Power Switches
Voltage, current and speed pull the physics in different directions — which is why five device families are still in production, and why SiC broke the trade rather than moving along it.
Skip to the animationA power switch is chosen on blocking voltage, conducting current and switching speed together — three requirements the semiconductor physics pulls apart, which is why several device families remain in production and why wide-bandgap materials changed the map.
Where the loss comes from
Conduction loss happens while the device is on and is set by the load current. Switching loss happens during each transition, when voltage and current are simultaneously non-zero — and it is proportional to switching frequency.
That gives the fundamental trade of the whole subject: higher frequency shrinks the magnetics and capacitors — a transformer's size falls roughly as 1/f — and heats the semiconductor. The heatsink sets the ceiling.
The device families
| Device | Voltage | Frequency | Character |
|---|---|---|---|
| Thyristor (SCR) | Up to 8 kV | Line frequency | Latches on; gate cannot turn it off |
| GTO / IGCT | kV | Low kHz | Turn-off capability, at the cost of a large gate drive |
| MOSFET | Up to ~600 V | Hundreds of kHz | Fast, voltage-driven, parallels safely |
| IGBT | 600 V – several kV | Tens of kHz | MOSFET gate, bipolar conduction; current tail at turn-off |
| SiC / GaN | Wide range | Very high | Thin drift region: faster and cooler at once |
Why the thyristor shaped a generation of circuits
An SCR latches on from a gate pulse and cannot be turned off by the gate. In AC circuits the current passes through zero every half-cycle and it commutates naturally. In DC circuits, an auxiliary forced commutation circuit had to drive the current to zero — which is why classic DC drives were far more complex than their AC equivalents.
The MOSFET's ceiling and the IGBT's answer
Blocking more voltage needs a thicker, more lightly doped drift region, so a MOSFET's R_DS(on) rises roughly as V^2.5. Above about 600 V, silicon MOSFETs stop being sensible.
The IGBT fills that gap with a MOSFET gate driving a bipolar output stage, whose conductivity modulation keeps the drop low at high voltage. Its weakness is a current tail at turn-off as stored charge recombines, capping it at tens of kilohertz.
Wide-bandgap devices
Silicon carbide and gallium nitride have critical fields roughly ten times silicon's, so the same blocking voltage needs a ten times thinner drift region. Thinner means lower on-resistance *and* less stored charge — faster and cooler simultaneously.
That breaks the trade rather than moving along it, which is why SiC displaced silicon IGBTs in electric-vehicle traction inverters. SiC also conducts heat about three times better, so what heat is produced leaves more easily.
Thermal design
Thermal design is an exact analogue of a resistive circuit: T_j = T_amb + P·ΣR_th, with heat flow for current and temperature for voltage. Junction temperature is what limits the current rating.
The case-to-heatsink interface is frequently the largest single term — and the cheapest to improve, with better thermal interface material or a flatter mounting surface.
The numbers you will be asked for
- Total loss
P = I²·R_on·D + (E_on + E_off)·f_sw
- Junction temperature
T_j = T_amb + P · ΣR_th
- MOSFET on-resistance
R_DS(on) ∝ V_breakdown^2.5
- Transformer size
roughly ∝ 1 / f_switching
- Switching energy
E ≈ ½·V·I·t_transition
Watch it work
Check yourself
question 1 / 4
One question at a time. Pick an answer to see why it is right or wrong, then move on — there is no score to keep and nothing is saved.