Fast vs slow attack

At a glance

What does attack time mean on a compressor?

Attack is how long a compressor takes to turn the sound down once it goes over the threshold. A fast attack gets most of the way there in a fraction of a millisecond. A slow one takes tens of milliseconds, and while it's getting there the sound plays through at close to its original level.

Whether that delay matters depends on the transient, the short burst of level at the start of a note. On the drum loop in the lab, the loudest hit rises from a quarter of its height to its peak in 2.3 ms, then falls away into a longer, quieter body. An attack much shorter than 2.3 ms is already turning down before that peak arrives. An attack of 20 ms or more is still on its way down when the peak has been and gone.

So the choice is between catching the front of each note and letting it go. The maths of how the gain moves over time, and what the release does afterwards, are in attack and release on a compressor. This page stays on fast against slow, and the lab has its knee locked so the shape of each hit comes down to the attack.

What does a fast attack do?

With a fast attack the compressor reacts to the transient itself. In step 1 of the lab, with a −18 dB threshold, 4:1 and a 0.3 ms attack, the loudest drum hit loses 9.7 dB off its spike and 10.3 dB off the body behind it. The crack comes down with the rest of the hit, so the compressor adds no punch, and because the quieter hits under the threshold are left alone, the loud and quiet hits end up closer in level.

Once makeup gain brings the level back, the drums sound smaller and further from the microphone. That's a loss if you wanted punch. It's the right sound when a part has peaks you don't want, though: a snare with a crack that hurts on headphones, a bass player who digs in hard on a few notes, a vocal that jumps on certain words.

A fast attack is also how you get the most level control out of a compressor. A peak can only be held down if the compressor is already working when it arrives, which is why limiters use the fastest attacks of all, and many add lookahead so they can start turning down before the peak gets there.

Fast attack with a fast release

Pair a fast attack with a fast release and the result is closer to an effect than to level control. In step 6, a 30 ms release lets the gain come back up straight after each hit is clamped, so the decay of the drum and the room sound around it rise up next to the hits. The compressor is working on only 40% of the loop now, against 86% with a 100 ms release, because it lets go between hits. That's the usual setting for crushed drum room mics, and it sounds rough and big.

What does a slow attack do?

With a slow attack the transient gets out before the compressor has properly reacted. Step 3 sets 30 ms at the same −18 dB threshold. The spike of the loudest hit loses 0.6 dB, almost nothing, while the body is still turned down by 3.6 dB. The front of the hit now stands 3 dB further above its body than it did in the original, and that's what people mean by punch: the compressor makes the start of a sound bigger by turning down what comes after it.

You can read the change in the crest factor, the gap between the peaks and the average level. The drum loop starts at 18.6 dB. Compression usually shrinks that gap, since it brings the peaks down. With a slow attack the peaks survive while the body comes down, so on drums the gap grows, to 20.1 dB at 30 ms. The fast attack nudges it up a little on this loop too, to 19.8 dB, because the 100 ms release keeps the gain down through the body of each hit as well as the spike. The slow attack pulls clearly ahead once both are doing the same amount of work: in step 4 the crest reaches 22.6 dB.

Why does a slow attack compress less?

Look at the gain reduction in that slow step. At 0.3 ms the compressor takes up to 11.7 dB off the loop. At 30 ms, with the same threshold and ratio, it takes 4.2 dB. A slow attack always does less on short sounds, because the sound is partly over by the time the compressor gets there.

That makes a straight swap between the two settings unfair. The slow attack sounds more natural partly because it's doing about a third as much. To compare them properly, lower the threshold on the slow setting until the gain reduction is similar. In step 4, a −32 dB threshold with a 30 ms attack takes up to 12 dB off, close to the fast setting's 11.7.

Now the numbers line up. The body of the loud hit is turned down by 10.5 dB with the slow attack and by 10.3 dB with the fast one, so the compressor is holding the drums down by the same amount. The spike is where they differ: it loses 9.7 dB with the fast attack and 2.2 dB with the slow one. At the same amount of compression, the slow attack leaves the front of the hit 7.5 dB taller.

The middle ground, 3 to 10 ms

Medium attacks catch part of the transient. At 3 ms on the drum loop, with the −18 dB threshold back in place, the spike loses 3.5 dB and the body 8.3 dB. That takes the harshest edge off the hit and still leaves it in front of the body. At this threshold, 2 to 3 ms also gives the biggest crest factor of any attack, about 21.7 dB, because the body is pulled down hard and most of the spike gets out.

This range is where a lot of vocal and snare compression sits. Around 5 to 15 ms on a vocal catches the loudest syllables without flattening the consonants at the start of each word.

When the attack is too slow

Past a point, slowing the attack stops shaping the hit and the compressor simply misses it. In step 5, a 100 ms attack at the −32 dB threshold takes at most 6.3 dB off, about half what the 30 ms attack did, and the body of the loud hit drops by only 5.1 dB. Drum hits are mostly over within 100 ms, so the gain is still coming down when the next one starts.

Very slow attacks still make sense on sustained sounds. On a pad, a string part or a mix bus, they let the compressor follow the overall level of a passage and ignore the quick details inside it.

Fast vs slow attack on different sounds

The table shows the same compressor on four of the loops in the lab, with a −18 dB threshold, 4:1 and a 100 ms release, changing only the attack. Max GR is the most the compressor takes off at any moment. Crest is the gap between peak and average level, and the figure in brackets is the loop without compression.

Drums

Drums show the difference most clearly, because so much of their character is in the transient. On a kick, a slow attack of 20 to 30 ms keeps the click at the front and turns the boom down a little, which helps the kick cut through on small speakers. On a snare, 5 to 15 ms is a common start, slower for more crack and faster when the crack is too sharp. The settings in step 7 of the lab, 20 ms at −20 dB, are a good place to begin on a drum loop or a drum bus.

Bass

Bass has two problems with fast attacks. The first is the same as with drums: the pluck or pick at the start of each note is what lets you hear a bass line on a phone speaker, and a fast attack turns it down. The second is distortion. A low note is a slow wave, about 18 ms per cycle for an A at 55 Hz, so a very fast attack can pull the gain down inside a single cycle and bend the shape of the wave.

You can measure that as wobble in the gain. On the bass loop at −24 dB and 6:1, a 0.1 ms attack with a 30 ms release makes the gain swing by about 3.8 dB, on average, within any 30 ms stretch where the compressor is working. A 30 ms attack with a 100 ms release brings that under 1 dB. To hear it, choose Bass in the lab and try both settings; the first one has a buzzy edge.

Vocals

On the vocal, the fast attack drops the crest factor from 18.7 to 13.3 dB, the biggest change of any loop in the table. That's very even, and it can sound squashed, with the start of each word pushed back into the rest of it. The slow attack leaves the vocal at 16.3 dB, with the consonants clear and the loudest syllables still poking out. A common fix is two compressors in a row: a fast one that only catches the loudest peaks, then a slower one that evens out the performance. Parallel vs serial compression covers that kind of chain.

Full mixes and buses

On the full mix, the slow attack raises the crest factor from 16.4 to 17.8 dB, so the drums come forward against everything else. That's why bus compression usually starts slow; 30 ms is a common setting on an SSL-style bus compressor. A fast attack does the opposite and drags the whole track down on every kick.

Fast and slow attack on real compressors

Fast and slow are relative to the compressor in front of you. The Universal Audio 1176 has an attack range of 20 to 800 microseconds, so even its slowest setting is fast by the standards of this page, and its knobs are reversed, with fully clockwise being fastest. The SSL bus compressor has stepped attack times of 0.1, 0.3, 1, 3, 10 and 30 ms. The Teletronix LA-2A has no attack knob at all; its optical cell reacts in roughly 10 ms, which puts it in the middle ground.

Manufacturers also measure attack time in different ways, so 10 ms on one compressor can behave differently from 10 ms on another. Set the attack by listening to the front of the note and watching the gain reduction meter, and treat the number on the knob as a rough guide.

What the compressor's detector reads changes things too. A peak detector sees every transient at full height, so the attack setting decides how much of it gets through. An RMS detector averages the level first and reads short spikes lower, so it behaves like a slower attack even with the knob turned right down. Peak vs RMS compression goes through it.

Other ways to shape the front of a note

If all you want is more punch on drums whose level is already even, a transient shaper turns the front of each hit up directly, with no threshold, so quiet hits get the same treatment as loud ones. If all you want is to shave peaks, a clipper does it instantly without turning down what comes after them.

A fast attack and punch can also live together through parallel compression. Compress a copy of the drums hard with a fast attack and blend it in under the original: the untouched copy keeps its transients on top, and the squashed copy lifts the quiet detail underneath. The lab's Mix slider does this, so try 0.3 ms with a low threshold and the mix at around 40%.

How to choose the attack, step by step

1. Start slowSet the attack to 30 ms and the release to around 100 ms, then lower the threshold until the loudest moments show 3 to 6 dB of gain reduction.
2. Bring the attack down while listening to the front of each noteStop when the peaks feel controlled. If the sound starts to lose its edge or move further away, go back up a little.
3. Check the gain reduction againSpeeding up the attack adds gain reduction, so raise the threshold if the meter now shows more than you wanted.
4. Listen to the low endOn bass, kick and full mixes, make sure a fast attack isn't adding a buzz. If it is, slow it down or lengthen the release.
5. Compare at matched loudnessUse makeup gain, or the lab's Original and Compressed switch, so the compressed version isn't winning by being louder.

Common mistakes

Setting the threshold before the attack on drums

If the attack is still fast when you start lowering the threshold on a drum part, the hits get flattened before you've judged anything else. When a drum sounds smaller with the compressor on, slow the attack first.

Judging slow against fast at the same threshold

As step 4 shows, a slow attack can do about a third of the gain reduction at the same threshold, so it wins a quick comparison by doing less. Match the gain reduction before you decide.

Slowing the attack to fix a squashed sound

If the whole part sounds lifeless, the threshold may simply be too low, with the compressor working all the time. Look at the gain reduction meter before reaching for the attack. More than 6 dB across most of the part is a lot.

Using a slow attack to control peaks

A slow attack lets peaks through by design, so it won't stop a vocal jumping out or a mix from clipping. Use a faster attack, a limiter or a clipper for that job, and keep the slow-attack compressor for shaping.

Train your ears

The quiz under the lab hides the settings and matches the levels. Every round plays the loop with either a 0.3 ms or a 30 ms attack and asks which it is. Listen only to the first few milliseconds of each hit. If it clicks, the attack is slow. Once you can get eight out of eight on the drums, try the same quiz on Vocal.

Questions people ask

Keep going