In a synchronous buck converter, the hotter MOSFET is the one with the larger loss multiplied by its effective thermal resistance. At low duty cycle the low-side switch conducts for most of the period, while the high-side switch often dissipates more energy during transitions. Compare the two gate waveforms, current intervals and copper temperatures before selecting a replacement.
Estimate conduction loss on each side
For continuous conduction with triangular inductor ripple, use I² ≈ IOUT² + ΔIL²/12, where IOUT and ΔIL are average and peak-to-peak inductor current in amperes. Multiply I² by the hot on-resistance in ohms and by the fraction of the period when that channel conducts. With dead time neglected, that fraction is D for the upper device and 1 − D for the lower one. D is approximately VOUT/VIN in the ideal buck.
For a 12 V to 1.2 V rail at 15 A, assume D = 0.10, small ripple and 4 mΩ on both switches at their operating temperatures. The upper channel dissipates roughly 0.09 W, and the lower channel roughly 0.81 W. These numbers cover channel conduction only. Recalculate with the actual on-resistance at the applied gate voltage and the maximum expected junction temperature. The buck MOSFET selection guide gives the broader part-selection tradeoffs.
Check what changes as frequency rises
High-side turn-on can commutate the low-side body diode and discharge capacitances at the switch node. The upper switch may experience voltage and current overlap on both edges. A rough overlap estimate is VIN × IL × (tr + tf) × fSW/2 when slopes are approximately linear, with time in seconds and frequency in hertz. With 12 V, 15 A, 20 ns total transition time and 500 kHz, the estimate is 0.9 W. This already changes the predicted temperature ranking in the example.
The low-side device may turn on near zero drain voltage after its body diode has started conducting. Its diode still dissipates energy during dead time, and recovery can add loss elsewhere in the bridge. At light load the converter may skip pulses or allow current reversal. Use the controller’s observed operating mode rather than applying the continuous-conduction estimate at every load. Review synchronous rectifier dead-time adjustment before changing the timing.
Use waveforms to explain an outlier
Measure each VGS directly between the corresponding gate and source. For the high side use an appropriately rated differential or isolated measurement with adequate common-mode performance. Measure the switch node and inductor current at the same operating point. A sagging high-side drive near the end of long on-times can indicate bootstrap supply trouble. A gate bump while the opposite switch turns on calls for a check of Miller injection, return inductance and shoot-through.
Run a small load sweep at fixed input, then a line sweep at fixed load. A large temperature rise with current points toward channel or copper loss. A rise with frequency at comparable current points toward transition, drive or commutation loss. Compare the two gate transitions with actual current waveforms. If using scope math to calculate device energy, deskew voltage and current channels first.
Compare cooling as well as power
Thermal images show surface temperature, which depends on emissivity and airflow. A nearby inductor can warm one device even when it dissipates less power. Compare equal package locations after thermal stabilization and inspect the drain copper, vias and exposed-pad solder. Small switch-node copper may be intentional for EMI, so review the electrical tradeoff before enlarging it. The MOSFET PCB thermal path guide explains how board construction affects the comparison.
Record input, output, switching frequency, load, ambient and each measured case temperature. Use the resulting loss budget to compare power MOSFET options at the real gate drive. A lower-resistance part is most useful where conduction dominates, and a lower-charge part may be more effective where transitions dominate.
