Mixing Tips
Attack and Release by Ear, Not by Numbers
Somebody once told me that 30 ms was the right attack for a snare. It got written on a piece of tape stuck to my desk and it stayed there the better part of a year, because on the first snare I tried it on it was perfect. Tight, forward, exactly the record I had in my head. On the second one it turned the drum into a wet cardboard box. I spent most of that afternoon convinced the compressor was broken, then a while longer convinced the drummer was, and it was neither of them. The tape stayed on the desk for months after that, which tells you how long a bad habit lives.
The number was never the setting. It is one input into a machine whose other inputs are the source, the threshold, the ratio, and whatever the designer decided a millisecond meant.
Here is what the two knobs actually do, what wrong sounds like in each direction, and how to set them with your eyes shut.
What the two knobs actually do
Threshold and ratio decide how much gain reduction a signal earns. Attack and release decide how fast the compressor is allowed to travel to that figure and back. That's the whole division of labour. A compressor's gain never jumps; it is smoothed, and the time constants are the smoothing. Set the threshold to −20 dBFS at 4:1 and a peak hitting 0 dBFS has earned exactly 15 dB of reduction. Whether it ever gets 15 dB depends entirely on the attack.
That is worth sitting with, because it is the part most people skip. The ratio is a promise about the destination. The attack decides whether the compressor arrives before the transient has already left.
Put three attack settings on the same drum hit and the difference is not subtle.
What too fast and too slow sound like
At half a millisecond the compressor catches the whole hit: the full 15 dB, reached about 4 ms in, before the stick noise has finished. That is exactly what you want on a bass guitar that jumps around, and exactly what kills a snare, because the crack of a snare is the first few milliseconds. Clamp those and the drum goes flat and distant while the meter insists nothing changed. The usual mistake is then reaching for the fader, which makes it louder and still flat.
At 50 ms the opposite happens. Three milliseconds after the hit the compressor has moved less than a decibel, so the transient sails through at full size, and the gain arrives afterwards, peaking at 8.8 dB some 76 ms later, by which point it is pulling down the tail rather than the peak. On a slow ballad that reads as punch. On a fast part it reads as a hole behind every hit, and it eats the room sound you paid for.
Ten milliseconds sits between them, giving up 2.2 dB of the available reduction in exchange for a few milliseconds of transient. That trade is the decision. Not the number.
Release is where the groove lives
Release gets set last and matters more. It decides whether the compressor lets go between events, and letting go is what your ear reads as rhythm. The three curves below are the same six eighth notes at 120 bpm, the same 10 ms attack, and nothing different except recovery time.
With 60 ms the gain climbs back to within 1.4 dB of zero before the next hit lands, so the compressor swings through roughly eleven decibels every 250 ms, over and over, in time with the part. Whether that reads as groove or as seasickness depends on the tempo and the source.
With 600 ms it never recovers past 10 dB. The track sits permanently ducked, all movement gone, which can be exactly right under a dense vocal and is deadly on drums.
Too short has a second symptom that people misdiagnose constantly. On bass and kick, a release under about 50 ms starts tracking individual cycles of the waveform instead of the envelope, and the gain modulation shows up as low-frequency distortion. It sounds like a bad amp. It is a release knob.
The rule I use: if you want to hear the beat, the gain reduction should be most of the way home before the next one. If you want a starting point in numbers before you start listening, the sweet-spot ranges we published earlier are a reasonable place to begin — just treat them as a first guess rather than a setting. If you want glue, don't let it get home at all.
Why the same number sounds different on two units
Three reasons, and none of them is mysterious. The first is that nobody agrees what a millisecond means. A time constant is normally the time to travel 63 % of the way to the target, which is the convention in the charts above, but plenty of manufacturers quote 10 % to 90 % instead, and that is a little over twice as long for the same circuit. Two panels can read 10 ms and behave two and a half times apart before anything else is considered.
The second is where the detector listens. Feed-forward designs measure the input and apply the result; feedback designs measure the output, which means the detector sees a signal that is already being reduced, so the compressor slows down as it works harder. Our BlackAsh SC-5 models that feedback loop along with a dual time-constant detector, and it is why its 200 µs to 20 ms range behaves nothing like the same span on a feed-forward plugin.
The third is that some compressors don't have time constants to quote. The VLA-2A has no attack or release control at all — Peak Reduction, Gain, a mode switch. Its T4A optical cell responds faster to loud transients, slower to sustained material, and never fully resets between hits. You are not setting a time, you are choosing a temperament.
Setting them without looking
Get the amount first. Threshold and ratio until the reduction meter shows roughly what the track needs, four to six decibels on most things, and ignore the timing knobs entirely while you do it. Then turn both time constants to their fastest position and listen to how bad it is. Fastest attack, fastest release: the transients are gone and the sustain is pumping. Good. Now you know what each knob is fixing, which you cannot know from the middle of the range.
Open the attack until the front of the sound comes back, then go a touch further and pull it back to the last setting you liked. Then slow the release until the pumping stops, and stop there rather than at the first setting that sounds smooth — smooth is usually 100 ms too slow.
One condition: match your levels first. Compression changes loudness, loudness changes your opinion, and a compressor 1.5 dB louder in bypass wins every comparison. Use the make-up gain to get back where you started, then A/B.
The numbers are notes, not instructions
What I write in a session now is a description, not a setting: fast enough to hold the bass, slow enough that the pick still reads. That survives a change of plugin. “4 ms” does not.
Want to feel the difference between a compressor with times on the panel and one with a mind of its own? Put a vocal through the VLA-FET with its 1 to 7 ms attack, then the same vocal through the VLA-2A, matching the gain reduction rather than the settings.
The FET will grab the consonants. The opto will lean on the line. Same five decibels, two different singers, and no number on either panel would have told you that.
Add a comment
You need to be logged in to add comments.