0
\$\begingroup\$

In many SMPS application (BOOST converter in particular), power MOSFET is used as a switch and it is driven by a PWM/IC at a frequency in a range of hundreds kHz (f = 200kHz for instance). This means that if a SMPS is used, says for 1 hour, the MOSFET would have switched \$f \cdot 1 \cdot 3600 = 72 \cdot 10^7\$ times !!

Does this amount of switching shocks anyone? I imagine that each MOSFET is designed to sustain a certain amount of switching (for instance \$10^9\$ times). If this limit is reached, what will happen to the MOSFET?

Thank you in advance for your answeer !!

\$\endgroup\$
  • 1
    \$\begingroup\$ it is not a mechanical switch .... components in your computer switch several billion times per second or more \$\endgroup\$ – jsotola Jun 9 '18 at 6:30
  • \$\begingroup\$ Yeah I am aware of that. But according to UTE C 80 810 (an data handbook for reliability estimation), usage of power MOSFET (> 5W) is limited in maximal number of switchings. So I am really confused how to interprete this notion and real life application... \$\endgroup\$ – Tuan Jun 9 '18 at 6:34
  • \$\begingroup\$ High power switching regulators tend to operate at lower frequencies. I am not an expert on high-power converters, but I believe 30kHz might be more realistic. I am not familiar with UTEC C 80 810, but when they say > 5W, is that greater than 5W of power dissipation in the transistor, or greater than 5W of power delivery? \$\endgroup\$ – mkeith Jun 9 '18 at 6:39
  • \$\begingroup\$ >5W as in maximal power dissipated (conduction and switching loss) \$\endgroup\$ – Tuan Jun 9 '18 at 6:43
  • 2
    \$\begingroup\$ I have an smps that is happily doing its duty after 2e13 switches, and surely there are some with twice that much \$\endgroup\$ – PlasmaHH Jun 9 '18 at 8:08
2
\$\begingroup\$

I imagine that each MOSFET is designed to sustain a certain amount of switching (for instance 10^9 times)

You imagine wrong then.

There are several mechanisms by which a MOSFET might be damaged. Any particular type of MOSFET is given ratings so that if the designer stays within them, the lifetime will be essentially unlimited, certainly exceeding the expected lifetime of any equipment they are built into.

Switching is a high-stress moment for a MOSFET, resulting in high channel dissipation as there's a current flowing with a voltage across the channel. However, the SOA (safe operating area) rating specifies the maximum voltage and current for any given time, and heat within this limit is absorbed safely and without exceeding the maximum junction temperature.

MOSFETs are made by diffusion at high temperature, and they can be unmade by the same process. The rate of diffusion increases exponentially with temperature. Most devices have a rated maximum junction temperature of 150C. This temperature is so far below the temperature at which diffusion happens rapidly that it would take 'a very long time indeed' for the structure to diffuse into something else.

This is in contrast to a switch or relay, whose contacts suffer from arcing and erosion on every switching operation, and so do have a lifetime rated in the number of switching operations before failure. 10^9 is a rather small number for switching events in a transistor, but quite a big number for switching events in a relay or switch.

You mention in comments that you are concerned by UTE C 80-810, which has the following comments under power transistors >5 watts, page 43 if you're looking at the 2000 version.

enter image description here enter image description here

The cycles referred to are thermal cycles, cold/hot/cold when switching equipment on and off. This subjects the package, its interface to the die, and seals to the leads, to mechanical stress, which can cause failure.

\$\endgroup\$
  • \$\begingroup\$ Thank you for your argumented answeer. I am well convinced by what you are saying. But what bothers me is why such standard like the UTE C 80-810 defines the power MOSFET (>5W) as be limited in number of switching? Is there any known failure that is due to the passing of this limit? \$\endgroup\$ – Tuan Jun 9 '18 at 7:48
  • \$\begingroup\$ @Tuan It's a long document isn't it. A quick search for '5W' yields only 2 hits for 1.5W zener diodes, so it'll will have to wait until I have some leisure time to look at it properly. Unless you want to give me a section number, which might speed up my search. \$\endgroup\$ – Neil_UK Jun 9 '18 at 8:22
  • \$\begingroup\$ Page 43 (power transistor) in 2000 version, or search for "5 watts" or "5 w" text \$\endgroup\$ – Tuan Jun 9 '18 at 8:36
  • 1
    \$\begingroup\$ @Tuan do you mean Under cyclic operation, the number of cycles is limited to: 7 0.05 10 j t N e− ∆ = ∆t j : junction temperature range? That's cycles, thermal cycling from cold to hot and back to cold again, which mechanically stresses the package/die/lead interface. Too much to read and absorb on that page right now. \$\endgroup\$ – Neil_UK Jun 9 '18 at 8:43
  • \$\begingroup\$ There is a document that has more information on MOSFETS than some might want, but it does list the FIT for an example part of 0.086 @ 55C yielding MTBF @ 55C of 1.16 x 10^^10 hours. mouser.com/pdfdocs/Nexperia_MOSFET_FAQ.pdf \$\endgroup\$ – Peter Smith Jun 9 '18 at 10:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.