Polyphony: Voices & Voice Allocation

Until now our synthesizer has used one voice. Polyphony means the instrument can manage several independent voices at the same time.

From the arpeggiator: hold C + E + G → C → E → G. With polyphony: hold C + E + G → C + E + G.

1. Why polyphony complicated synthesizer architecture

A monophonic architecture

keyboard → oscillator(s) → filter → VCA/envelope → output

One note needs one active voice path.

A polyphonic architecture

voice 1 ┐
voice 2 ├ → mix → output
voice 3 ┘

Independent simultaneous notes require multiple voice paths plus a system for deciding which note goes to which voice.

Early analogue polyphony therefore multiplied circuitry, calibration and control problems. A landmark example was the Sequential Circuits Prophet-5 (1978): five programmable polyphonic voices, microprocessor voice control and patch memory.

Prophet 5
Important concept: polyphony is not simply “one oscillator playing a chord.” A polyphonic synthesizer allocates and manages multiple voices, then mixes their outputs.

2. Interactive polyphonic synth

Held: —

Voice limit

Try a triad with 1, 2 and 3 voices. Then choose 5 voices, hold five notes, and add a sixth.

Simple presets

Keyboard

Computer keys: A W S E D F T G Y H U J K. Unlike the arpeggiator page, simultaneous held keys now request simultaneous voices.

Live voice allocator

Architecture

Illustrative analogue drift

Small tuning differences between analogue voices, exaggerated here for demonstration.
Deliberate unison detune. Drift represents instability/variation rather than a purposeful detune setting.

3. Each voice has its own state

With a pad preset, release one key while holding the others. That voice enters its own release stage while the other voices continue independently.

VOICE = oscillator + filter + amplifier/envelope + its current note/state

This is why polyphonic implementation is structurally different from our earlier monophonic synth.

4. What happens when we run out of voices?

If every voice is busy and another note arrives, the synthesizer needs a policy. This demonstration uses a simple oldest-voice stealing rule, prioritising released notes before active notes.

5 voices busy + new note → oldest/released voice is reassigned

The stolen voice will flash in the allocator. Real synthesizers can use different allocation and stealing strategies.

5. Analogue voices were not perfectly identical

Analogue oscillators and other components can vary with temperature, calibration and component tolerances. When several analogue voices are combined, tiny differences between voices can contribute to the character of the instrument — while excessive drift becomes a tuning problem.

Our drift control is illustrative. It deliberately exaggerates per-voice pitch offsets so the idea can be heard and seen; it is not a circuit model of a particular vintage synthesizer.

6. Synthesis fundamentals complete

oscillators → filters → ADSR → filter envelope → LFO → arpeggiator → polyphony

Next we can stop thinking of synthesis as a keyboard instrument and use the same building blocks to construct drum-machine sounds: kick, snare and hi-hats.