Control Latency Budget
The latency budget accounts for every microsecond between a measurement and the resulting actuator response, and must fit inside the loop period.
Why latency dominates stability
Feedback control adds phase lag proportional to loop latency. For an unstable mode, too much lag pushes the closed loop from correcting the mode to reinforcing it. The maximum tolerable latency is set by the growth rate of the fastest instability the loop must catch. A faster-growing mode demands a shorter budget.
Where the time goes
- Sensor integration and digitization
- Signal transport over the data network
- State estimation and filtering
- Control algorithm computation
- Command transport to the actuator
- Actuator physical response and coil inductance
The last term, actuator response, is often the largest and least compressible. A coil with high inductance cannot change current instantly regardless of how fast the computer is. Good control design does not fight this; it chooses actuators whose bandwidth matches the mode to be controlled.
A worked example
Suppose a vertical instability grows with an e-folding time of two milliseconds. A common rule is to keep total loop latency well under one e-folding time, so a budget near a few hundred microseconds is required. If digitization takes 100 microseconds, transport 50, computation 100, and the coil response 200, the loop is already tight and there is little room for jitter. Engineers then either speed a stage up or reduce the plasma elongation to slow the instability.
Design tradeoffs
Shortening latency can raise cost and complexity, so it is spent where it buys the most stability. Slow loops, like current-profile control, get relaxed budgets. Fast loops get dedicated hardware. The Hyperion breeder, being a low-aspect-ratio elongated device, spends much of its latency budget on the vertical loop, which is the tightest constraint in the whole system.