Plasma Control System Architecture
The hardware and software organization that lets many control functions run at different rates while sharing sensors and actuators.
A real-time system of systems
A plasma control system is a distributed real-time computer system: sensor front-ends digitize diagnostics, a real-time network moves data with bounded latency, compute nodes run the control laws and state estimation, and actuator interfaces drive the hardware. All of this must meet timing guarantees, because a late command is a wrong command.
Multi-rate structure
Different functions run at different rates on the same infrastructure: vertical stabilization at tens of microseconds, magnetic control at a few milliseconds, scenario supervision at tens of milliseconds, prediction slower still. The architecture supports this multi-rate operation with clear timing domains, so a fast loop is never blocked by a slow one.
Hardware for determinism
- FPGAs and dedicated processors for the fastest, safety-critical loops
- Real-time operating systems for the millisecond control layers
- A deterministic real-time network with bounded latency and jitter
- Isolated timing domains so slow tasks cannot stall fast ones
- Redundant paths for critical signals
Software structure
Software is organized so that safety-critical code is small, simple, and separately verified, while more complex model-based and predictive code runs where a fault cannot endanger the machine. Configuration - which loops run, with what references and limits - is data-driven, so scenarios can change without rewriting control code.
Designed for evolution
Control systems live for decades and must accept new diagnostics, actuators, and algorithms over time. Good architecture separates the stable real-time core from the evolving algorithm layer, defines clean interfaces between them, and version-controls both, so the system can improve without compromising the timing and safety guarantees at its foundation.