StringFlux: DSP Architecture Case Study
StringFlux is a transient-aware, multiband granular delay and freeze plugin for stringed instruments. The design goal is to turn one performance into layered texture while preserving playable response.
Signal Architecture
- Dry/wet path split with wet-only advanced processing
- 3-band crossover before grain generation
- Host-rate history rings used as grain/freeze source buffers
- Grain bus shaper for harmonic character control
- FeedbackBus reinjection at host rate
- Mix -> Gain -> Limiter -> Output final stage
Oversampling policy (design table)
Not a benchmark; this documents the intended relationship between quality, CPU, and audio-thread safety.
| Factor | Goal | Engine behavior |
|---|---|---|
| 1x | Baseline CPU; default monitoring path | No resampler churn; simplest state machine |
| 2x / 4x | Reduce aliasing on nonlinear stages in the wet path | Reconfiguration queued; applied only at safe boundaries |
Grain Scheduling Model
- Density-driven scheduler for baseline grain cloud behavior
- Transient-driven scheduler for performance-reactive accents
- Band-aware source selection from history rings
- Shaping controls for grain length, density, pitch, and pan
Production Constraints
- Real-time safety: oversampling reconfiguration is queued and applied only at safe points.
- Latency and responsiveness: transient-following behavior must remain playable for string attacks.
- Stability under host variance: processing must tolerate differing host rates and plugin states.
- Feature discipline: avoid effect sprawl while core instrument behavior is still being refined.
Key Tradeoffs
Determinism over feature breadth
The engine prioritizes predictable scheduler behavior and stable routing over adding more effect modules early.
Safe oversampling transitions over immediate switching
Oversampling changes are deferred to safe boundaries to avoid audio-thread instability and state corruption.
Playable response over maximal density
Scheduler behavior favors transient readability and musical control rather than maximal grain saturation at all times.
Current Outcome Signal
Technical outcome: the current build supports multiband routing, transient and density-driven grain scheduling, freeze/history capture, feedback-bus processing, and safe oversampling state transitions. The next phase is focused on validating playability and consistency under broader host and performance conditions.