Acceptable mosfet temperature

[treez]

If a MOSFET in say a 60W Flyback SMPS (eg the IPP90R1K2C3 mosfet) runs for long periods regularly at 105degC junction temp, (dissipating around 3.5W) then it may ***** due to atmospheric moisture that ingressed into it getting vaporized. Is this true, or possible?

(after all, semiconductors are often shipped with dessicant bags in their package, to keep moisture out, precisely to stop them cracking during reflow soldering)

IPP90R1K2C3
https://www.infineon.com/dgdl/Infin...n.pdf?fileId=db3a30432313ff5e0123a89fe8085c04

As the following shows, manufacturers are very coy about saying what is an acceptable mosfet temperature, and the exact relationship between temp and life…
https://toshiba.semicon-storage.com/info/docget.jsp?did=13414

 

[barry]

Really? There is a pretty exact equation and explanation in that document relating life to temperature. I would suggest you contact the manufacturer for the specific parameters for your MOSFET.

 

[betwixt]

...also make sure your board assembler is following the device manufacturers guidelines on baking the devices and they are following the correct reflow profile.

In general though, moisture ingress is a very slow process, if the device is heated regularly, the chances of damage from expanding gasses is very low.

Brian.

 

[dick_freebird]

There's life, and there's life, and there's life. Many ways to
die, wearout being the longest and least likely. What you're
on about today, is more of an abuse scenario - "popcorn"
and MSL (moisture sensitivity level, which ought to be rated
in the datasheet although this alone is not that informative,
unless of course it says "hermetic, fuggedabooudit").

 

[treez]

Really? There is a pretty exact equation and explanation in that document relating life to temperature.

Thanks, yes there is, on page 8, but there are two unknowns in the expression called A and B, and these are not known.
We are a small company, and FET manufacturers never reply to us with their values for A and B.

Supposing you have to make a streetlight and the TO220 FET was at 115 degC junction...thats less than the maximum stated in the datasheet, but what lifetime woudl it give.
We wish to check becuase our contractor gives us equations, but wont state from where they got them.
They say its from their 1000's of previous superb designs, but we cannot know this, and wish to double check them.

- - - Updated - - -

Graph 1 on page 4 of the IPP90R1K2C3 datasheet (in top post) suggests that operation at 20W and 115degC junction temp is fine, but can this really be so?

- - - Updated - - -

The 4th page of the following Philips literature states that switching a product on and off many times can have a big impact on lifetime…due to the temperature cycling and the expansion/contraction to electronics components that may result….

https://www.docs.lighting.philips.c...ite_paper_LED_Driver_lifetime_reliability.pdf

We believe that with power mosfets in smps led drivers, it is this expansion/contraction that could more significantly affect FET lifetime than temperature itself (as long as that temperature is below 150degC).
Do you agree?

Basically, our competitors are ordering huge numbers of streetlights from overseas, and we feel that there should be a government checking agency to check that the mosfet temperature in the drivers is acceptable for the long product life that the taxpayer demands of these kind of items. We need 30 years lifetime minimum for motorway streelighting, because replacing these lamps places the replacement staffs lives at serious risk, due to cars travelling on the motorway at 80mph plus.
These lamps don’t have electrolytic capacitors in them, but obviously do have fets in them…so we must check the fets temperature out, and assess whether the fets really will last 30 years at whatever temperature…but where is this temperature/lifetime data for TO220 FETs?

 

[dick_freebird]

I would not propose any single "acceptable long term use
temp" because FET construction is widely variable; an old
HEXFET (cheap) and a new superjunction device (expensive
but superior) will have very different body fields and body
access resistances (which latter, temperature will swing a
fair bit and this is all that holds off the parasitic BJTs in the
MOSFET underbelly; the D-B leakage, including dynamics
and abnormal events, across this body resistance decides
whether the FET will enter a burnout operating region, if
no other current limiting steps in to save it.

Even a "recommended max" burnin / life test does not
fully or properly challenge the in-application robustness,
which has added electrical anomalies over time as a threat
element.

It's not out of reason to procure some competitor products
and perform your own testing (or hire it done, for a layer of
"I didn't cook the experiment, honest"), instrument up the key
components for temperature, and impose a reasonable worst
case voltage, temp and line anomalies. A year from now, you
may have information (if your competitor is solid; if you're
lucky you can pin a tail on something, sooner). You can send
a copy to any appropriate agencies and then out out a press
emission regarding observations and notifications, which will
be a nice FUD-bomb.

 

[treez]

Page 12 of the following app note from Infineon bases the calculation for specifying a heatsink for a MOSFET as using the Tj(max) as the temperature to design towards. In other words, a mosfet junction temperature of 150degC is deemed entirely acceptable. There is no declaration that operating at 150degC for long periods could affect the product lifetime.
This is in opposition to my experience of working in large electronics corporations, where mosfet junction temperature was kept below 85degC.

https://www.infineon.com/dgdl/an-1057.pdf?fileId=5546d462533600a401535591d3170fbd

Why is it that no manufacturer gives curves of mosfet lifetimes vs junction temperature?

 

[c_mitra]

(after all, semiconductors are often shipped with dessicant bags in their package, to keep moisture out, precisely to stop them cracking during reflow soldering)

.True. and .False.

Moisture is highly corrosive. But it is more corrosive in the liquid form (electrochemical corrosion). Moisture that is not allowed to condense is always acceptable.

But your idea that atmospheric moisture will be sucked into the resin bonded die is not substantiated. Why only moisture then, how about good old air?

Both tin and copper gets a surface oxide coat in presence of air; moisture that can condense will hasten the process.

The dessicant you mention (common silica gel) is a lousy dessicant; it can only absorb a little of moisture.

atmospheric moisture that ingressed into it getting vaporized. Is this true, or possible

The simple answer is therefore NO!

If the tin plating is exposed to condensing moisture, the oxide film will grow and you will have nightmares while soldering.

To be cost effective, you need to run you devices at the highest recommended temp ( perhaps with a few degrees margin).

Device failure is a statistical random process that depends on too many factors.

 

[treez]

Thanks, i've never designed a SMPS with a FET running at above 120degC.....does anybody know what is the long term outcome for such a FET?
(bearing in mind that the datasheet says 150degC is the max operating temperature, so 120degC is definetely below that)

 

[Nik_2213]

Slightly off-topic, but our techs would grumble unprintably about kit that ran much above body temperature 'on the bench'. Too many ways to go wrong if there was a really, really hot day, or ventilation was impeded. Murphy's Law awaits the unwary...

FWIW, my brother once said he'd had to beat off dive-bombing sea-birds who'd built a scraggly but extensive nest on top of a {Redacted} module waaay up an airport's mast. Under-floor heating, you see ? Downside was impairment of passive cooling to top, and a growing layer of guano progressively insulating the sides. Combination caused premature failure of module.

As he said, the yolk was on the guy who rudely enquired what was taking so f**g long to fix...