Can I power a MOSFET from my uC in this way?

Question

The uC is a PIC10F220, which if I read the datasheet correctly can supply 25mA absolute maximum on its pins. The MOSFET that drives the load is actually 4 pcs NCE0140KA in parallell, but drawn as one for simplicity. The combined capacitance of the MOSFET's is 13.6 nF. I have selected the values of R1 and R2 so that the transistor gate will charge and discharge as quickly as possible, in order to minimize the transition time when the the transistors will be half-open. I am doing this because this circuit will be potted, so I'm trying to minimize heat dissipation.

By my calculations the current drawn from the uC output should peak at around 20mA the moment the output is activated and stabilize at 10mA. The peak will be close to the absolute maximum of the pin, but I figure it will be OK since the peak will be very short: Time constant of the R1-M1 RC circuit is 250 * 13.6e-9 = 3.4 us.

The circuit will be used as a pwm control and the switching frequency will be around 2 kHz.

Have I got this right? What would you do differently?

Answer 1

You're (worse than- because of voltage drop in the output) halving the gate voltage, the particular MOSFET you mention is only guaranteed as low as 4.5V Vgs, so you're going to get more (perhaps much more, and relatively unpredictably) Rds(on) and thus I^2R heating during conduction than if you used the full 5V drive.

Secondly, you need to consider the gate charge to determine the switching time with a load, not just input capacitance because of Miller effect- your numbers will be optimistic by a large factor because the gate-drain capacitance couples the drain voltage swing back to the gate at a multiple because the drain voltage changes more than the gate voltage. From the link above:

There are many papers on predicting switching times, or just get a SPICE model and look at "typical" results to get a feel for it (keeping in mind that there will be a spread in parameters in reality- not all devices are "typical").

The MCU naturally drives the output push-pull- you're better off driving it that way with the full 5V (series resistor) and adding a driver circuit if the switching is not fast enough to meet your requirements. A (very) rough simulation for the MCU output with a 5V supply is a voltage source in series with a ~100 ohm resistor (maybe 200 ohms for sourcing current vs. 100 for sinking). If it turns out that guesstimate resistor value is fairly important, your circuit may be too dependent on characteristics that are not guaranteed (in reality the outputs act sort of like a constant current source over a range, and that current is not well controlled).

Here's an old Microchip output curve, the part you are using will likely be different, but the shape will be similar:

Edit: Suggested simple gate driver. You can also buy gate driver chips with much higher gate drive, and several manufacturers have many products (including Microchip), but for 2kHz it should not be necessary. You can optionally put a 1K resistor from MCU output to gate to get the tail of the 0/5V.

^{simulate this circuit – Schematic created using CircuitLab}