The 20–30A Crossover: Where the Math Flips

Start with the loss equations, because everything else follows from them. A MOSFET conducts like a resistor: P_cond = I² × R_DS(on). An IGBT conducts like a voltage source: P_cond = V_CE(sat) × I. One is quadratic, the other linear — and quadratic always loses at high current.

Work a real example at 650V class. A 40A MOSFET at 80mΩ burns 40² × 0.08 = 128W conducting. A 40A IGBT at 1.45V burns 40 × 1.45 = 58W. Same current, half the heat.

That's the entire argument for IGBTs above the crossover — and it's why every motor-drive inverter on the market runs IGBTs.

Now drop to 5A. The same MOSFET burns 2W; the same IGBT still burns 7.25W because its 1.45V floor doesn't shrink with current. Below the crossover, the millivolt-level drop of a good MOSFET beats any V_CE(sat), no matter how the datasheet is spun.

Where exactly does the flip happen? For 600V silicon devices, around 20–30A at 25°C — and the number slides down as temperature rises.

R_DS(on) roughly doubles from 25°C to 150°C while V_CE(sat) drifts only modestly. A design that crosses at 25A on paper can cross at 15A in a 90°C enclosure.

The Tail Current: Why 10kHz Motor Drives Are IGBT Country

Conduction is only half the story. The IGBT's low drop comes from minority-carrier injection — and those injected carriers must recombine before the device turns off.

That recombination shows up as a tail current: the collector keeps conducting for a slice of time after the gate signal ends, burning switching energy every cycle.

A MOSFET, unipolar, has no stored charge to sweep out. It turns off clean in tens of nanoseconds. That difference sets the frequency ceiling: IGBTs are efficient only below roughly 20–50kHz, while superjunction MOSFETs switch clean well past 100kHz.

Motor drives run at 1–20kHz. Welding machines, a few kHz. UPS systems, 10–20kHz. All of them are conduction-dominated, low-frequency, high-current — textbook IGBT territory, and exactly what the Shenzhen boards confirm. High-frequency PFC stages, resonant converters, and fast chargers run 100kHz+, which is MOSFET territory, period.

The Three Things Nobody Warns You About

1. The missing body diode

A MOSFET's body diode conducts reverse current for free. An IGBT has no built-in reverse path at all.

Every bridge circuit running IGBTs needs an antiparallel diode per device, or the freewheeling current has nowhere to go. That's extra parts, extra layout, and a failure mode that never appears in a MOSFET design.

2. The short-circuit window

IGBTs typically withstand a short circuit for 5–10µs; MOSFETs need protection in microseconds. In an industrial drive with contactor bounce or a stalled motor, that extra time is the difference between a protection trip and a burned module — a big reason fault-tolerant industrial designs favor IGBTs.

3. The linear-region trap

IGBTs hate linear operation — small dies, poor current sharing, thermal runaway. Electronic loads and soft-start stages need MOSFETs designed for linear work. The dummy-load threads on EEVblog steer everyone the same way: don't build a linear load out of IGBTs.

When the MOSFET Wins Even at 650V

Modern superjunction MOSFETs (CoolMOS, MDmesh class) made 650V a genuine contest, not a walkover. They switch fast, conduct well at moderate current, and come with the body diode built in.

At 650V with 5–10A and 100kHz, a superjunction MOSFET beats any IGBT on total loss — which is why PFC stages and flyback supplies at 650V are MOSFET country.

There's also the voltage-scaling physics: silicon MOSFET on-resistance rises roughly with V^2.5, which is why 100V MOSFETs are cheap and 1200V MOSFETs are impractical.

At 650V the superjunction technology keeps R_DS(on) manageable. Push to 1200V and the die area explodes — 1200V-class designs are IGBT (or SiC) territory by default.

The 4-Question Decision Framework

Question 1: What's the average current? Below ~20A, the MOSFET's millivolt drop wins. Above ~30A, the IGBT's flat V_CE(sat) wins. In between, the next three questions decide.

Question 2: What's the switching frequency? Under ~20kHz, conduction dominates and the IGBT looks good. Above ~50kHz, the tail current makes the IGBT a loser and the MOSFET's clean switching wins.

Question 3: Is there reverse-current conduction? Half-bridges, synchronous rectifiers, and motor commutation all push current backward through the switch. The MOSFET's body diode handles it for free; the IGBT needs an antiparallel diode per device — count the cost and the layout space.

Question 4: How hot will the junction actually run? Evaluate at the worst-case temperature, not 25°C. R_DS(on) doubles by 150°C, dragging the crossover down; a board that "passes at 25A on paper" can flip to IGBT territory in a sealed enclosure.

The one-line verdict: high current + low frequency + hot environment → IGBT. Low to moderate current + high frequency + tight space → MOSFET. At 650V, both are legitimate — the math at your operating point decides, not the marketing.

Frequently Asked Questions

Q1: Is 650V the crossover voltage between IGBT and MOSFET?

A: It's the contested zone — below ~500V MOSFETs dominate, above ~650V IGBTs take over, and in between the current and frequency decide. Superjunction MOSFETs made 650V viable for moderate current; 1200V-class designs are IGBT or SiC territory because silicon MOSFET resistance rises with the 2.5th power of voltage.

Q2: Can I replace an IGBT with a MOSFET in a motor drive?

A: Only at low power — a 650V MOSFET with enough current rating is expensive and the switching frequency doesn't help at 5kHz anyway. At motor-drive currents (10A+), the IGBT's flat V_CE(sat) wins on conduction loss, and the drive frequency never needs the MOSFET's speed.

Q3: Why do IGBTs have no body diode?

A: The bipolar structure doesn't create one — the P-N collector junction blocks reverse current. Bridge circuits must add antiparallel diodes per device. This is a real BOM and layout cost that never appears in MOSFET designs.

Q4: What does "tail current" actually cost?

A: It costs switching energy every cycle — typically 5–10× the per-cycle switching loss of a comparable MOSFET. At 10kHz that's tolerable; at 100kHz it dominates the loss budget and the IGBT becomes inefficient. It's the physical reason IGBTs stop at ~50kHz.

Q5: Do IGBTs need special gate drivers?

A: Not always — but bridge designs in noisy environments benefit from a −5V to −15V off bias to stop parasitic turn-on. EEVblog and Infineon discussions trace the false turn-on to Miller-capacitance coupling (dv/dt) and stray emitter inductance (di/dt); a negative bias keeps the gate below threshold through those spikes.

Modern trench field-stop IGBTs run fine with 0V off in benign appliance drives — the negative rail is a robustness measure for industrial noise, not a universal requirement. Miller clamps help with dv/dt but not di/dt, which is where the negative bias earns its keep.

Q6: Which is more robust in a fault?

A: IGBTs win on short-circuit withstand (5–10µs) but lose on linear-region operation. For fault-prone industrial environments the IGBT's wider protection window matters; for electronic loads and linear stages, MOSFETs are the only safe choice.