The Leslie speaker is the sound of a B-3 organ, the textural shimmer in countless rock and gospel records, the thing every "rotary" plugin tries to imitate. What gets lost in the marketing is that it isn't a chorus. It isn't a vibrato. It isn't a tremolo. It is all three at once, generated mechanically by two rotating mechanical sound sources running at different speeds inside a wooden cabinet — and that shared origin is why software emulations sound either convincing or obviously fake.
If you understand why two rotors are necessary, you understand the entire effect.
What is actually rotating
Don Leslie bought a Hammond organ in 1937, took it home, and was disappointed. The showroom had been alive; his living room sounded flat. He concluded the problem wasn't the organ — it was that the speaker presented a single, fixed sound source. So he built a cabinet that moved the sound source through space, and in 1941 began selling it. Laurens Hammond hated the device, refused to market it, and spent decades trying to make Hammond organs "Leslie-proof" by changing connectors. Players bought adapters anyway.
A typical Leslie cabinet (the iconic 122 and 147 models) splits the audio into two bands at around 800 Hz with a passive crossover. The high band feeds a compression driver firing into a rotating horn assembly — actually one functional horn and one dummy counterweight, mounted to a horizontal disc that spins around a vertical axis. The low band feeds a 15-inch woofer pointing down at a rotating drum, which is a wooden cylinder with one open scoop cut into the side. The driver doesn't move; the drum that redirects its sound does.
The two rotors run at independent speeds, driven by separate motors. On the "tremolo" (fast) setting, measurements of an original Leslie 147 give the upper horn around 400 RPM and the lower drum around 340 RPM. On the "chorale" (slow) setting, the horn drops to about 48 RPM and the drum to about 40 RPM. They are deliberately not synchronized.
The non-synchronization matters. If both rotors spun at exactly the same rate, the two bands would modulate together and the brain would parse it as one moving source. With the rates offset, the bass and treble drift in and out of phase, and the ear hears a complex, three-dimensional motion that no single moving object could produce. This is also why the chorale-to-tremolo transition is so distinctive: each motor has its own ramp time (the drum has more inertia and accelerates more slowly), so during the speed change the relative phase relationship between bass and treble sweeps through every possible combination.
Three effects from one cabinet
A microphone parked next to a running Leslie picks up three simultaneous modulations:
1. Doppler shift. As the horn rotates, its mouth alternately approaches and recedes from the listening point. Approach raises pitch, recession lowers it. The maximum Doppler deviation is governed by the tangential velocity of the horn mouth: a horn with a 0.18 m radius spinning at 400 RPM gives v ≈ 7.5 m/s, which against the speed of sound (≈343 m/s) produces a peak frequency deviation of roughly ±2.2%, or about 38 cents. A small interval, but modulated at 6.7 Hz it's instantly audible.
2. Amplitude modulation. Speakers are directional. A compression driver firing into a horn has significant gain on-axis and rolls off as it points away. As the horn rotates, the on-axis lobe sweeps past the listener once per revolution, producing a periodic amplitude variation — classic tremolo, locked in frequency to the Doppler.
3. Comb filtering. Sound from the moving driver also bounces off the inside of the cabinet, the walls of the room, and the second (dummy) horn. The reflected paths have time-varying lengths because the source is moving. Sum the direct and reflected signals at the listening point and you get a moving comb filter on top of everything else. This is the "spatial" character: not a stereo effect, but a constantly shifting timbre.
All three modulations share a frequency (the rotor's RPM divided by 60), but they have different waveforms — Doppler is roughly sinusoidal, AM is more peaked, comb position is irregular — and crucially, the upper and lower bands are doing all three at slightly different rates.
How plugins reproduce it
Stanford CCRMA's Smith, Serafin, Abel and Berners published the canonical DAFx paper on Leslie simulation in 2002, and most modern emulations follow some variant of their structure.
The Doppler effect is implemented as a variable-delay line per band. A cosine LFO running at the rotor frequency modulates the read pointer. Linear interpolation between samples is too noisy here — the interpolation error itself becomes audible as harmonic distortion modulated at the LFO rate, which sounds like a buzzy aliasing artifact riding on the swirl. Cubic Hermite or 4-point Lagrange interpolation is the usual fix; the better implementations use sinc interpolation with a small windowed kernel.
The amplitude modulation is a tremolo with a non-sinusoidal envelope, shaped to match the directional response of the actual driver and horn. Some plugins go further and model the frequency-dependent directionality: high frequencies are more directional than mid-highs, so the AM depth should increase with frequency.
The comb-filter component is the hardest to fake convincingly. Naive approaches add a fixed feedback delay and call it done, which is why cheap rotary plugins sound static even with the LFO moving. The Smith-CCRMA approach is to model the cabinet with a small set of moving image sources — the dummy horn, the cabinet walls — each contributing its own short variable delay, summed with attenuation. This is essentially a tiny moving-source reverb running at the rotor speed.
Two-microphone simulation is now standard at the high end. A real Leslie is usually miked with two SM57s or 421s on the upper horn (wide stereo) and one mic on the drum below, and the most credible plugins synthesize the same configuration: two listening points 90 degrees apart on the upper horn, each with its own delay/AM curves shifted by a quarter rotation, and a separate close-mic model for the drum. Convolution-based hybrid approaches use real cabinet impulse responses and then add the time-varying delay on top.
What this means for using it
Three things follow from the physics.
First, stereo width on a Leslie is not a width control — it's the difference between two listener positions on the same rotating source. Narrowing the mics narrows the swirl. Don't reach for a stereo widener after the fact; move the virtual mics in the plugin instead.
Second, the slow speed is the more demanding test of a plugin's quality. At 400 RPM, almost anything sounds plausible because the ear can't track the detail. At 40 RPM, every Doppler artifact, every interpolation glitch, and every flatness in the comb-filter model is exposed. Audition rotary plugins on chorale, sustained chord, no other modulation in the chain. The bad ones reveal themselves immediately.
Third, the ramp matters as much as the speeds. The classic Leslie sound on a held chord is the moment a player kicks the speed switch and the cabinet takes 1–2 seconds to slide between modes. If your plugin's speed transition is instantaneous or linear, you've removed half the expressive vocabulary. Look for separate ramp-up and ramp-down times for the horn and drum — it's the single most common compromise in cheap emulations.
The Leslie is one of the rare effects whose mechanical complexity has no shortcut. You can fake a chorus, fake a phaser, fake a flanger. You cannot fake a moving sound source convincingly with a static one. Every good rotary simulation is, underneath, a small physics engine running at audio rate.