MOSFET, IGBT or SiC: Which Switch to Choose

Key points

• Below 200 V, silicon MOSFETs win on every axis including cost
• IGBT tail current makes it lose above about 20 kHz; conduction loss makes it win above 400 A
• SiC beats super junction silicon at 650 V above roughly 30 kHz, not at 65 kHz and 100 W
• The crossover is a loss calculation, and it moves with cooling, cost of magnetics and EMI effort

The three-way trade-off in one table

Property	Silicon MOSFET	IGBT	SiC MOSFET
Voltage range	30–900 V	600–6500 V	650–1700 V
Current range	up to ~200 A per device	10–3600 A	10–100 A per device
Conduction behaviour	resistive (I²R)	fixed forward drop + resistive part	resistive, lower per area than silicon
Switching frequency	up to ~500 kHz at low voltage	2–20 kHz typical	30–200 kHz realistic
Tail current	none	yes, dominates turn-off loss	none
Body diode recovery	present, low Qrr in modern parts	present in the anti-parallel diode	low but not zero
Gate drive	10 V (4.5 V logic-level)	15 V, often bipolar	15–18 V, negative off-bias recommended
Ruggedness	high, avalanche rated	very high, short-circuit rated	lower short-circuit withstand time
Relative cost per amp	lowest below 200 V	lowest above 400 A	highest, falling

Read the table as a set of boundaries rather than a ranking. Each technology is the best answer inside a region, and the regions overlap at their edges — which is where the arguments happen.

The conduction argument

At high current, conduction loss decides. A MOSFET dissipates I² × RDS(on); an IGBT dissipates VCE(sat) × I plus a small resistive term.

Take 600 V, 200 A. A silicon super junction MOSFET rated for that current would need a very large die and several devices in parallel; the IGBT does it with a 1.7 V forward drop, dissipating 340 W at 200 A, which is manageable with a proper heatsink. A single SiC MOSFET at 200 A is possible but expensive.

Now take 600 V, 20 A. The IGBT still drops 1.7 V and dissipates 34 W. A 650 V super junction MOSFET at 190 mΩ dissipates 76 W at 20 A — worse. But at 50% duty with a freewheeling path the average conduction is lower, and once switching loss is added at 65 kHz, the IGBT’s tail current makes it lose badly. This is why 20 A offline power supplies moved from IGBT to super junction silicon, and then to SiC as frequency rose.

The switching argument

The IGBT’s weakness is tail current: after the gate is off, stored charge in the drift region must recombine, and the collector current decays over a microsecond or more. At 20 kHz that is tolerable; at 100 kHz it is the dominant loss.

SiC’s advantage is that it has no tail current and its output charge is small, so it can switch in tens of nanoseconds. The price is electromagnetic: dv/dt of 50 V/ns turns every parasitic capacitance in the layout into a current source, and every nanohenry of common-source inductance into a gate voltage error.

SiC also changes the magnetics economics. Doubling frequency halves the transformer core, and at 30–100 kHz that saving often pays for the more expensive devices. In a 3 kW telecom rectifier, the inductor and heatsink savings frequently exceed the device cost delta.

The frequency crossover, roughly

Bus voltage	Frequency	Practical choice	Reason
48 V	100–500 kHz	Silicon trench or SGT	Conduction dominates, SiC offers nothing
400 V	20–65 kHz	Super junction silicon	Switching loss still small, cost lowest
400 V	65–200 kHz	SiC MOSFET	Tail-free switching, smaller magnetics
800 V	any	SiC	Silicon 800 V parts lose on RDS(on) per area
400 V, 400 A	2–10 kHz	IGBT	Conduction dominates, cost per amp lowest
400 V, 400 A	20 kHz+	IGBT or SiC module	Depends on thermal budget; both are used

The numbers are starting points, not rules. A design with excellent cooling can justify silicon where the table says SiC, and a sealed enclosure with no airflow can justify SiC where the table says silicon — because efficiency buys thermal headroom, and thermal headroom buys reliability.

What usually decides it in practice

Three factors override the loss calculation more often than engineers expect.

Cooling cost. A 1.5% efficiency gain in a 3 kW unit is 45 W less heat. If that removes a fan, the saving is not just the fan — it is the acoustic specification, the dust filter, the service interval and the warranty exposure. SiC often pays for itself this way rather than through the device price.

EMI effort. SiC’s fast edges make conducted emissions worse, and fixing that costs filter components and board area. A design that switches at 65 kHz with silicon and passes CISPR 32 on the first attempt may be cheaper overall than a 200 kHz SiC design that needs two rounds of filter work.

Supply risk. SiC die come from a small number of foundries, and lead times move. For a product with a ten-year life, the second-source question is harder with SiC than with silicon. That is a commercial fact, not a technical one, and it belongs in the decision.

Where each technology is simply wrong

A silicon MOSFET above 900 V is possible and rarely sensible: the drift region becomes so resistive that the die is large, expensive and thermally awkward. An IGBT at 300 kHz is possible and always wrong: the tail current turns the device into a heater. A SiC MOSFET driven from a 10 V silicon gate driver will fail at turn-off, and the failure will be blamed on the device rather than the drive.

If your design sits near a boundary, describe the circuit — bus voltage, current waveform, frequency, cooling and the EMI standard you must pass — and ask for a loss comparison rather than a recommendation. We publish the device data for silicon MOSFETs and SiC MOSFETs on separate pages precisely so the comparison can be made on numbers; the glossary explains FOM, Qrr and Eoss with the practical consequence of each.

This article describes general manufacturing practice for machined and fabricated components. It is not a substitute for a drawing-specific engineering review — tolerances, material selection and process capability depend on the individual part.