OT - MOSFET choice

[Scampy]

Thank you for the explanation, and the link. At that price it's worth getting a pack and testing them on the breadboard

Cheers

[HenrikOlsson]

In switching applications there are more to it than just the Rds-on.

The gate of the MOSFET looks like a capacitor to the circuit driving it. To turn the MOSFET on you need to "charge" the gate. The longer it takes to charge the gate the more switching loss you'll have, the MOSFET is in the linear region for a longer period of time.

To charge the gate quickly you want a MOSFET with low gate charge and/or a driver that can charge/discharge it quickly - that is pump current into and out of the gate. You're limitied to 10mA current at 5V which really isn't much in the scheme of things.

If you want it efficient, don't use a PWM frequency higher than needed and look for a MOSFET with low gate charge (small "capacitor"). Usually, the "larger" the MOSFET is the higher the gate charge is so don't use a 100A MOSFET for a 3A load just because the 100A one has 5mOhm less Rds-on.

/Henrik.

[Scampy]

I have knocked up an output circuit that works, but may not be ideal. All I had in the way of MOSFETs in my component box was an IRF3708.

I've used a BC639 to drive the FET using the attached schematic. I've tested this with a six bulb 1/12th chandelier which drew just over 310ma, and a twenty-one bulb light that drew 1100mA and it worked really well... OK the additional 16 transistors and 48 resistors is not much to accommodate in the design, but it would be better for a single component.

As each channel could end up driving 60 x 50mA bulbs, I've rated rated the device at 3 to 4 amps, but as most of these MOSFETs seem to be rated at 30Amp - 600A or more I don't think I'm going to be taxing them too much. Is it worth looking at these multi-fet devices in SOIC / DIP packages, or would they need cooling ? That TO 292 package that Gusse linked to looks an alternative if no DIP device is suitable.

[Gusse]

Hi Scampy,

Based on IRF3708 datasheet, Rdson = 9.5mOhm (typ), Vgs= 4.5V, Id= 12A, therefore PCA9865 should be OK to drive FET without pre-driver stage. Gate resistor might need fine tuning. Ten(s) of ohm might be OK. Resistance on gate and gate input capacitance (~2-3nF), will make a RC-filter.

Gate capacitance is important parameter as Henrik stated. I didn't include it to my previous post due to fact that PCA9865 max output frequency is low (1526Hz).
IRF3708 applications are high freq DCDC (~3-6MHz switching freq), so there is no need to worry about that parameter too much.

With multi-FET packages it is more important to keep Rdson low if all FETs are on same time. It doesn't matter if it is through hole or surface mounted device, same rules applies.
Reserve enough PCB area for GND under the FET to do proper cooling.

[HenrikOlsson]

I didn't include it to my previous post due to fact that PCA9865 max output frequency is low (1526Hz).

Ah, yes at such low frequency it's probably not a concern.

IRF3708 applications are high freq DCDC (~3-6MHz switching freq), so there is no need to worry about that parameter too much.

Yes, but not with 10mA gate drive - which is my whole point. The total gate charge for IRF3780 is 24nC so if my math isn't failing me it takes 2.4us to charge/discharge it at a constant 10mA charge rate. Driven by a PCA9865 the current into the gate isn't constant of course so in reality it'll take longer than 2.4us. Trying to switch it at even an low(ish) 20kHz would make it spend a considerable portion of the switching cycle in the linear region. At 1500Hz Gusse is most likely correct and it's not a concern.

/Henrik.

[Scampy]

LOL - a lot of what you guys have discussed went right over my head

However it gave me the confidence to remove the BC639 and take the output directly to the gate of the MOSFET and give it a try. Although I kept the series resistor and gate to GND resistor in place, it all worked fine

Thanks guys

[Gusse]

Yes, but not with 10mA gate drive - which is my whole point. The total gate charge for IRF3780 is 24nC so if my math isn't failing me it takes 2.4us to charge/discharge it at a constant 10mA charge rate. Driven by a PCA9865 the current into the gate isn't constant of course so in reality it'll take longer than 2.4us. Trying to switch it at even an low(ish) 20kHz would make it spend a considerable portion of the switching cycle in the linear region. At 1500Hz Gusse is most likely correct and it's not a concern.

You are right, 10mA pre-driver is weak even for this purpose, but still usable. Any application using higher switching frequency would compromise efficiency.

Scampy, charge calculation for future applications:
C = A*s -> s=24nC/10mA=2.4us

C = Coulomb
A = Ampere
s = second

However it gave me the confidence to remove the BC639 and take the output directly to the gate of the MOSFET and give it a try. Although I kept the series resistor and gate to GND resistor in place, it all worked fine

If PWM source is able to source 10mA @ 5V then it means that internal resistance of source is 500 Ohm (R=U/I).
Therefore you should adjust external components according to this information. Use much smaller serial resistor and remove/increase GND resistor a lot.