before the compressor: mix prep that makes compression do less work

the part of the chain producers skip

a producer opens their session. they have a vocal track that “needs compression.” they reach for a compressor, set 4:1 with -18 dB threshold, attack 5 ms, release 100 ms, and start tweaking. the vocal still sounds wrong. they swap to a different compressor. then a different preset. then a different plugin from the same company. an hour later they have learned nothing about the vocal and they are about to blame compression for not solving a problem compression cannot solve.

the problem is upstream. the vocal had a 3.4 kHz sibilance peak, a phase issue between two double tracks, and a body resonance at 280 Hz that nobody had touched. the compressor was not failing — it was being triggered by content the producer had not addressed yet. and once the compressor is triggered by the wrong material, no amount of tuning attack and release fixes the underlying problem.

this is what “the step before your compressor” actually means as a position. compression cleans up dynamics. it does not clean up resonance, phase, stereo image, or harmonic content. those are different jobs that have to happen before compression, or compression inherits them. (you have been here. you know the move.)

i make KERN PUSH, a spectral compressor positioned as “the step before your compressor” — and the position is meant literally. PUSH is built to do what most producers want compression to do (control dynamics across frequency, recover detail) without inheriting the problems mix prep was supposed to handle. but PUSH still works better when the work upstream of it has been done. so does any compressor. that is what this guide is about.

the four jobs that come before compression

mix prep before compression breaks into four categories. each one solves a problem that, if left for the compressor to handle, would produce uneven or unmusical gain reduction.

job 1: resonance cleanup

the most common upstream issue is resonance — a frequency or narrow band that protrudes above its surroundings consistently across the performance. a vocal recorded in a small room has a 200 Hz boxiness. a steel-string acoustic has a body resonance around 800 Hz. a bright synth has a harshness peak at 3-4 kHz. these are real, audible problems, and a compressor cannot remove them.

what a compressor does with a resonance: every time the resonant frequency sits at the loudest moment of the signal, the broadband detector sees a peak and triggers gain reduction. the compressor reads “the signal is loud right now” and pulls the entire spectrum down 4 dB. the resonance is still there — it is just compressed against the rest of the mix and now everything else has been pulled down with it. you have not fixed the resonance; you have hidden it under a layer of broadband ducking.

the fix is to remove the resonance directly, at the source, before the compressor sees the signal. three tools handle this:

• static EQ cuts for resonances that are constant across the entire performance (a guitar body resonance, a room mode).
• dynamic EQ for resonances that are only sometimes problematic (a vocal that goes harsh on certain phrases but not others).
• a resonance suppressor like KERN SMOOTH for full-spectrum resonance work that would take many dynamic EQ bands. SMOOTH operates across 40 ERB-spaced bands and only acts on bands that are actually problematic, in real time.

(this is the part most tutorials skip. they tell you “high-pass filter and compress” — which is right but only handles 1 of the 40 bands you might need to think about.)

read taming harshness on vocals and instruments for resonance-cleanup techniques per source type.

job 2: stereo and phase

most modern compressors run in stereo, with either a linked detector (left and right share the same gain reduction) or independent per-channel detectors. either mode is sensitive to phase issues between the channels.

if a track has correlated mono-incompatible content (left and right partially out of phase, summing to less amplitude in mono than either side alone), the compressor’s detector will see a smaller peak than your ears hear. the compressor will then do less work than the signal needs — and the mix will pump unevenly between mono and stereo playback. the listener on a phone speaker hears something different from the listener on monitors.

the inverse case: a stereo source with poor correlation (left and right almost uncorrelated — a hyper-wide pad, a Haas-shifted double, an over-aggressive widener) will trigger different gain reduction on the L and R sidechain. the result is a stereo image that wobbles in time with the compression.

both cases are not compression problems. they are stereo image problems that need fixing first. tools:

• a correlation meter (CHECK is free, mint #66DFBA, designed exactly for this — you will know in 5 seconds if your stereo image has a problem).
• M/S routing on a stereo widener like KERN WIDE to manage mid and side independently.
• a basic stereo-to-mono check on every track in the session before deciding if it needs compression.

(no compressor in the world fixes a wobbly stereo image. the wobble is the problem.)

read how to widen stereo without phase issues for the full treatment.

job 3: saturation

saturation adds harmonic content. it changes the spectral balance of the signal. and depending on where it sits in the chain, it can either feed the compressor with a fatter signal or follow the compressor and add character to compressed material.

saturation before compression is the analog approach: a tape machine or a console preamp would add harmonics first, then a downstream compressor would glue the harmonically-rich signal together. on modern plugins this still works for vintage-leaning material — drums, guitars, bass, anything where you want a “single-instance” cohesion that feels like analog tape.

saturation after compression is the modern approach: compress for level first, then add harmonic colour. preserves transient definition (because compression has already shaped it), gives you precise control over how much harmonic content sits on the final result. better for clinical pop and hip-hop where transparent compression matters more than analog character.

both are right, and the question of which to use is a separate guide. the relevant point here: if you are committing to saturation-before-compression, the compressor will react to a different signal than the dry one. set compressor parameters after the saturation is dialed in, not before — otherwise the threshold and ratio you picked are wrong for the actual material the compressor is processing.

(KERN WARM ships with three character modes — tape, tube, transformer — and each one changes what the downstream compressor “sees” by a different amount. tube saturation adds even harmonics; tape adds compression-like soft clipping; transformer adds odd harmonics with high-frequency lift. the right pick depends on what you want the compressor to inherit.)

read tube, tape, and transformer for the tonal differences per character.

job 4: compression

after the previous three jobs, what is left for the compressor to handle is the actual dynamic-range issue: peaks that are too loud, sustain that needs evening, transient material that needs taming. the compressor now acts on signal that has been cleaned, balanced, and harmonically shaped — and the gain reduction it produces is mostly responding to genuine dynamic problems, not artifacts of upstream issues.

the result is usually less compression. a vocal that “needed” 5 dB of gain reduction with broadband compression often needs only 1.5-2 dB once the sibilance, the body resonance, and the mono-incompatibility have been addressed. the work done upstream did not just fix the upstream problems — it changed how much downstream work was required.

this is the compounding effect of mix prep. each upstream fix reduces what the compressor has to do. and a compressor doing less work is almost always a more transparent, more musical compressor.

the order rule that survives every genre

different mix engineers have different chains. that is fine. but the order rule that survives every genre i have heard discussed is this:

corrective work first, creative work second.

corrective work removes problems: resonance cuts, high-pass filters, phase corrections, mono-compatibility fixes, dynamic EQ on harshness. these are decisions about what the signal should not contain.

creative work shapes the signal: compression for level and feel, saturation for character, EQ for tone, reverb for space, stereo expansion for image. these are decisions about what you want the signal to sound like.

the corrective layer comes first because creative work amplifies whatever it inherits. compress a signal with a 280 Hz boxiness, and now you have a compressed boxiness. saturate a signal with a phase issue, and now the harmonics encode the phase issue. add reverb to a track with sibilance, and now the reverb has sibilance. fix the source first, then shape it.

most compressor problems in real mixes are inherited from the corrective layer being skipped. once you stop skipping it, compression becomes a much smaller, much more boring decision — and that is the goal.

the high-pass-filter heuristic

there is a heuristic some mix engineers swear by: high-pass filter every track before compression. the idea is that low-frequency content below ~80 Hz on most tracks (vocals, guitars, keys, hi-hat overheads) is rumble, mic-stand vibration, or sub-bass that is genuinely not part of the intended source — and that rumble triggers compressor gain reduction inappropriately, since broadband compressors are sensitive to low-frequency energy out of proportion to perceived loudness.

high-passing every non-bass track at 80-100 Hz removes that energy, gives the compressor a cleaner dynamic-range picture, and tightens the low end of the mix considerably. the trade-off is that you have to be careful with sources where genuine low content matters (kick, bass, low piano notes, tom-toms) — but for most other tracks it is a free win.

a related move: high-passing the sidechain of a bus compressor at 100-200 Hz. the SSL bus comp sidechain HPF is the canonical example. it stops the kick from triggering broadband gain reduction on the entire mix while still letting the kick pass through the audio path unchanged. SETTLE in KERN PUSH implements the same sidechain HPF (6 dB/oct rolloff below 200 Hz) for the same reason.

(this is the kind of detail that turns a 4 dB pumping bus compressor into a 1.5 dB transparent one without changing any other setting. it costs nothing.)

what the compressor inherits when prep is done

run a careful prep pass on a problematic vocal. resonance cleanup with KERN SMOOTH or a dynamic EQ, mono-compatibility check with KERN CHECK, optional saturation with KERN WARM. then bring the compressor in.

what the compressor sees: a signal where the loud moments are only the genuinely loud moments. no sibilance peaks faking dynamic-range events. no body resonance triggering gain reduction. no phase wobbles producing inconsistent L/R envelope readings. the compressor’s threshold can sit higher (you do not need it as low to catch the same peaks). the ratio can be gentler (less compression is needed). the attack and release can be slower (the signal is more musical to begin with).

a 4:1 compressor at -18 dB threshold becomes a 2:1 compressor at -12 dB threshold. 4-5 dB of gain reduction becomes 1-2 dB. the compression you hear is more transparent, more musical, and audibly smaller. the listener hears the performance, not the compressor working.

this is the same compression plugin doing 60% less work. the difference is what came before.

a note on AI plugins and “intelligent” mixing assistants

a class of plugins markets itself as automated mix prep — Ozone Master Assistant, Nectar’s Vocal Assistant, Sonible’s smart series. these plugins analyze the signal and pick parameter settings automatically.

they can be useful as starting points. they cannot replace the prep pass. an AI assistant looking at a vocal can identify a sibilance peak and apply a de-esser, but it cannot tell you that the room you recorded in has a 280 Hz boxiness that also needs a static EQ cut, or that the doubled track you have running in parallel is creating a comb filter at 1.5 kHz, or that the reverb send is feeding back into the dry channel through a routing mistake. the prep pass is partly perceptual (what does my ear hear?) and partly diagnostic (what is wrong here that should not be?). neither is a fully automatable problem.

(use the assistants if you want; just do not skip the prep listening pass on the trust they have done it for you. they have not.)

the recap

the work upstream of compression has compound effects:

1. clean up resonance first — every harsh peak the compressor would otherwise react to is removed at the source.
2. fix stereo and phase — the compressor’s detector reads a clean signal, not a wobbly one.
3. decide on saturation — set its character before compression, so the compressor reacts to the right material.
4. then compress, with less depth and gentler settings than you would have used otherwise.

if you have done the first three, the fourth is small. if you have skipped the first three, the fourth is doing all the work and inheriting all the problems.

(the part most tutorials skip: this is most of mixing. compression is the most-discussed part of the chain, and prep is the most-skipped — but mixing is mostly prep, and the compressor is mostly cleanup.)

frequently asked questions

frequently asked questions

what should I do before reaching for a compressor?

three things in order. first, clean up resonance — every harsh peak and ringing frequency that the compressor would otherwise trigger on. second, fix stereo and phase issues that bus compression would amplify. third, decide whether the source needs harmonic glue (saturation) or harmonic restraint (clean leveling) before compression. if any of these is wrong, the compressor inherits the problem and you blame the compressor.

should I EQ before or after compression?

both. corrective EQ — high-pass filters, resonance cuts, problem-frequency removal — goes before compression so the compressor is not triggered by content you were going to remove anyway. tonal EQ — air shelves, broad scoops, presence boosts — goes after compression so the compression decisions do not get re-balanced by the EQ. one rule: if you are removing problems, do it before. if you are shaping tone, do it after.

why does my compressor sound mushy on bright vocals?

because the compressor is reacting to a sibilance peak or a harsh resonance, not to the vocal level. when 7 kHz spikes 8 dB above the surrounding energy, the broadband detector responds and the entire vocal ducks — including the body of the syllable that had no problem. fix the resonance with a dynamic EQ or a resonance suppressor before compression, and the compressor only acts on the parts of the vocal that were actually peaking.

do I need to clean up the stereo image before bus compression?

almost always. broadband bus compressors are sensitive to mono-incompatible content because phase cancellation across left and right makes the level detector see something different than your ears do. if your reverb tails are de-correlated to the point of mono collapse, the compressor reads the spectral peak inconsistently between the L and R sidechain feed, and you get audibly uneven gain reduction. tighten correlation issues first.

where does saturation go in the chain — before or after compression?

depends on what you want it to do. saturation before compression adds harmonics that the compressor then "glues" together — useful when you want a single-instance vintage feel. saturation after compression preserves the dynamic shape of the compression and adds final harmonic colour — useful for transparent compression with character on top. for most modern pop and hip-hop, saturation goes after compression. for analog-leaning material, it often goes before.

references

a note from the developer

this guide was the hardest one in the compression path to write because it is mostly about plugins that are not the compressor. the case for “the step before your compressor” only lands if i can be honest about how much of mixing happens before any compression at all. for me, that took a long time to admit.

i was a producer who blamed compressors for a long time. the snare wasn’t punching, so i swapped compressors. the vocal felt buried, so i swapped compressors. eventually i learned to bypass the compressor and check whether the problem was actually a compression problem — and most of the time, it wasn’t. the vocal was buried because the room was boxy, and the snare wasn’t punching because there was a 250 Hz mud cloud in the kick that nobody had cleaned up. compression had been a scapegoat for skipped prep work.

building KERN PUSH with the “step before your compressor” tagline was partly a reminder to myself. PUSH is a compressor; SETTLE is glue compression; DRIFT is upward compression. but the position is meant to anchor the plugin upstream in the producer’s mental chain — to be the thing the producer reaches for before their habitual broadband compressor, so that the broadband compressor downstream gets to do less work. the same logic works whether the compressor downstream is PUSH itself, an SSL emulation, an 1176, or nothing at all.

the test i run on every mix now: bypass the compressor. listen. if the bypass sounds 90% as good as the compressed version, you have a healthy mix and the compressor is doing finishing work. if the bypass sounds dramatically worse, the compressor is doing prep work it should not be doing — and you are about to inherit problems on every track in the session.

i am a solo developer in Copenhagen. if you have a prep workflow that took years to settle on, i want to read it. jonas@kernaudio.io. and if you have a take that disagrees with the order rule above, send it — i learn most from the producers who tell me i got something backwards.