Pulsar timing data from the ATNF catalogue drives an experimental synthesis engine. Each voice is a layered FM synthesizer with vocal formant filtering and a stochastic modulation engine — parameter shifts happen at Poisson-distributed random intervals, not on any musical clock.
| Pulse Period (P) | Carrier oscillator base frequency | 1ms–20s → 800–30 Hz |
| Period Derivative (Ṗ) | FM depth + slow pitch drift | irrational ratios: φ, √2, √3, e/2 |
| Dispersion Measure | Delay time + reverb depth | 0–1000 pc/cm³ → 80ms–1.5s taps |
| Characteristic Age | Carrier waveform + formant rate | young=sawtooth/fast → old=sine/slow |
| Surface B-field | Filter cutoff + shimmer intensity | 10⁸–10¹⁵ G → bright/dark |
- FM synthesis with irrational carrier/modulator ratios — never locks into stable harmonics
- Formant filter bank (F1/F2/F3) interpolates between six vowel positions — /a/, /ε/, /i/, /o/, /u/, schwa
- Breath noise layer — filtered white noise pulsed stochastically for texture
- Soft-clip waveshaper — tanh saturation adds harmonic warmth
- Three-tap cascade delay at irrational time ratios (t, t×φ, t×e) — never repeats
- Shimmer convolver — long reverb IR with harmonic halo buildup in the tail
- Stochastic Poisson engine — 7 event types (vowel morph, FM jump, shimmer bloom, breath pulse, pitch drift, freeze, double event) at exponentially-distributed intervals
Press F for present mode. Up to 4 simultaneous voices. Shift+click to remove a voice from the diagram.