Silicon MOSFET vs IGBT vs SiC MOSFET vs GaN HEMT
Choose the switch that meets your circuit’s voltage, loss, thermal and protection requirements—not the technology with the most impressive headline. Silicon MOSFETs are a useful starting point for cost-sensitive, low-voltage designs; IGBTs for high-power conversion at moderate switching frequency; SiC MOSFETs for demanding high-voltage switching; and GaN HEMTs for compact, fast-switching stages.
Those are starting points, not fixed boundaries. The right choice depends on topology, current waveform, operating temperature, reverse-current path and the changes your board can support.
A power module contains semiconductor dies and internal interconnections. Both the device and its package affect electrical and thermal performance.
How do silicon MOSFETs, IGBTs, SiC and GaN compare?
The main differences are the voltage drop while conducting, energy lost at each switching transition, and behavior when current reverses. Compare exact parts only after checking that they can withstand the same real circuit stresses.
| Technology | On-state loss model | Switching considerations | Reverse-current path | What to check first |
|---|---|---|---|---|
| Silicon MOSFET | Current squared × hot on-resistance. | Gate charge, output capacitance and body-diode recovery vary substantially between families. | Channel when suitably driven; intrinsic body diode when it conducts. | Hot RDS(on), gate bias, commutation behavior and safe operating area. |
| Silicon IGBT | Average of collector current × on-state collector-emitter voltage. | Stored charge can produce a turn-off current tail and significant turn-off energy. | Usually a separate or co-packaged antiparallel diode in an inverter; verify module topology. | VCE(sat) at working current and temperature, turn-off loss, diode and fault limits. |
| SiC MOSFET | Current squared × hot on-resistance, at the specified gate bias. | No IGBT-like minority-carrier tail; fast edges make package and circuit parasitics important. | Channel and body diode. Do not assume zero body-diode recovery or zero dead-time loss. | Matched switching-energy data, gate limits, body-diode operation, layout and protection. |
| GaN HEMT | Current squared × applicable on-resistance; consider dynamic behavior. | Low charge can support fast switching, but driver, layout and architecture remain decisive. | Bare enhancement-mode GaN has no intrinsic p-n body diode. Reverse channel conduction still dissipates power. | Exact GaN architecture, drive interface, dynamic RDS(on), reverse conduction and transient limits. |
Scroll horizontally on a small screen. SiC means silicon carbide; GaN means gallium nitride. MOSFET, IGBT and HEMT describe different transistor structures.
Why do these power transistors behave differently?
Their internal structures trade conduction loss against voltage blocking and switching behavior. Understanding the mechanism helps explain why the lowest room-temperature resistance is not always the lowest-loss choice.
Silicon MOSFETs use a resistive conduction path
Once adequately driven on, a power MOSFET behaves approximately like a resistance. That makes I2R loss especially important in high-current circuits. The relevant resistance is at the intended junction temperature and gate voltage, not automatically the headline value measured at 25°C.
At higher blocking voltages, the silicon device structure becomes a more important part of the trade-off. Superjunction MOSFETs improve on-resistance, but their capacitance and body-diode characteristics still need to suit the commutation path. Do not treat every silicon MOSFET as the same switching device.
For the temperature-adjusted conduction model, see TI’s MOSFET selection discussion, Section 5.7.
A TO-220 package provides leads and a mounting tab. Its outline alone does not specify the pin assignment, electrical limits or cooling requirement.
IGBTs combine insulated-gate control with bipolar conduction
An IGBT uses gate control similar in principle to a MOS device, with bipolar conduction in the power path. Carrier injection helps conduction in high-voltage structures, but stored charge must clear during turn-off. The resulting current tail is one reason frequency can make IGBT switching losses costly.
Its on-state voltage is not a single fixed resistance. Read the collector-current curves at the relevant temperature and gate drive. A device that works well near rated load may compare differently at light load. Infineon’s loss-calculation guidance separates conduction, switching and diode losses.
An IGBT combines an insulated control gate with a bipolar power-conduction structure. Stored carriers influence its turn-off behavior.
SiC changes the high-voltage MOSFET trade-off
SiC supports a more conductive voltage-blocking region than silicon for a comparable blocking requirement. It retains MOSFET-like, unipolar conduction without the IGBT’s turn-off tail. That can reduce switching loss, but it does not make on-resistance independent of temperature.
Gate-oxide limits, package inductance and reverse operation still need attention. Infineon’s CoolSiC technical discussion explains these mechanisms for its illustrated device generation; its numerical gate-drive and ruggedness examples are not universal SiC limits.
GaN HEMTs need an architecture-specific drive approach
A GaN HEMT conducts through a high-mobility electron channel. Bare enhancement-mode devices, cascodes and integrated power stages can have different control interfaces and reverse behavior. A cascode includes a silicon transistor; an integrated stage may include the driver and protection.
Identify which structure the actual part uses before comparing gate voltage or claiming “zero reverse recovery.” TI’s GaN architecture explanation shows why a single generic GaN gate-drive rule is unreliable.
How do you compare power losses at the same operating point?
Define the circuit conditions first. For each candidate, record bus voltage, device-current waveform, switching frequency, junction temperature, gate drive, gate resistance and the commutating device. A lower number measured under easier conditions is not a fair comparison.
Use the actual current waveform for conduction loss
MOSFET or resistive HEMT channel
Pcond ≈ IRMS2 × Ron(Tj)Amperes squared × ohms = watts.Use the RMS channel current over the complete switching period. If this RMS value already includes the off-time, do not multiply by duty cycle again. Use resistance at the applicable temperature and drive condition.
IGBT with approximately constant on-current
Pcond ≈ D × Ion × VCE(sat)Duty fraction × amperes × volts = watts.Here Ion is current during the on-time, and D is the fraction of time conducting. For a varying waveform, average vCE(t)×iC(t) over the full period, using the appropriate temperature-dependent characteristic.
For GaN, also examine dynamic RDS(on): the effective on-resistance following switching or off-state stress can differ from a static measurement because of charge trapping. Ask for data relevant to voltage, temperature, duty cycle and switching mode; do not add an arbitrary “GaN penalty.” TI’s reliability white paper, Section 3, discusses this mechanism and device-specific validation.
Convert switching energy into watts at your frequency
For one turn-on and one turn-off per period, Psw ≈ fsw×(Eon + Eoff). Read the energy-test conditions, including current, voltage, gate resistance, temperature and commutation path. Check whether capacitance and diode effects are already included before adding separate loss terms.
Switching-frequency sensitivity example
Two imaginary switches each have a fixed 16 W conduction loss. Switch A loses 0.8 mJ per switching cycle; Switch B loses 0.2 mJ. The calculation is 16 + frequency in kHz × energy in mJ, in watts. A and B do not represent named technologies or products.
Enter 5 to 100 kHz; decimals are accepted. Figures are rounded to one decimal place. Both bars share a fixed 0–96 W scale.
Static example at 20 kHz. Interactive controls activate when JavaScript is available.
At 5 kHz the two-loss totals are 20 W and 17 W; at 100 kHz they are 96 W and 36 W. The example holds current, temperature and energy per cycle constant. It excludes thermal feedback, reverse-path, gate-drive, passive and auxiliary losses. It does not calculate converter efficiency or select a part.
Distinguish hard switching, soft switching and light-load operation
Hard switching can involve substantial voltage-current overlap. Zero-voltage or zero-current switching changes that loss balance, but only over the operating range where those conditions are achieved. Check startup, low load, transients and reverse operation; do not apply one favorable full-load waveform everywhere.
Gate-drive supply power, often first estimated as Qg×gate-voltage swing×fsw, is a separate budget item, not automatically heat dissipated entirely in the power transistor. Include driver quiescent power and where gate-drive energy is actually dissipated. The complete converter also has magnetic, interconnect and auxiliary losses.
When can SiC or GaN reduce power losses?
They deserve a closer comparison when switching or commutation loss is a major constraint and the design can support the required driver, layout and topology. Manufacturer examples help quantify the opportunity, but simulation and tested-board performance are different kinds of evidence.
SiC versus IGBT in an onsemi solar-boost simulation
onsemi’s December 2020 note AND90082/D compares the NXH100B120H3Q0 IGBT and NXH40B120MNQ0 SiC module designs in the same Q0 package. Table 3 reports simulated losses at 500 V / 25 A input, 800 V output, 16 kHz, 600 µH, 95°C case temperature and 5 Ω gate resistance.
| Loss | IGBT | SiC MOSFET |
|---|---|---|
| Conduction | 13.33 W | 12.17 W |
| Turn-on | 3.80 W | 3.17 W |
| Turn-off | 34.66 W | 3.06 W |
| Total switch loss | 51.79 W | 18.39 W |
Values are reproduced as published: the displayed SiC entries sum to 18.40 W, while the source lists 18.39 W total.
Most of this difference is in turn-off loss, not conduction. It is a simulation of these switches under these conditions, not a measured whole-inverter efficiency improvement. The note also discusses driver and PCB changes despite pin compatibility.
Read onsemi Tables 2–3 and design considerationsGaN in TI’s tested 4-kW totem-pole PFC stage
TI’s TIDA-010203 is a single-phase, continuous-conduction-mode totem-pole power-factor-correction reference design. Its published range is 200–277 V nominal AC input and 400 V DC output, with up to 4 kW output. TI reports peak efficiency of at least 99.1%.
“Peak” is not a guarantee at 4 kW or every input voltage. TI identifies efficiency, thermal, AC-drop, surge and conducted-EMI testing for its reference hardware. Use the design guide’s relevant operating point and test arrangement when comparing your target.
This is evidence for an engineered PFC stage with its control, sensing, layout and protection. It is not the efficiency of a complete isolated power supply, nor proof that inserting a GaN device into an existing silicon design reproduces the result.
Review TI’s design and performance scopeDesign guide TIDUEZ3B, revised April 2023.
How do reverse current and dead time affect switch selection?
An inductive current needs a path even while both half-bridge gates are off. Dead time prevents simultaneous turn-on, but the resulting reverse-conduction interval can add loss. Identify the actual conducting element at each stage.
During dead time
Current may flow through a MOSFET body diode, an IGBT’s antiparallel diode or a GaN reverse channel. The path’s voltage drop and duration determine conduction energy. A longer dead time is not automatically more efficient.
At commutation
Previously conducting diodes may have reverse-recovery charge. Output capacitances also need to charge and discharge. Inspect the energy accounting for both devices in the switching leg; “no body diode” does not eliminate capacitance-related energy.
With the reverse channel enabled
Suitable synchronous gate control can lower the reverse-path drop. Timing must still prevent shoot-through under propagation-delay, temperature and operating-point variation. Verify the actual controller, driver and device combination.
Can SiC or GaN replace an existing MOSFET or IGBT without redesign?
Not on technology, voltage rating or footprint alone. A proposed replacement must preserve the required function and survive the existing circuit’s stresses. Faster switching can change overshoot, electromagnetic interference (EMI), gate behavior and fault response.
| Review area | Evidence to compare or obtain | What keeps the alternate on hold |
|---|---|---|
| Package and gate drive | Full MPN, pinout, Kelvin connections, recommended on/off bias, absolute gate limits, driver current and isolated-driver common-mode transient immunity. | Unverified pin mapping, gate overshoot, incompatible drive interface or an unsupported negative-bias assumption. |
| Switching loop and timing | Gate and power-loop layout, local decoupling, dead time, measured overshoot and unintended turn-on across the operating range. | Voltage or gate stress outside the approved limits, or a timing change not reviewed in both operating directions. |
| Protection and robustness | Applicable safe operating area, overload limits, fault detection and shutdown time, plus any explicitly specified short-circuit or avalanche capability. | Reliance on the previous part’s ruggedness or a technology-wide “SiC/GaN is tougher” assumption. |
| Thermal and product qualification | Loss and temperature review, mounting or PCB thermal path, relevant EMI tests, assembly requirements and recorded engineering approval. | A better typical datasheet value without evidence for the finished design and its required qualification scope. |
Qualified engineers should use measurement methods suitable for the circuit voltage, edge speed and energy. Probe artifacts can resemble real overshoot; an inadequate measurement cannot establish safe operation.
Which technology should you shortlist for your application?
Start with the constraint that matters to your design: conduction loss, switching loss, cooling, size, cost or compatibility. The following engineering starting points are not manufacturer selection guarantees.
| Application or constraint | Useful starting comparison | Question that decides the next step |
|---|---|---|
| Low-voltage battery conversion or high-current switching | Silicon MOSFET first; include GaN if switching speed or size can deliver a measurable benefit. | Does hot conduction loss dominate, and can the board support the proposed package and transients? |
| Industrial inverter at moderate switching frequency | IGBT versus SiC MOSFET, with the correct freewheel path. | Do turn-off losses, load profile or cooling justify a different switch and driver? |
| Mains PFC or charging front end | Silicon MOSFET, SiC or GaN according to topology and voltage stress. | What happens during commutation, surge, light load and EMI testing? |
| Compact high-frequency DC/DC stage | GaN versus suitable silicon or SiC parts at the required voltage class. | Does the frequency increase actually reduce system volume after magnetic, drive and thermal changes? |
| Existing qualified board with limited redesign time | Exact approved device or a documented engineering-approved alternate. | Is there time and evidence to complete the necessary validation before production? |
Compare the load profile, not only full-load efficiency
A converter that spends most of its life at partial load can have a different optimum from one operating continuously near rated power. Compare losses across representative load, line voltage, temperature and operating mode; then weight them by expected time in service.
As a planning calculation, energy lost over a duty profile is the sum of loss in watts at each operating point multiplied by hours spent there. Keep standby and auxiliary consumption in the same system boundary. A peak-efficiency number alone cannot answer this operating-cost question.
How do cooling and total system cost change the final choice?
A lower switch-loss estimate is useful only if the junction remains within its approved temperature range in the actual assembly. Case, PCB, heatsink and ambient temperature are different quantities.
Tj ≈ Tc + Pdevice × RθJCSteady-state estimate using case temperature, device dissipation and the applicable junction-to-case thermal resistance.Do not substitute ambient temperature for Tc. Use the defined case measurement point and a matching thermal model. Pulsed loading requires transient thermal impedance; multi-die modules can also require thermal-coupling analysis.
Illustrative example: with a measured case temperature of 90°C, 30 W device loss and an assumed RθJC of 0.5°C/W, the estimate is 105°C. This is not a prediction for any device pictured here. Because losses can change with temperature, iterate the loss and thermal estimates rather than stopping at one cold calculation.
A power module needs a specified mounting and thermal interface. Package photographs cannot establish junction temperature, losses or qualification.
Compare the complete bill of materials and engineering effort: switch, driver and bias supply, cooling, magnetics, decoupling and EMI components, insulation, assembly and qualification. A more expensive transistor can enable a smaller system, but only if the design realizes those savings. Faster switching can also increase other losses or validation work.
The final choice is an approved device in an approved design. Record the conditions, unresolved risks and validation owner. High-voltage and high-energy converter development requires qualified personnel, suitable equipment and controlled test procedures.
Need to source a defined power-switch requirement?
YURUNOX is an independent electronic-component sourcing partner, not the manufacturer of these devices. Share the approved full part number, quantity and delivery needs. If you are evaluating an alternate, identify the open engineering questions separately.
- Full manufacturer part number and package
- Quantity, required date and destination
- Exact-match or approved-alternate policy
- Packing, condition and evidence requirements
RFQ text to adapt to your BOM
Please quote the full manufacturer part number, including package and grade suffixes, quantity, required delivery date and destination. Confirm packing, condition, lot policy and the source documentation available. List any alternate separately with its full code and official supporting documents. Do not substitute without our written engineering approval. For an engineering comparison, use our stated topology, bus-voltage range and transients, current/load profile, switching frequency and mode, gate drive, cooling arrangement, protection and qualification constraints. Identify any missing information before recommending a candidate.
You can also select and copy the text directly.
Review quality expectations, the purchase process and shipment information when defining the commercial requirement.
Technical references and scope
Technical sources reviewed on September 6, 2026. Numerical examples without a named manufacturer are illustrative calculations, not test results. Historical application notes remain limited to their described devices and conditions; recheck the current datasheet for every candidate.
- TI SLUA865B, Section 5.7 — MOSFET RMS current and temperature-dependent conduction-loss calculation.
- Infineon: IGBT loss calculation — current, temperature, switching energy and diode-loss terms.
- Infineon: high-performance SiC MOSFET technology — device structure, hot resistance, reverse operation and packaging.
- onsemi AND90082/D, December 2020 — the cited boost-stage simulation and associated design considerations.
- TI TIDA-010203 and TIDUEZ3B design guide — the defined GaN PFC reference hardware and reported testing.
- TI SSZT092: GaN architectures — differences between discrete, cascode and integrated drive approaches.
- EPC: GaN transistor fundamentals — reverse operation and capacitance-related considerations.
- TI SNOAA68, Section 3 — dynamic on-resistance, stress conditions and device-specific reliability evidence.
Credited photographs and manufacturer references do not imply endorsement, current inventory or approval of a particular replacement.