Vertical Stability Control
An elongated plasma is vertically unstable on millisecond timescales; L1 runs a fast feedback loop on the breeder to arrest vertical displacement before it grows.
The instability
A vertically elongated tokamak plasma sits in an unstable equilibrium in the vertical direction: a small upward or downward displacement produces a force that pushes it further. Left uncontrolled this grows on the vertical instability timescale — milliseconds — and can lead to a vertical displacement event (VDE) and disruption. Active feedback is mandatory.
The fast loop
L1 estimates the plasma's vertical position z from magnetics each cycle and drives a fast vertical-control coil (or a designated PF-coil combination) to produce a restoring radial field. The loop is a stabilizer: u = -(k_p z + k_d dz/dt), where the derivative term damps the growth. It runs well inside the instability timescale, with a period around 100 µs and worst-case latency inside the 10 µs protection class for its innermost action.
Why it is a protection-class loop
- Growth is exponential; late action cannot recover.
- Runs on deterministic fabric, not on any learning component.
- Derivative damping requires low-jitter, coherent sampling.
- Coordinated with radial and shape loops on the shared clock.
Because a missed vertical-control deadline can end in a disruption, this loop exemplifies why L1 exists. It cannot wait on model inference or tolerate scheduling jitter; it must act on the deterministic pipeline every cycle. The controller gains come from an offline model, but their execution is pure fabric.
Relationship to disruptions
Vertical control is the first defense against VDE-driven disruptions. When it detects that displacement is exceeding what it can recover, it escalates to disruption avoidance and, if needed, mitigation. The 9.66 MA plasma current makes a disruption energetic, which is why the vertical loop is engineered with margin and coordinated tightly with the current-feedback loop.