Power MOSFET Selection Guide — Match the Switch to the Job

Why the MOSFET choice decides your power stage #

A power MOSFET looks like the simplest part in a power supply — three terminals, one switch. In practice it is the component that sets the efficiency ceiling, the thermal budget and often the bill-of-materials cost of any switching converter, motor driver, DC load switch or battery protection circuit. Pick the wrong one and you get a part that runs hot, needs an oversized driver, or fails under the first voltage spike. Pick the right one and the rest of the design becomes straightforward.

This guide walks through the decision path buyers and engineers actually use: start from the operating point, then trade conduction loss against switching loss, then pin down the package, thermal and grade constraints before writing the orderable code.

Start from the operating point, not the datasheet headline #

Before comparing parts, fix the numbers the application demands:

• Drain-source voltage (Vds). The rated Vds must clear the highest rail voltage plus overshoot from stray inductance and ringing. A 24 V nominal rail can spike well above 30 V during switching, so a 40 V or 60 V part is the usual comfort zone rather than a 30 V device. Budget 20–50 % headroom and add more if the layout is loose.
• Continuous and pulsed current (Id). Datasheet Id is a thermal capability tied to a specific board and case temperature — it is not a "runs forever at this current" promise. Derive the real limit from junction-temperature rise, not the headline number.
• Topology. High-side, low-side, synchronous or bridge changes which losses dominate. A low-side switch in a non-synchronous buck mostly sees conduction loss; a high-side switch in a synchronous stage sees more switching and gate-drive burden.

On-resistance vs gate charge: the core trade-off #

The single most important trade-off in MOSFET selection is between conduction loss and switching loss.

• Rds(on) drives conduction loss: roughly `I² × Rds(on) × duty cycle`. Lower Rds(on) means a larger die and usually a bigger package.
• Gate charge (Qg) drives switching loss: roughly `f_sw × Qg × V_drive` per transition. Lower Qg means a smaller, cheaper driver and less heat at high frequency.

A single die cannot minimize both at once. The useful figure of merit is `Rds(on) × Qg` (sometimes `Rds(on) × Qgd`): the lower the product, the better the part balances the two. As a rule of thumb, prioritize low Qg for high-frequency or light-load designs where switching dominates, and prioritize low Rds(on) for heavy DC or low-frequency paths where conduction dominates.

Gate drive and threshold: logic-level vs standard #

A MOSFET only delivers its quoted Rds(on) when the gate is fully enhanced. `Vgs(th)` is the threshold where the channel merely starts to conduct — it is not the drive voltage to design around. The gate must sit comfortably above threshold under worst-case drive, and that drive voltage decides the part family:

• Logic-level parts are specified for 3.3 V or 5 V gate drive and suit microcontroller-driven loads and small POL stages.
• Standard parts want 10–12 V gate drive from a dedicated driver and suit higher-power bridges.

In half-bridges, the Miller capacitance (`Qgd` / `Crss`) matters: during the drain-voltage transition the gate can be pulled back on, causing shoot-through. That is why gate-resistor tuning and a driver with strong sink current are part of the selection, not an afterthought.

Package, thermal and layout constraints #

The package is not cosmetic — it sets current capability, junction-to-ambient resistance (`RθJA`) and parasitic inductance:

• TO-220 is through-hole, easy to heatsink, and forgiving for prototype and repair, but bulky.
• DPAK / D²PAK / LFPAK / PQFN / PG-TDSON are surface-mount, offer low parasitic inductance for high-frequency switching, and rely on PCB copper for heatsinking.
• Smaller leadless packages win on board area and switching speed but lose raw current capacity.

Total dissipation is the sum of conduction and switching loss; the junction temperature must stay below the rated limit at the worst-case ambient. If the math says you need more copper or a heatsink, that is a package-and-layout decision, not something a "better" part number fixes for free.

N-channel vs P-channel and the body diode #

N-channel devices give the lowest Rds(on) for a given die and are the default for low-side and (with a bootstrap or isolated drive) high-side switches. P-channel parts simplify high-side load switches because they need no bootstrap, but they carry a penalty: for the same die, a P-channel MOSFET has higher Rds(on) than its N-channel counterpart.

Every MOSFET also has an intrinsic body diode. In synchronous rectifiers and bridge legs its reverse-recovery charge (`Qrr`) shows up as loss and noise, so Qrr becomes a selection parameter wherever the diode conducts during dead time.

Automotive, grade and lifecycle suffix checks #

For automotive or harsh-environment builds, confirm AEC-Q101 qualification rather than assuming an industrial part will survive the temperature cycling. Check the operating-temperature grade, moisture-sensitivity level (MSL) and date code for surface-mount parts — reflow of a high-MSL device without baking is a classic field-failure cause.

Second-sourcing needs care: a "pin-compatible" cross from another brand is rarely a true drop-in. Die, Rds(on) curve, Qg and body-diode behavior can all differ. Re-qualify any alternate before committing it to a BOM.

Representative examples from stock #

These three published parts show how package and family signal the use-case class. Verify the exact Rds(on), Vds and Qg on the datasheet before RFQ — the family and package tell you the job, the numbers confirm it.

Procurement risk points #

• Suffix carries the real part. Package, grade, lead-free finish, reel versus tube and automotive flag are all in the orderable code. `STP80NF55-08` is not interchangeable with a different suffix on the same base number.
• Commodity parts get cloned. Long-established series such as IRF and STP are heavily counterfeited and remarked. Match date code, lot consistency and supplier documentation, and prefer traceable stock.
• Grade and lead-time diverge. Automotive-qualified or specific-package variants can carry very different lead times from the generic equivalent.
• Verify before purchase. Confirm package, suffix, date code, MSL and traceability, then send the exact orderable code with quantity and destination in the RFQ.

Conclusion #

Selecting a power MOSFET is a sequence, not a single lookup: define the voltage and current margins the application forces, choose the Rds(on)-versus-gate-charge balance that fits your switching frequency, then let the package, thermal budget and grade (AEC-Q101 where required) eliminate the rest. Shortlist by the `Rds(on) × Qg` figure of merit, confirm the datasheet numbers and any second-source, and only then write the full orderable code. Do that, and the power stage behaves on the bench the way the simulation promised.