The hoover sound: why yours is wrong, and the one trick that fixes it

The hoover (Mentasm, Dominator, every early-90s hardcore record) comes from one patch: the Roland Alpha Juno-2 factory sound "What The", weaponised by Joey Beltram in 1991. It's one of the most recognisable sounds in dance music, and it's one of the easiest to fake badly. We tried four times. Here's what we learned, including the part where we gave up and did the smart thing instead.

What actually makes a hoover a hoover

Five things, and you need all five:

1. True PWM: a pulse wave whose width sweeps back and forth. This is the

heart of it.

1. A wide, detuned saw stack underneath for body.
2. A pitch dive into each note, so it swoops down onto the pitch.
3. A resonant lowpass sweeping down over the note.
4. A fat ensemble chorus to smear it into that hollow, metallic swirl.

Get PWM wrong and it doesn't matter how good the other four are. It won't sound like a hoover. So let's talk about the mistake.

The mistake that kills most hoover patches

Pulse-width modulation means the duty cycle of a pulse wave moves (the wave spends more time high, then more time low, back and forth), and that motion is what gives the hoover its churning, hollow character. In synthesis terms, PWM is even-harmonic energy fading in and out: a pulse at exactly 50% width has no even harmonics, and they come back as the width sweeps away from 50%.

Here's the trap. Web Audio's oscillator (and a lot of simple synths) has no pulse-width control. Its square wave is locked at 50%. So when you reach for the nearest knob, the obvious one is detune, and you modulate that instead.

That's not PWM. That's vibrato. Modulating detune wobbles the pitch and leaves the timbre completely alone. It sounds seasick, not metallic. This is the single biggest reason a hoover patch sounds wrong, and it's what our first version did. We measured it later: the broken version's second harmonic never got past 0.29× the fundamental and modulated at a rate that matched nothing in the patch. It was pitch-wobbling a static tone.

The fix

Real PWM is a subtraction of two saws:

pulse(t) = saw(t) − saw(t − d)

The delay d sets the duty cycle. Modulate d and the width sweeps: real PWM, even harmonics moving in and out. Two conditions or it collapses: both saws must start at the exact same instant (or they're not phase-aligned and the subtraction gives you silence), and you modulate the delay, not the pitch.

When we did it right, the second harmonic pushed to 5× the fundamental, modulating at the LFO rate, even harmonics swelling in and out, which is literally what a hoover is.

One more thing everyone overrides

A hoover is BRIGHT. The instinct is to close the lowpass down to make it "warm". Don't. Closing it too far kills the sound stone dead. The filter should sweep, but it lives high. The brightness is the sound.

And then we deleted it

Here's the honest ending. Even with true PWM, our synthesised hoover was expensive (three PWM saws plus a chorus insert, the heaviest preset we had) and never quite nailed it. The last 10% of a hoover is in the specific Juno chorus chip and the exact filter tracking, and chasing it was a rabbit hole.

So we did what you should do when a sound is defined by its exact timbre rather than by parameters you want to move: we sampled a real one. Four rounds of synthesis were beaten by the idea of just grabbing a clean hoover and playing it from a sampler, where you can still pitch it, chop it, filter it and drive it, but you start from the real thing.

That's the actual lesson, and it's the one coproducer is built around: synthesise what you want to control, sample what you want to nail. A hoover is a nail-it sound.

In coproducer, say "find a hoover and put it in the breakdown", and it searches, isolates a clean one, and drops it on the timeline, pitched to your key.