Which isn't what they're seeing, because the cap is directly across the MOSFET.insta wrote:respond to variable input voltage
Time for some back of the envelope numbers.
When the MOSFET turns on, it connects the + side of the cap to ground. That puts a dead short across the cap, so the MOSFET must absorb all the energy stored in the cap in one brief pulse. However, the fan sees the full supply, because the MOSFET has connected the black fan wire to ground, with the +24 V (or whatever) supply on the red fan wire.
When the MOSFET turns off, the cap voltage rises (and the fan voltage decreases) as the fan current passes through it. The 24 V fans on my (heavily modded) M2 say "DC Brushless 24 V 0.09 A", so two of them in parallel initially put 180 mA into the cap.
In round numbers, 180 mA charges a 680 uF cap at 260 V/s. IIRC, Marlin runs the fan PWM at 500-ish Hz, so at 50% PWM it's off for 1 ms, during which time the voltage changes by 260 mV. Even with 10% PWM, it'll rise maybe half a volt.
So good news: the cap voltage can't rise much while the MOSFET is off, which means the MOSFET will short a fully charged cap only when the fans start up. The rest of the time, it must handle only a fraction of a volt, which is survivable.
But bad news: the cap defeats the entire purpose of the exercise, because the fans see nearly the full supply voltage all the time, because the full range of PWM values changes the fan voltage by 2%.
The IRLML6346TR fan MOSFETs live in little bitty SOT32 packages and have an on-state resistance around 0.06 ohm, so the initial current from a fully charged cap is 400 A = 24 V / 0.06 ohm. I'd expect the wiring & connector resistance to reduce that, but for a total resistance of 0.2 ohm, the peak current is still 120 A.A 680uF isn't going to blow up a mosfet
Their continuous current rating is 3.4 A and their absolute maximum pulsed current rating is 17 A.
You treat 'em your way, I'll treat 'em my way ... [grin]