster if you want more control, slower if you want more punch. then adjust the release: faster if the compression sounds sluggish, slower if it pumps. finally, fine-tune the ratio and threshold to get the exact depth you want. timing first, depth second.
makeup gain and auto-compensation
after compression reduces the peaks, the overall signal is quieter than it was. makeup gain adds a fixed level boost to bring the compressed signal back to its original perceived loudness.
the simplest approach: add makeup gain equal to the average gain reduction. if the compressor is typically reducing by 4 dB, add 4 dB of makeup gain. the peaks are now at roughly the same level they were before, but the quiet parts are 4 dB louder relative to them.
many compressors offer auto-makeup gain that estimates the correct amount based on the threshold and ratio settings. this is convenient but not always accurate, especially with dynamic material where the gain reduction varies widely.
more sophisticated compressors use loudness-matched auto-compensation. instead of estimating from the settings, they measure the perceived loudness (using standards like ITU-R BS.1770) of the input and output and adjust the gain to match. this is more accurate because it accounts for the actual signal behavior, not just the compressor settings.
key takeaway
always level-match before judging compression. your ears perceive louder signals as “better” regardless of whether the compression improved the sound. if the compressed signal is even 1 dB louder than the bypass, you will think the compression sounds better even when it does not. auto-compensation removes this bias.
common mistakes
over-compression
the most common mistake. the gain reduction meter is showing 8 to 12 dB on every phrase, the signal sounds flat and lifeless, and you cannot figure out why the mix has no energy. reduce the amount. 2 to 4 dB of gain reduction handles most situations.
wrong attack for the source
fast attack on drums kills the transients and makes them sound soft. slow attack on a sibilant vocal lets the harsh consonants through. match the attack to the source: fast for things that need smoothing, slow for things that need punch.
fighting the release
the compressor pumps on every beat because the release is too fast. or the compressor holds everything down because the release is too slow. the release should match the tempo and phrasing of the material. on rhythmic sources, try timing the release so the compressor recovers just before the next transient arrives.
compressing to make things louder
compression is not a volume tool. if a track needs to be louder, turn up the fader. if a track needs to sit more consistently in the mix, compress it. these are different problems with different solutions.
frequently asked questions
frequently asked questions
what does attack time do on a compressor?
attack time controls how quickly the compressor responds after the signal exceeds the threshold. a fast attack (under 1 ms) catches transients immediately, reducing punch. a slow attack (10-30 ms) lets the initial transient through before clamping down on the sustain. the attack time is the most important control for shaping how your compression sounds.
what does release time do on a compressor?
release time controls how quickly the compressor stops compressing after the signal drops below the threshold. a fast release (under 50 ms) recovers quickly, preserving dynamics but risking distortion on low frequencies. a slow release (200+ ms) creates a smoother, more consistent sound but can cause audible pumping if the gain reduction is heavy.
what compression ratio should I use for vocals?
start with 2:1 to 4:1 for most vocal applications. 2:1 is gentle, transparent compression that controls peaks without obviously changing the character. 3:1 to 4:1 is more assertive and works well for vocals that need to sit tightly in a dense mix. go above 4:1 only for aggressive styles or limiting.
what is the knee on a compressor?
the knee controls how gradually compression engages as the signal approaches the threshold. a hard knee means compression kicks in abruptly at the threshold. a soft knee means compression fades in gradually starting a few dB below the threshold. soft knee sounds more natural and transparent. hard knee sounds more aggressive and precise.
how do attack and release interact with each other?
attack and release together define the compressor envelope. fast attack plus fast release creates tight, aggressive compression that responds to every transient. slow attack plus slow release creates gentle, transparent compression that smooths the overall dynamics. fast attack plus slow release is the most common vocal setting: it catches peaks quickly but releases smoothly without pumping.
references
a note from the developer
i spent an embarrassing amount of time studying attack and release curves before i understood what they actually do to music. the math is simple. the listening is hard.
when i was building the three characters in KERN PUSH, each one needed different timing behavior. SETTLE needed slow, gentle timing that holds the mix together without you hearing it work. TIGHT needed fast, aggressive timing that snaps onto transients and lets go. DRIFT needed adaptive timing that lifts quiet detail without pumping up the noise floor. same four parameters, three completely different results.
i am a solo developer in copenhagen. if anything in this guide is unclear, or if you have a tip about attack and release settings that took you years to learn, i want to hear it. reach out at jonas@kernaudio.io. the best guides are written by the community, not one person.
try it yourself
KERN PUSH: three compression characters across 40 spectral bands. $29, no iLok, no subscription.