The Pump

What it is

A kick and a bass at 128 BPM, with a compressor on the bass keyed from the kick, so the bass drops out every time the kick lands and swells back in between. That’s the pump, and it’s the shape of house.

The reason given for it’s loudness. The kick and the bass want the same headroom at the same moment, and if the bass gets out of the way they stop stacking, so the whole mix can come up before it hits the same peak. This page measures that. Both mixes, pumped and not, are brought to the same peak and their loudness read in the units from day 63.

The pump buys about half a loudness unit, when everything is set right. Set the way most people set it, it costs one and a half.

How it works

The kick is the 808 from part six. The bass is the saw from part one through the filter from part two, playing straight through. The compressor listens to the kick with a peak follower, and once the kick is over a threshold it turns the bass down by seven eighths of the excess, a ratio of eight, and lets it back at the release time.

The loudness is K weighted with the published coefficients at 48 kHz, the head’s high shelf and then the RLB highpass, exactly as day 63 did it, and the meter check is a 1 kHz sine at full scale reading -3.0 LUFS. The loudness gain is the pumped mix against the plain one after both are normalised to the same peak. The meter reading is the swing of the ducked bass’s momentary loudness, a 400 ms window sliding every ten milliseconds.

The audio path is the same recipe with the real nodes. The compressor’s gain curve for one beat comes out of the render and is applied to the bass as a value curve, a lookahead before each kick. Rendered offline as a whole and measured the same way, the real graph gives -1.56, 0.21, 0.56, 0.88 and -1.12 LU of gain across five settings against the maths’ -1.50, 0.21, 0.58, 0.89 and -1.11, and the meter readings agree to a tenth.

What surprised me

That the pump costs loudness before it buys any. With a 5 ms attack and no lookahead, the setting most people reach for, the pumped mix is 1.5 LU quieter than the plain one at the same peak. The bass hasn’t ducked yet when the kick’s transient lands, so the peak of the mix barely drops and no headroom is freed, and the bass then loses loudness for every moment it’s away. You pay for the pump and get nothing back.

Lookahead is what makes it pay. Two milliseconds ahead of the kick, -0.41 LU. Five, +0.21. Twenty, +0.54. Once the duck lands before the transient, the mix peak drops and the whole thing comes up. A real sidechain compressor does this by delaying the audio a few milliseconds, and this is the whole reason that feature exists.

Then the balance decides the sign. With the kick louder than the bass, which is the usual balance, five milliseconds of lookahead buys +0.58 LU. With the bass 6 dB louder than the kick, the same setting costs -0.41, because what you lose ducking the loudest thing in the mix is more than the headroom is worth. And the release matters the way you would expect: 50 ms buys 0.89 LU, 300 ms buys 0.47, because the shorter the bass is away the less loudness it gives up.

So the most the pump buys, in this mix, is under a decibel. The headroom it frees is about 3 dB, and nearly all of that’s spent again on the bass being quieter for a third of every beat. Pumping isn’t a loudness tool. It’s a groove, and the loudness is a rounding error.

The second surprise is how little of the pump a meter can see. The compressor pulls the bass down 15 dB every beat, and the momentary loudness of the bass swings by 2.4 LU. A sixth of it. The meter’s window is 400 ms and the duck is over in about 150, so the meter averages the duck with the bass either side of it and reports a gentle wobble where the compressor is doing something violent. At a 300 ms release the meter sees 1.4 LU of a 15 dB duck. If you are setting a pump by watching a loudness meter, you can’t see what you’re setting.

One more, smaller. A 150 ms release on a kick that decays for 180 ms doesn’t quite get back inside a 469 ms beat: ten milliseconds before the next kick the bass is still 2.4 dB down. Most of the way, and not all of it, and the bass in a pumped house track is never actually at full level.

What I would do next

That’s the series. Ten parts, from stacking sines into a saw to ducking that saw under a kick, and every part measured something its own knob doesn’t tell you. The one thing I would go back and add is a listener to each of them, because the ear was the missing instrument throughout, and thresholds like where a beat becomes a shimmer or when a click becomes a knock are exactly the kind of measurement the earlier days of this blog did with staircases.

For the pump itself, the interesting version is the one with a real track’s balance rather than two synthesised parts, since the sign of the result depends on it. And the lookahead finding suggests a measurement I haven’t seen made: the loudness gain against lookahead for a kick with a given attack, which would say how much lookahead a compressor needs to be worth using at all.