MOSFET junction temperature in SMPS?

[Faringdon]

Hi,
Do we at least agree that a MOSFET in an offline SMPS should preferably not run with junction any hotter than 105degC?......otherwise it will succunmb fairly quickly, to expansion/contraction damage which is likely to result in significantly shortened lifetime.?

BTW, that above comment is what you hear from "old-timer" engineers, all over the place...you want see it in MOSFET literature.....its "in_the_club_knowledge"....a Trade secret.

[Psi]

Are you making the assumption to remove silicon carbide MOSFETs from this discussion?

[Faringdon]

Thanks, no i mean SiC aswell.

[jonpaul]

yet another FTTS
nonsense questions as always
read the ancient and plentiful papers , books and app notes on thermal management of power devices.

The Hotspot temps are in the device datasheets, with all thermal résistances jct>> Mt, amb, etc.

Calculations take perhaps a minute

Use your brain and eyes not ours!!!!!!

j

[Faringdon]

Thanks, as you now, the MOSFET datasheets and App Notes give the Abs Max Tjunction, ..... can be anywhere between 125degC and 175degC.
But allowing that will mean a short lived SMPS.
It is known (but not written publicly anywhere), that you should not run a MOSFET junction above 105degC if you want a reasonably long lasting
SMPS.

As you know, we simply cannot make  MOSFETS physically rigid and rugged and robust  like the stantion supports in big buildings or big bridges...the MOSFET internal structures, are, of necessity, highly delicate structures, and repeated expansion/contraction will kill them prematurely. The interfaces between the fine structures of a MOSFET are just too fine.....they will be busted by repeated expansion/contraction....As i discuss, you are never going to build the inner support structure of a skyscraper, etc etc,  with "MOSFET material".

The 105degC figure, is the "in_the_club" figure, that you hear over and over.
It makes sense too...those internal structures are extremely fine and delicate....necessarily so.

As i discuss, there are no public books or websites  that will tell you this...it is just an "in_the_club" fact.
There is no way  that the "engineer_in_the_street" can calculate it either.

[Gyro]

Cut the "in the club" crap treez!

The sort of figures you are quoting are Absolute Maximum Ratings. Datasheets contain a disclaimer with wording along the lines of:

CAUTION: Stresses above those listed in “Absolute Maximum Ratings” may cause permanent damage to the device. This is a stress only rating and operation
of the device at these or any other conditions above those indicated in the operational sections of this specification is not implied.

You need to derate the temperature and other parameters for reliability, just like any other component - you sound (as always) like you've never designed anything before!

[jonpaul]

105C.... no secret not a rule.
Check spec sheet for device.
SiC, Gas, Si differ in absolute max Tj
Reliability...see Arrhenius  1889
https://en.m.wikipedia.org/wiki/Arrhenius_equation


We have lost another 5 min down Farringtons toilet bowl of time.


FTTS where can I send you my invoice?

Visa, MC, Amex card accepted....

Cheers!

Jon

PS Why are the British so eccentric....(:-:)

[temperance]

You've asked the same question in 2018. Don't tell me you didn't find an appropriate answer in a span of 4 years.

[Siwastaja]

You need to derate the temperature and other parameters for reliability, just like any other component - you sound (as always) like you've never designed anything before!

That's literally and exactly what he wrote, numerous times. Why do you bother replying to a message you did not read at all? If treez posting triggers you so badly, consider using the ignore feature.

The question itself is valid - how much derating is needed?

I would not agree that 105degC is some kind of hard limit. Many MOSFETs are rated at Tj=175degC absolute max and actually characterized at +125degC. I would not hesitate to use them at +125degC, but of course then I need to be pretty sure I'm not significantly exceeding that - with careful thermal modelling using worst-case values.

But if you have to choose some magic number for a design limit, then 105 is not bad.

[Gyro]

@Siwastaja, you are of course correct. I have become overly sensitized to someone who keeps telling us 'what we all agree on', 'what we all know' and generally treats basics that the rest of us learned in our first jobs (if not before) as mystical secrets posessed only by some masonic inner circle, from which he is deliberately and persistently excluded. I have taken your advice.

The degree of derating of components is of course dependent on many factors, including operating environment, required MTBF, market sector (Mil?), frequency of thermal cycling etc. even house rules, and isn't a 'magic' number.

[ejeffrey]

As you know, we simply cannot make  MOSFETS physically rigid and rugged and robust  like the stantion supports in big buildings or big bridges...the MOSFET internal structures, are, of necessity, highly delicate structures, and repeated expansion/contraction will kill them prematurely.

Thermal expansion is not a big problem related to junction temperature.  The thermal expansion of the silicon is fairly small and the the die itself is fairly rugged, and should be mounted to the package in a way that allows the needed thermal expansion.  It's a bigger problem for package to pcb interface and the pcb itself than the silicon.

Thermal stress is also much worse with cycling.  So a device that cycles between 25 and 125 C 500 times per day will have a much different effect than one that sees steady load for 8 hours a day

Derating of Tj is done first because often performance is specified more at lower temperature.  Check the data sheet, and if the parameters you care about are defined at 85C that's probably where you should operate.

Derating also of course lowers temperature dependent degradation, but it also provides some margin against unanticipated effects such as operation at higher ambient temperature, atmospheric pressure, clogged fan filter, not fully accounted for heat generated by other parts of the system, manufacturing variation in connection to the heat sink and so on.  If you design you system for 105 C when the max is 125 C you are less likely to go over 125 without anticipating it.  But if Tj max is 175 you can certainly go well over 105 C.

You may also not be limited by the junction temperature at all.  If you take a system without inadequate heat sinking, a pcb that isn't designed for high temperature, and lots of plastic components and electrolytic capacitors, then replace the MOSFETs with high temperature SiC devices you may not be able to practically exceed Tj of 105 C without breaking something else. 

105 is not the worst generic rule you could pick as a target where you don't need to do too much modeling and testing.  But you can certainly design reliable power supplies that operate above that, even at higher ambient temperatures.

[Faringdon]

Thermal stress is also much worse with cycling.  So a device that cycles between 25 and 125 C 500 times per day will have a much different effect than one that sees steady load for 8 hours a day

Thanks, this is  a great point.

As the below philips literature says, thermal cycling is indeed a major failure point.
https://www.docs.lighting.philips.com/en_gb/oem/download/xitanium-led-drivers-outdoor/White_paper_LED_Driver_lifetime_reliability.pdf
(...theyre talking about LED modules, but the same thing applies to FETs)

This is indeed it..the above is about streetlights which only turn on/off  once per day...but thermal cycling is even a danger for them.
So indeed, if your product coudl thermal cycle just a few times a day, every day, then stick to 105degC, this is what i repeatedly hear, on the "experienced grapevine"

[TimNJ]

As ejeffrey mentioned, other factors like the temperature rating of the PCB and nearby components often drive the maximum operating temperature of the power semiconductors. You can of course pick more exotic PCB materials, 125/130C rated electrolytics and so on, but this can get expensive pretty quickly. Usually makes sense just to get the MOSFET temperatures down first. For 150C rated semiconductors, I might allow up to ~110-115C under worst case conditions. But it depends on whether it’s mounted to a heatsink, SMD, whether it’s close to something sensitive to temperature etc. Keep your rules flexible to the circumstances.

[jonpaul]

Bonjour, a tous!

It seems Impossible to perceive  FTTS as all questions are theoretical and dreams    rather than practical engineering.

Since 1970s..1990s, the Limiting factor in reliability and life expectancy of modern SMPS, EBU power devices is not the Tj of the switch!

 Actual Switch Tj  is not near Tjmax!


Our  designs are soft switching or resonant converters,  thus low Psw and modest Pon, rather than hard switching.

The real limit in power supplies life/reliab. is rather

1/ The use of electrolytic caps .....we used special EBU 135 C rated ELKO in EBU for CFL/HID that had life 2000 hrs at rated Irippple/ambient 135 deg C.

2/ Catastrophic failure of rectifiers and EMI filters due to mains transients

Just the ramblings of an old EE !

Jon


(PS: Anniversary, 1 Mar, was 40 yrs my firm, and  55 yrs as EE!)

[mzzj]

Thermal stress is also much worse with cycling.  So a device that cycles between 25 and 125 C 500 times per day will have a much different effect than one that sees steady load for 8 hours a day

Thanks, this is  a great point.

As the below philips literature says, thermal cycling is indeed a major failure point.
https://www.docs.lighting.philips.com/en_gb/oem/download/xitanium-led-drivers-outdoor/White_paper_LED_Driver_lifetime_reliability.pdf
(...theyre talking about LED modules, but the same thing applies to FETs)

This is indeed it..the above is about streetlights which only turn on/off  once per day...but thermal cycling is even a danger for them.
So indeed, if your product coudl thermal cycle just a few times a day, every day, then stick to 105degC, this is what i repeatedly hear, on the "experienced grapevine"

As getting any reliable data on thermal cycling is time consuming (just like maximum junction temperature) the designers usually play it safe. There can be an enormous variation between models, brands or even batches.

Cycling data is often available for larger IGBT’s and SIC mosfets as it can be easily the limiting factor VFD use.
https://library.e.abb.com/public/9e7410602bcc4e67b0d11f7d75028026/On%20Superior%20Power%20Cycling%20capability%20of%20a%20High%20Power%20Density%20SiC%20Power%20Module%20for%20e-Mobility%20Application_Zhang_Biwei.pdf?x-sign=pDDq9Ah3UcyvwNneeDD5nGjow0WH2yzHVmX7c1yNo7s923hcmK1IsMG0YfuoD3Sx

https://www.osti.gov/servlets/purl/1325461
https://ntnuopen.ntnu.no/ntnu-xmlui/bitstream/handle/11250/2587123/EPE2018_Final_Gothner_Peftitsis_0041.pdf?sequence=2
https://lutpub.lut.fi/bitstream/handle/10024/86334/isbn%209789522653550.pdf?sequence=1&isAllowed=y

In SMPS use you might get 1 thermal cycle per decade or 1000 cycles per hour.

[Faringdon]

2/ Catastrophic failure of rectifiers and EMI filters due to mains transients

Thanks, thats interesting, i used to work in streetlighting, and we regularly sent our power supplies for transient testing by littelfuse......they upped the voltage till they failed, then returned the failed units to us.......but of all their returns, and of all the field returns i looked at, i never once saw a failed mains rectifier bridge.

[DEENGFANYUN]

The rectifier will burn out if it continues to work due to its performance degradation under high temperature（150 centigrade）. Not easy to break at normal temperature

[jonpaul]

Mains rectifier rating determine transients immunity
Voltage Derating, and use of avalanche diodes improved.

Many instruments and SMPS have had failed mains rectifier

j

[JohnG]

If you are starting a from-scratch design, then 105C is a rule of thumb useful for a first cut. You can't do a size, cost or reliability analysis without one. Then the real work starts, meaning you need to put in some real time and money to figure out the rest, at least if you have some challenging constraints.

John