Why does a high PWM frequency heat the driver more?
Switching loss grows in direct proportion to frequency because energy is spent at every transition; conduction loss is frequency-independent and follows I².
At every turn-on and turn-off transition of a MOSFET, voltage and current are both high at the same time for a short moment and energy turns into heat at that instant. Since the energy spent at each transition is roughly constant, increasing the number of transitions per second — that is, the frequency — increases the switching loss in direct proportion. Conduction loss, by contrast, is independent of frequency. The heating of the driver is the sum of these two items.
The two components of the loss
- Conduction loss: The power the current flowing through the switch dissipates in the channel resistance while it is on, P = I² × RDS(on). It is unaffected by frequency but grows with the square of the current: 10 mΩ at 15 A means about 2.25 W, and doubling the current makes this loss four times larger.
- Switching loss: The loss occurring at the transition moments. A fixed amount of energy is spent at every transition; since the frequency determines how many transitions occur per second, the total loss is proportional to the frequency.
The practical consequence is this: if you hold the current constant and raise the frequency, the conduction loss stays where it is, the switching loss grows and the total heat rises. The lower the channel resistance (RDS(on)), the smaller the first item becomes; the way to shrink the second item is not resistance but frequency.
What raising the frequency gains, and what it costs
| Carrier frequency | Gain | Price |
|---|---|---|
| Low | Little switching loss, the board stays cool | Audible hum, high current ripple |
| High | Quiet running, low ripple, smoother torque | Switching loss and heat rise, radiated noise grows |
The motor hums when the carrier frequency is inside the audible band; moving the frequency above that band, to around 20 kHz, stops the sound. Current ripple also varies inversely with the frequency. So the selection is a trade-off, and how the PWM frequency is chosen falls between the noise tolerance of the application and its thermal budget.
Why the heating feeds itself
As the board temperature rises, the RDS(on) of the MOSFET increases; the increased resistance grows the conduction loss, and the growing loss raises the temperature further. That is why doing the thermal calculation with the 25 °C datasheet value is misleading; use the normalised RDS(on) together with the junction temperature curve. A closed panel, a mounting that blocks the airflow and a dusty surface run the same circuit at a completely different thermal point.
Where a frequency setting does not solve it
Trying to solve the heating with frequency alone is usually taking hold of the wrong end, because the variable the user can influence most is the current. A correctly sized motor, an adequate cable cross-section, tight terminals and a current selection with a safety margin left in it lower the heating more than a frequency setting does. Lowering the frequency has its price too: once the carrier drops into the audible band the motor hums, the current ripple grows, and on low-inductance motors that ripple produces extra heat in its own right. So there is a limit in both directions, and the application decides where the middle lies.
The second frequent mistake is to measure the board over a short test and decide. Keep the cycle running with its real load until the temperature settles, and read the body temperature at the end of the cycle, at the hottest point. If the problem is in the mounting rather than the driver, frequency changes nothing: a closed box, a heap of cable blocking the airflow and a mounting resting on a hot surface take the same board to its limit far sooner.
What to do
Measure the current first and check your safety margin against the driver's continuous current rating. If the hum does not bother anyone, do not raise the frequency unnecessarily. If you have to raise it, improve the cooling at the same time: open up the airflow, move the board away from hot surfaces and watch the temperature regularly over long cycles.
The full topic: Locked rotor and thermal protection