This bit from Bitwig's modular synthesis tutorial covers pulse width modulation, starting with a clear explanation of how pulse waves differ from sawtooth waves in their harmonic content. Where a sawtooth presents all harmonics, a square wave (the default pulse wave at 50% width) drops every other harmonic. Shift the width to one-third and every third harmonic disappears instead, making the relationship between pulse width and harmonic spectrum predictable and usable.
The real interest comes from modulating that width over time. The host pairs a pulse oscillator with a slow LFO to demonstrate classic PWM, where the gradual sweep creates smooth timbral movement and an instant pad-like quality. Swapping the LFO for a step sequencer pushes things further, with more extreme width values producing waveforms so narrow they become almost inaudible at the edges.
Per-voice randomization of the step modulator adds another layer, giving each new note its own pulse width state and a more unpredictable, textured result. The core takeaway is that PWM is one of the most direct ways to make a static oscillator feel alive, with the character of the movement shaped entirely by what you use to drive the modulation.