MOSFET Fried, how can i tell why?

[Chet T16]

As i have mentioned elsewhere i was making a fuel injector pulser based around an IRFZ34N and today i gave it a full test and it fried, d'oh.

The connected load was 4, 3.5 ohm injectors in parallel, powered from a car battery with 13V.

From the mosfet datasheet it says that the Rds is 0.04ohms which should mean only a tiny amount of power being dissipated (around 8W) so i hadn't even considered a heatsink or anything

It worked for a while, 30-45 seconds than let out the magic smoke without any pop bang or fizz and the mosfet is now short circuited.

I was using a suitable flyback diode across the load

Everything else in the circuit works fine still

Any ideas where i've gone wrong?

(Schematic coming as soon as i figure out how to make one)

Chet

[Rufus]

From the mosfet datasheet it says that the Rds is 0.04ohms which should mean only a tiny amount of power being dissipated (around 8W) so i hadn't even considered a heatsink or anything

8W is not tiny. A junction to ambient thermal resistance value is given in the datasheet you linked.

62 C/W meaning that with no heatsink the device would have to run at almost 500C to dissipate 8W.

[Simon]

yes you need a heatsink. Also if you are turning the mosfet on and off quickly this will make more heat because the mosfet will not always have a resistance of 0.04 ohms as it will take some time to turn off during which the resistance is higher and so more heat is made.

[Zero999]

I make it closer to 9W, 8.83W to be precise.

Yes, that's to be expected, I don't see why you're surprised, you exceeded the maximum operating junction temperature of 175^oC by around 400^oC. I suggest reading the datasheet again, doing the temperature rise calculations and using a suitable heat sink.

What's the gate source voltage?

[Neilm]

yes you need a heatsink. Also if you are turning the mosfet on and off quickly this will make more heat because the mosfet will not always have a resistance of 0.04 ohms as it will take some time to turn off during which the resistance is higher and so more heat is made.

Switching losses are pretty hard to calculate correctly. One explanation I found to calculate this was in a TI datasheet http://focus.ti.com/lit/ds/symlink/bq24600.pdf If you go down to page 23 you will find some data on how to calcultate switching losses. This is tailored for the application in the datasheet but you should be able to use it to approximate the losses you will be experiencing in your application. I have used these formula to calculate how hot MOSFETs were going to get with a reasonable degree of accuracy.

Yours

Neil

[Simon]

well I just tend to "judge by eye" although I've never done any complex stuff. I remember building a 555 based PWM speed controller for a car blower (8A motor) and added an optional capacitor in parallel with the timing capacitor to reduce the frequency. The difference in temperature of the mosfet was noticeable even though on a heatsink being cooled by the fan itself. I learnt a valuable lesson early on and thankfully have never burnt a mosfet out

[Chet T16]

Clearly i f'd up here. I think this may be (in part) because i was originally intending to drive only one injector so the load was much less and the specs for that mosfet might have been confused with the curent one along the way. Vgs is 5v which i believe would have lead to a higher value for Rds

I came across this (available from farnell) which looks like it would work much better with a small heatsink. The ones i fried came from ebay in some questionable packaging so i'm not 100% certain they are what they say they are

[jahonen]

Vgs is 5v which i believe would have lead to a higher value for Rds

I think this is also a big problem. Consider using a logic-level MOSFET, which achieves specified Rds(on) with 5 volts, or, gate driver circuit which raises the gate-source voltage to appropriate levels.

Regards,
Janne

[Simon]

yes logic level mosfet just to be sure. I got away with plenty with IRF540 devices run from a 5V pic but for any real power I'd use the equivalent logic level mosfet by the similar name of IRL540

From a personal point of view, the packaging don't always mean much, I've sold mosfets I bought from major uk suppliers and mosfets bought from china in ziplock bags and it don't mean they were bad and i've had no complaints, I think you just abused the mosfets you bought.

Put a heat sink on and try again. What you need to remember is that with a heat sink the heat will gradually dissipate away. With no heatsink it builds up fast, even that little extra dissipation area helps alot - a lot like a car radiator (I work for a radiator company) the in and out temps are only a few degrees apart, but taking those few degrees off at a time is enough to maintain a stable temperature

[Chet T16]

Oh yeah, they were certainly abused and the failure was totally my fault!

I'll retry with a heatsink and see how it goes. I haven't any new ones but i seen a couple on an old board somewhere

[Zero999]

Vgs is 5v which i believe would have lead to a higher value for Rds

I had a feeling you were doing that but no one knew because you didn't post a schematic.

The IRL540 springs to mind but there are probably better, more cost effective MOSFETs available. Going from the worst case, you should use a heat sink with a thermal resistance of 4^oC/W or less.
http://www.vishay.com/docs/91300/91300.pdf

[Simon]

so does Rds get lower for Vgs that is higher than 4 or 5V ? I hope so because the datasheet for MY irl540 states 44mR so i assumewith something like 10V

[Zero999]

so does Rds get lower for Vgs that is higher than 4 or 5V ?

What do you think?

I hope so because the datasheet for MY irl540 states 44mR so i assumewith something like 10V

Are you sure it isn't the IRF540? It's a similar MOSFET but with a lower on resistance and is not logic level so has a higher threshold voltage.
http://www.irf.com/product-info/datasheets/data/irf540n.pdf

[Simon]

not sure, but then I have seen a few variants of the IR*540 family, adding an N to the end often produces a part that can handle more amps, it could be I'm thinking of the resistance spec for that variant although any of the two vary slightly from one manufacturer to the next

[Zero999]

Yes, the IRF540N seems to be a newer, updated version - manufacturers often do that.