My IRL540N is smoking

Question

I am using an IRL540N MOSFET to drive a 12 V pump (about 6 A) and a 12 V solenoid valve (about 1 A max). It's smoking after just two uses. I need to know two things.

1. Is it still alright? It's still working and I have switched off power. When do I know it's best to replace it?

2. How do I prevent the smoking? It's getting too hot. What's a good heat sink for such applications? I'm surprised it got so hot as it's winter and the ambient temperature is close to 12-13°C. If it's in this condition in this weather, then when summer brings 45°C it will just go nuclear. I'm scared.

Below is my schematic.

Answer 1

When do I know it's best to replace it?

After it smoked, it's necessary to replace it. See the "Absolute Maximum Ratings", first page of the datasheet. You've clearly exceeded that maximum temperature.

Physically, failure modes of overheated semiconductor devices involved migration of dopants, degradation of insulation, degradation of thermal coupling, … So, things you don't easily see. An IRL540N is between 1 and 2€ in single quantities – so definitely not worth risking it.

How do I prevent the smoking? It's getting too hot.

It would seem that way, yes! So, you need to stop it from getting so hot.

What's a good heat sink for such applications? I'm surprised it got so hot as it's winter and the ambient temperature is close to 12-13°C.

You didn't install it on a heatsink to begin with? I'm a bit surprised: The Junction-to-Ambient thermal resistance from the datasheet says 62 °C/W, so, with a loss of only 3 W, you'd be far into "beyond repair" territory.

Anyways, I'm not sure why you have that loss. The currents you're switching are low, and at your (starting) temperature and this current, the on-resistance should be far below 1 Ω. In fact, looking at Fig 1. from the datasheet, at 7A your Drain-Source voltage would be < 0.3V for any reasonable Gate voltage, so you'd be converting 0.3 V · 7 A = 2.1 W into heat. Sure enough, that's a lot, but according to the datasheet thermal resistance should heat up your case to little more than 120°C hotter than ambient – not smoking hot.

So, three things are possible: you're using a gate control voltage that's too low, the current is higher than you say, or you're swithing the MOSFET on and off a lot – but these are motors, so I'm pretty sure you wouldn't want to PWM them?

In any case, when you have something that produces ~ 2W of heat, yeah, a heatsink is desirable – but a cheap one should do. All heatsinks will come with some specified thermal resistance. That, added to the thermal resistance from the MOSFET datasheet from junction to cooling surface, will tell you how hot the chip gets when you generate some defined heat power.

Answer 2

At 7A, and assuming you are supplying at least 5V Vgs (Which is where the on state resistance is specified), I make the power dissipation about 5 watts, so you will want a small heatsink to keep the thing happy, bolting it to the side of a metal case should suffic.

Note that a common trap for beginners is looking at Vgs(th) and thinking that is all the gate drive you need, when it is actually the point at which the thing just starts to conduct a tiny bit (250uA in this case), to get the mosfet hard on you need typically double whatever Vgs(th) is, details will be in the graphs.

If you try to switch this with 3.3V logic it will absolutely smoke.

There are much better mosfets for this that will run with no heatsink.

The front page of mosfet datasheets doesn't usually lie exactly, in the same sense that a government minister addressing parliament doesn't usually lie exactly...
They are however heavily influenced by marketing's desire to publish the best possible numbers, and particularly for things like maximum drain current a smart designer will go to the graphs instead.

Answer 3

Look out also for temperature coefficient of Ron. If you get a MOSFET whose Ron increases with temperature and you don't drive it hard enough on or give enough cooling, this can cause magic smoke very quickly. The Ron you see on the graph on the datasheet rises, the higher it goes the more power it tries to lose ...

Basically Ron, tempco of Ron and thermal resistance form a complex system which needs to remain in a stable state. Being conservative about calculations and generous with heatsinking goes a long way to achieving this.

This app note gives a decent overview. In the past when I did this, I always did some prototyping, making sure to try a few different devices and measure temperature rise on the device body at maximum current (plus some error margin) - I feel that real world tests go a long way to confirming whatever paper calculations or simulations you came up with, to make sure you didn't miss anything important. This is not stuff you want to make an error with (especially in equipment which might be left running unattended, which is most equipment).

Also don't forget to be pessimistic about ambient temperature - especially if you are designing gear which might will run in a rack, with many devices all conspiring to push "local" ambient temperature up above what you expect.