Electronics Guide

Articles in This Category

Confinement Approaches and What They Demand

Two confinement strategies dominate the field, and they impose almost opposite electronic requirements.

Magnetic Confinement

Tokamaks and stellarators hold a plasma in a toroidal magnetic bottle for seconds to minutes, and eventually for continuous operation. The tokamak drives a large current through the plasma itself to generate the poloidal field component, which makes the configuration efficient but also unstable: the plasma column must be actively stabilized against vertical displacement and against a family of magnetohydrodynamic modes. That requirement produces the defining electronic signature of magnetic confinement, namely fast closed-loop control. Position and shape controllers typically run at update rates from a few hundred hertz to several kilohertz, driving poloidal field converters that push and pull tens of kiloamperes.

The stellarator shifts the burden from control to fabrication. Its twisted external coils generate the entire confining field, so no plasma current is required and the vertical instability disappears. Wendelstein 7-X in Germany, the largest optimized stellarator, realizes this with fifty non-planar superconducting coils whose shapes were derived from numerical optimization. The control problem becomes gentler; the magnet engineering, coil-tolerance metrology, and cryogenic instrumentation become harder.

Inertial Confinement

Inertial confinement compresses a millimeter-scale fuel capsule so rapidly that it fuses before it can fly apart. There is no feedback loop during the event, because the event lasts a few hundred picoseconds. The electronics instead concentrates on enormous pulsed power and on precise timing. The National Ignition Facility fires 192 laser beams that deliver roughly two megajoules of ultraviolet light in a shaped pulse a few nanoseconds long, corresponding to hundreds of terawatts of peak power. A capacitor bank storing several hundred megajoules charges the flashlamps that pump the amplifiers, and beam-to-beam timing must be held to picoseconds so the drive is symmetric. Pulsed-power machines such as Sandia's Z machine take a different route, driving roughly twenty-six megaamperes through a wire array in about a hundred nanoseconds to compress fuel magnetically.

A laser fusion power plant would need to repeat this several times per second rather than a few times per day, which turns diode-pumped solid-state lasers, high-repetition-rate switching, and target injection tracking into the central electronic problems.

The Radiation and Electromagnetic Environment

Deuterium-tritium fusion sends eighty percent of its energy into a 14.1 MeV neutron, and that neutron is the reason fusion electronics cannot simply reuse industrial hardware. Neutrons at this energy knock atoms out of crystal lattices, producing displacement damage that raises leakage current in silicon, degrades bipolar gain, and shifts the threshold voltage of transistors. The accompanying gamma field adds ionizing dose. A research device such as ITER accumulates a relatively modest lifetime neutron dose, but the first-wall structures of a power plant must tolerate tens of displacements per atom, a regime in which most conventional components simply cannot survive.

Designers respond with a layered strategy. The first and most effective measure is distance: nearly all electronics is moved out of the neutron field into shielded port cells, gallery spaces, or a separate diagnostic building, leaving only passive sensors, mineral-insulated cabling, and radiation-hard optics near the machine. What must remain close is built from tolerant technologies and inspected for the failure modes that matter, including optical fiber darkening under dose, photomultiplier gain drift, and connector insulation breakdown. Motion systems inside the vessel favor resolvers over optical encoders, because a wound resolver has no semiconductor junction to damage.

Magnetic fields create a second, often underestimated constraint. Stray fields of tens or hundreds of millitesla extend well beyond the coils, saturating transformer cores, deflecting electron trajectories inside photomultiplier tubes, and exerting forces on any current-carrying conductor. Equipment sited near the machine needs magnetic shielding, non-magnetic construction, and, in some cases, a switch to solid-state photodetectors.

Finally, the plasma itself is a formidable interference source. Radio-frequency heating systems radiate megawatts in bands from tens of megahertz to well above one hundred gigahertz, and a plasma disruption dumps stored energy in milliseconds while inducing large currents in surrounding structures. Grounding topology, cable routing, differential signaling, and fiber-optic isolation are treated as first-order design decisions rather than as late-stage remediation.

Magnet Power and Quench Protection

Superconducting magnets are the largest single electronic subsystem in a magnetic confinement device. ITER's eighteen toroidal field coils, wound from niobium-tin cable-in-conduit conductor, operate at 68 kA with a peak field near 11.8 tesla, and its central solenoid runs at roughly 45 kiloamperes with a peak field of about 13 tesla. The toroidal field system alone stores on the order of forty gigajoules of magnetic energy. Delivering that current requires thyristor or hybrid converters rated in the tens of kiloamperes, fed from a dedicated substation, with output ripple tight enough not to disturb the plasma.

The corresponding hazard is the quench, a local loss of superconductivity that turns the conductor resistive and converts stored magnetic energy into heat at the fault site. Quench protection must detect the event and extract the energy before the coil is damaged. Detection is a signal-integrity problem disguised as a protection problem: the resistive voltage that signals a quench may be a fraction of a volt while inductive pickup from ramping neighbors reaches hundreds of volts, so co-wound voltage taps, bridge networks, and careful common-mode rejection are used to separate the two. Once detected, fast discharge units open the circuit and divert current into dump resistors, often with pyrotechnic breakers as an independent backup path.

High-temperature superconductors based on rare-earth barium copper oxide tape have changed the calculus. They tolerate higher fields and operate with far greater thermal margin, which is what allowed Commonwealth Fusion Systems to demonstrate a 20 tesla peak-field toroidal field model coil in 2021 and to design the compact SPARC tokamak around an on-axis field of about 12.2 tesla. That same thermal margin, however, makes quenches propagate far more slowly than in low-temperature superconductors, so a normal zone can grow hot locally before conventional voltage detection notices it. Fiber-optic distributed temperature and strain sensing, acoustic emission monitoring, and Hall-array field mapping are all being developed to close that gap.

Heating and Current-Drive Power Electronics

Ohmic heating from the plasma current stops being effective well below fusion temperatures, because plasma resistivity falls as temperature rises. Auxiliary heating supplies the rest, and each method carries a distinctive power-electronics problem.

Neutral beam injection accelerates ions, neutralizes them so the magnetic field cannot deflect them, and fires them into the plasma. ITER plans two heating injectors delivering 16.5 megawatts each, for 33 megawatts total, using deuterium beams at one megavolt produced from negative ions. Generating and regulating a stable megavolt-class direct-current supply, protecting the accelerator grids against breakdown within microseconds, and transmitting that potential through an insulated transmission line constitute one of the hardest high-voltage engineering tasks in the field.

Ion cyclotron resonance heating couples tens of megawatts in the range of roughly 40 to 55 megahertz through antennas mounted at the vessel wall. Because the plasma edge moves, the antenna load impedance moves with it, and matching networks must retune faster than the plasma changes to avoid reflecting power back into the transmitters. Electron cyclotron resonance heating uses gyrotrons, vacuum tubes that convert an electron beam into millimeter-wave power; ITER's system is built around one-megawatt units near 170 gigahertz driven by well-regulated cathode and body supplies, with fast protection to limit fault energy into the tube. Because millimeter waves can be steered by movable mirrors and deposited on a chosen flux surface, electron cyclotron systems double as a surgical tool for suppressing neoclassical tearing modes.

Real-Time Plasma Control

A tokamak control system reads magnetic sensors, kinetic diagnostics, and actuator states, computes an equilibrium reconstruction, and commands converters, gas valves, pellet injectors, and heating systems, all inside a loop whose period is measured in fractions of a millisecond. Vertical stabilization is the tightest of these loops, since an elongated plasma is passively unstable on a growth timescale of a few milliseconds and will drift into the wall without continuous correction. Shape and position control, current profile control, density control, and divertor detachment control layer on top, each with its own bandwidth.

Magnetic measurement introduces a subtle constraint. Pickup coils sense the time derivative of flux, so the control system must integrate their signals, and any small offset accumulates as drift over a long pulse. Devices intended for hundreds or thousands of seconds therefore use chopped or digitally corrected integrators, or supplement magnetics with Hall sensors and other steady-state field measurements.

Disruption handling sits apart from ordinary control. A disruption terminates the plasma current in milliseconds, converting stored thermal and magnetic energy into localized heat loads, halo currents in the vessel, and beams of runaway electrons. The response is prediction followed by mitigation: a real-time classifier flags an approaching disruption, and shattered pellet injection fires cryogenic pellets of neon or argon into the plasma to radiate the energy away evenly before it can concentrate. The entire chain, from detection to material arriving in the plasma, must complete in tens of milliseconds, which sets hard latency budgets for the diagnostics, the algorithm, and the injector valves.

Machine learning has moved into this loop over the past several years. Deep reinforcement learning has been used to synthesize magnetic controllers that hold unusual plasma shapes on the TCV tokamak, and learned controllers have demonstrated real-time avoidance of tearing modes on DIII-D. These results also raise the question that will govern their adoption in a licensed power plant, namely how a learned controller can be verified and how a conventional protection layer should backstop it.

Diagnostics, Timing, and Data Acquisition

ITER will field roughly fifty diagnostic systems, and their outputs serve three distinct customers: real-time control, machine protection, and physics analysis. The measurement techniques span most of experimental physics. Thomson scattering fires a pulsed laser through the plasma and analyzes the Doppler-broadened scattered light to obtain electron temperature and density profiles. Interferometry and reflectometry infer density from the phase of a microwave or far-infrared beam. Bolometers measure total radiated power. Neutron cameras and activation foils determine the fusion rate, which is the direct measure of performance. Magnetic coils, flux loops, and Rogowski coils supply the fast signals that control depends on.

Each of these places different demands on the front end. Thomson scattering needs photodetectors with nanosecond response and gated integration synchronized to the laser. Neutron counting needs pulse-shape discrimination to separate neutrons from gammas. Magnetic diagnostics need microvolt-level stability across a pulse lasting minutes. What unifies them is timing: a facility-wide distribution system provides a common clock and event triggers so that thousands of channels sampled at different rates can be aligned to a common timebase during analysis.

The resulting data rates are substantial, reaching terabytes per experimental day at the larger devices, and they are managed by dedicated control and data infrastructure. ITER's Control, Data Access and Communication system defines a standard software and hardware framework, built largely on established open control frameworks, so that components supplied by seven partners across dozens of countries interoperate. Alongside it run two independent layers with different authority: an interlock system that protects the investment, and a safety system that protects people and the environment and is designed and licensed to nuclear standards.

Fuel Cycle, Tritium, and Nuclear Instrumentation

Deuterium is abundant in seawater, but tritium is radioactive with a half-life of about 12.3 years and exists in only trace quantities in nature. A deuterium-tritium power plant must therefore breed its own fuel, using the fusion neutrons to transmute lithium in a blanket surrounding the plasma. Sustaining operation requires a tritium breeding ratio above unity once losses and decay are accounted for, and demonstrating that number is one of the central open questions in fusion engineering. ITER will test candidate blanket concepts as test blanket modules rather than breeding at scale.

The instrumentation this implies is closer to nuclear plant practice than to laboratory physics. Tritium inventory must be tracked continuously through ionization chambers and gas-analysis systems, in a facility where the isotope permeates hot metals readily. Neutron flux monitoring, activation measurement, and coolant chemistry monitoring feed both operations and regulatory reporting. Because the environment is activated, maintenance is performed by remote handling systems whose cameras, actuators, and position feedback must themselves tolerate radiation. Above all of this sits a safety instrumented system built to functional safety practice, with the redundancy, diversity, and proof-testing regime that a licensed nuclear installation requires. ITER is licensed as a basic nuclear installation under French law, and its safety-classified electronics is engineered and documented accordingly.

Where the Field Stands

Fusion has recently passed several thresholds that had stood open for decades, and each one rested on electronics.

In December 2022, the National Ignition Facility achieved scientific ignition, producing 3.15 megajoules of fusion energy from 2.05 megajoules of laser energy delivered to the target. Subsequent campaigns improved substantially: an experiment on 7 April 2025 yielded about 8.6 megajoules from roughly 2.08 megajoules of laser energy, a target gain above four. These figures compare fusion output to light on target, not to the electrical energy drawn from the grid, which remains far larger.

On the magnetic side, the Joint European Torus set a record of about 69 megajoules of fusion energy in a single 5.2-second pulse on 3 October 2023, from roughly 0.2 milligrams of fuel, during its final deuterium-tritium campaign before shutdown at the end of that year. Long-pulse operation advanced in parallel: the EAST tokamak in China sustained high-confinement plasma for 1,066 seconds in January 2025, and the WEST tokamak in France sustained a plasma for 1,337 seconds in February 2025. Those durations shift the engineering emphasis from pulse-mode performance toward steady-state issues such as integrator drift, actively cooled components, and thermal management of the electronics themselves.

ITER, the largest of these efforts, adopted a revised baseline that was presented to its council in 2024. Under that plan, the start of research operation is scheduled for 2034, full magnetic energy and 15 megaamperes of plasma current for 2036, and the start of deuterium-tritium operation for 2039, with a more complete machine available at first operation than the earlier schedule envisaged. In parallel, privately funded ventures pursue faster and smaller routes, including high-field tokamaks built on rare-earth barium copper oxide magnets, field-reversed configurations with direct electrical energy recovery, and laser-driven concepts. Whichever approach reaches the grid first, the same electronic disciplines underpin all of them.

Conclusion

Fusion is often described as a plasma physics problem, but the path from a laboratory plasma to a power plant runs through electronics. Confinement is a control problem, heating is a power conversion problem, performance measurement is an instrumentation problem, and fuel self-sufficiency is a monitoring and safety problem. Each subcategory in this section examines one of these domains in the detail that its engineering deserves.

Related Topics