SMD MOSFET thermal selection

[Simon]

As those of you that can help with this know, datasheets are bollocks interpolated with just enough information for you to work out if the part is suitable - maybe.

The first assumption I work from is that I cannot afford the ridiculous board space most data sheets put in their notes at the bottom of the page if not whole document to justify the unlikely current carrying capacities they claim.

So I generally work from the power that the package alone can dissipate and the power generated by my current given the channel resistance taking into account the working temperature I am happy for the channel to get to.

I start with the junction to ambient temperature difference and then, from the thermal resistance the power dissipation that allows. From that the maximum current I can pass that will dissipate that power given the channel resistance at that power and chosen junction temperature.

Is this about right? am I missing a trick? generally I find things run cool enough, often cooler than I thought. I suspect that ignoring any dissipation from the PCB may be making my decisions more cautious than need be but when data sheets qualify their results with statements like 6sqcm of copper attached to the drain, clearly I will not be replicating these conditions.

Sometimes my calculations come close to "a" number on the datasheet, often they are much lower than what the more fanciful datasheets predict.

[Refrigerator]

I've found any estimations go right out the window when you introduce even the slightest bit of airflow.
Max current depends on what you mean by that. If you need max pulsed current then there's a handy chart in most datasheets, which shows how much for how long you can pass. IIRC same charts also show DC (constant load) operating conditions. It all depends on the thermal mass of the die, the heat spreader, the heat sink and thermal resistance between each one of those.
Don't forget that most (if not all) FET's can only carry half their rated current at 80^oC as compared to 20^oC.

[Simon]

I'm looking purely at constant current. I see that a lot of specifications are based on transients as this makes it look better than it is. I looked at one MOSFET that claims 20 °C/W but in the notes on that "item" one note says  "Maximum under steady state conditions is 54 °C/W", So what was the 20 spec for?

https://www.vishay.com/docs/64814/si7145dp.pdf

Does the thermal resistance change at steady state compared to an unspecified transient? It actually only reduces the current capacity by 40% by my calculation but is a drop by 60% in thermal performance.

[PGPG]

I looked at one MOSFET that claims 20 °C/W but in the notes on that "item" one note says  "Maximum under steady state conditions is 54 °C/W", So what was the 20 spec for?

For someone (certainly not me), who knows that he can find 1" x 1" copper for this transistor and that he uses shorter than 10s pulses with enough long gaps to assume that transistor get back to ambient temperature before the next pulse.

[HwAoRrDk]

The first assumption I work from is that I cannot afford the ridiculous board space most data sheets put in their notes at the bottom of the page if not whole document to justify the unlikely current carrying capacities they claim.

Yeah, it's annoying when you have to hunt for the PCB mounting conditions for which the thermal figures are specified.

And then the headline figures turn out to be for something totally unrealistic, like "only when mounted on ceramic substrate with 4oz copper of 7000mm² area - anything else, your MOSFET will burn like the sun, LOL", or something irregular (for some reason always 2oz copper?).

And that's if they even state the conditions. Sometimes it'll be in a separate document ("as per our standard test configuration, see obscure document FU-42069 - good luck finding it on our terrible website!"), and sometimes never stated anywhere at all.

[Simon]

I suspect that transient specs are easy as they rely purely on the thermal mass of the device which is a known quantity. But when I have a spec that is for up to 10s well on such a small device we have been steady state for several seconds now, so I assume they are actually guessing at the effect of the thermal mass of the PCB.

Now the PCB is a totally different discussion. 4 layers is much better than 2, that is obvious from how hard it is to solder 4 layer.

[Solder_Junkie]

As those of you that can help with this know, datasheets are bollocks interpolated with just enough information for you to work out if the part is suitable - maybe.

Really? Datasheets are what designers refer to.

Have you tried contacting the manufacturer directly? In my experience, manufacturers are generally willing to help, even a hobbyist provided they are asking sensible questions. Analog Devices helped me with a design query via their forum, with a response in a couple of days. I have only bought a handful of their chips, but they don't know if I am a student who may soon be designing for a major manufacturer (I'm not).

SJ

[Simon]

Datasheets are sales adverts to a degree. The first page is total rubbish generally, the further in the more real the information is but also more cryptic. I've never had a problem with a MOSFET overheat but if I can better select one I will be able to get more current through less extreme solutions.