MOSFET on-resistance temperature behavior is one of the most consequential facts in power design. The resistance of a conducting MOSFET channel rises considerably as the junction heats, so the attractive figure on the datasheet front page, measured at room temperature, understates the conduction loss you will actually experience. Design from the hot value, or the thermal calculation is wrong from the start.
Why the resistance rises with temperature
Conduction in the MOSFET channel and in the drift region depends on carrier mobility, and mobility falls as temperature increases because lattice vibrations scatter carriers more effectively. Lower mobility means higher resistance for the same geometry and the same gate drive. The effect is not small and it is characterized in every power MOSFET datasheet as a normalized curve of on-resistance against junction temperature, usually referenced to the room-temperature value. Reading that curve at the junction temperature you expect gives a multiplier to apply to the headline number. Higher voltage devices generally show a stronger temperature dependence than low voltage ones, because more of their resistance comes from the lightly doped drift region where the effect is pronounced.
The feedback loop this creates
Conduction loss is the square of the current multiplied by the on-resistance, so as resistance rises the loss rises, which raises the junction temperature, which raises the resistance again. In a well-designed thermal system this settles at a stable operating point, but it settles somewhere hotter than a room-temperature calculation predicts and the gap widens as the thermal path worsens. A design that appears to have comfortable margin on paper can have very little in practice. The correct procedure is iterative: estimate the junction temperature, look up the resistance multiplier, recalculate the loss, recalculate the temperature, and repeat until the numbers stop moving. Then confirm the result by measuring the real assembly.
The same property makes paralleling work
A positive temperature coefficient of resistance is genuinely useful when devices share current. If two MOSFETs are connected in parallel and one takes more than its share, it dissipates more, heats up, and its resistance rises, which pushes current toward the cooler device. This negative feedback makes paralleling MOSFETs for steady conduction reasonably well behaved, unlike bipolar transistors, which have the opposite tendency and require emitter ballast resistors to avoid current hogging. The balancing is not perfect and does not extend to switching transients, where differences in gate drive routing and threshold voltage can make one device turn on first and take the entire switching loss. Paralleled devices therefore need symmetric layout and individual gate resistors even though their conduction sharing is self-correcting.
Gate drive interacts with all of this
On-resistance is specified at a stated gate-source voltage, and it rises quickly if the gate is driven with less. A device characterized at a full gate drive will have substantially higher resistance when driven from a lower logic rail, and the datasheet may provide a separate figure or curve for that condition. Driving a MOSFET partially on is a reliable route to excessive dissipation, and because a hotter device also has a lower threshold voltage, marginal drive conditions can behave unpredictably as the part warms. Always confirm the on-resistance figure corresponds to the gate voltage your driver actually delivers under load, including any drop across a gate resistor during the switching interval and any reduction in the driver’s supply.
Measuring conduction loss in a working circuit
Inferring dissipation from the drain-source voltage during conduction is possible but demands care, because the measurement must be made while the device is on, the voltage is small, and the surrounding switching produces large transients. Probes need adequate bandwidth and short ground connections, and the scope’s vertical resolution must resolve a small voltage while surviving a large one. An easier route is thermal: measure the case temperature at known load, calculate dissipation from the thermal resistance, and compare against the expected conduction plus switching loss. Separating conduction from switching loss can be done by varying the switching frequency at constant current, since conduction loss stays roughly constant while switching loss scales with frequency. Repeat that comparison at both temperature extremes, because the balance between the two shifts as the device warms and the resistance climbs.
What to check before you act
- Apply the temperature multiplier from the curve, not the room-temperature figure.
- Iterate the loss and temperature calculation until it converges.
- Confirm the on-resistance figure matches your actual gate drive voltage.
- Give paralleled devices symmetric layout and individual gate resistors.
Common questions
How much does on-resistance increase at high temperature?
It varies by device and voltage class, and the datasheet’s normalized curve is the only reliable source. Expect a substantial increase and read the exact multiplier for your part.
Can I parallel MOSFETs without balancing resistors?
For conduction, the positive temperature coefficient shares current reasonably well. Switching is not self-balancing, so symmetric gate drive layout and individual gate resistors remain necessary.
The practical takeaway
Start every conduction loss calculation from the on-resistance at your expected junction temperature and at your actual gate drive voltage, then iterate until the thermal answer is self-consistent. Confirm it on hardware, and rely on the positive temperature coefficient for conduction sharing while still designing the gate drive for symmetry.
References and further reading
- Department of Energy: how electronics losses turn into wasted energy and heat
- NFPA: electrical safety research and guidance on overheating equipment and fire risk
- CPSC: safety guidance for electronics and electrical products indoors
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
