A MOSFET transistor IRF530 is dissipating a pulse like power curve(green plot below) for around a second during its switch-off(red and blue are voltage and current):

enter image description here (left-click to enlarge)

And here closer look: enter image description here This MOSFET is not manufactured to be used in linear region but for some reasons I have to use it this way even though it switches off slow.

I have asked a related question before here: A question about heating of a slowly turning off MOSFET and I'm sure the answers were great and clear for many but unfortunately there are some confusing points that I still couldn't grasp.

That previous question of mine led me to believe that I wasn't able to interpret this bell-shaped power pulse(green plot) when using a data-sheet.

As you see, during transition this power in linear region makes a peak to around 30W and decreases with the same rate and the whole event takes a second or a bit more. It is not periodic(not repeating, only one single pulse), or should I say the period is infinity. In this case if I take the period infinity the average power goes to zero. But if I take the average across the 1 sec transient time the average power becomes around 15W. So depending on the integration interval we get different average powers. As you know LTspice calculates and shows this average power for the selected time interval automatically and you can see that average in my above plots as well.

After this short introduction, let me explain my confusion:

I have been trying to find answers to the following two questions for the last week without any achievement:

1-) Remembering that this is a single bellshaped transient(not a periodic) and is not exaclty a typical square pulse mentioned like in the data-sheets: If I want to use this MOSFET's power dissipation and use this power to check anything with the data sheet which power should I use? Average power or peak power? If average power, what should be the integral interval? I currently calculate the heatsink by considering the power dissipation as 15W even it is a single pulse. And if I take the average through 1 sec and say the MOSFET dissipates 15W average, then how will it fit to the SOA graph? The bell shaped power pulse should be considered as a 15W power for 1 second pulse width? I hope I made this one clear enough, pardon my english.

2-) I found out that LTspice has a new feature which can simulate the temperature of an nMOSFET called: SOAtherm which can also be used with SOAtherm-Heatsink.

Here is the information about it:

"Modeling Safe Operating Area Behavior of N-channel MOSFETs" at: http://www.linear.com/solutions/5239 And here is the video how to use this feature: http://www.linear.com/solutions/5445

It seems they only have this option for some MOSFET models(not for IRF530) but they also have a "User defined" option which makes me to think one can add any MOSFET?

When I look into their library SOAtherm-NMOS.lib for each MSOFET the SPICE code starts as for example:

.subckt BSC009NE2LS D G S D2 G2 S2 Tj Tc

* The following two lines are customized for each  model * 
.param Tambient=85 RthetaJA=50 Cheatsink=0 
.param Imult=1.00E+01 Iexponent=3.00E-01 R6=3.18E-02 C6=1.77E-04 R5=1.01E-01  C5=2.77E-04 R4=2.74E-01  C4=1.60E-03 R3=8.49E-01  C3=8.83E-03 R2=7.34E-02  C2=3.30E-03 R1=1.05E-03  C1=1.82E-03

* The remaining lines do not change between models
... the rest of the code same for all MOSFETs...

It means one can modify couple of lines for a custom model. I wanted to do the same thing for IRF530 but what are these parameters I couldnt figure out. I only know RthetaJA from data-sheet but have no idea what are C1 C2 R1 R2 ect...

Has anybody dealt with this tool and know how to customize? I've sent an email and asked about this to them but didn't receive any reply after a week.

In their web-site there is great explanation about how to use this tool but no info at all about how to customize.

  • \$\begingroup\$ I haven't used that feature, but the values are for a Cauer RC ladder (RClowpass, followed by RC lowpass, etc). They set the thermal time constant, i.e. how fast/slow the device heats up during usage. I don't know about Imult and Iexponent, either, but I think it's safe to start with them as defaults. There is an .xls file on that page (and subsequent links) which should give you more info about which MOSFET uses what values. Maybe that could help. \$\endgroup\$ – a concerned citizen Nov 10 '16 at 6:57
  • \$\begingroup\$ Looking at the SOA chart, it specifies up to 10ms pulse. Extrapolate another line by equal spacing for 100ms, that more or less hits the 88W Maximum Power Dissipation limit. So take a 100ms slice around the peak of the pulse, integrate/calculate the power and compare that to 88W. \$\endgroup\$ – rioraxe Nov 10 '16 at 8:06
  • \$\begingroup\$ @rioraxe thanks for the answer. but couple of things and would be glad if you can comment on: 1-) do you mean that those lines(like 100ms 88W in your example) are power lines which indicate the maximum durations the MOSFET can handle? if they are, are they considering no heatsink or with heatsink??(because the graph says 175 C° for the junction) \$\endgroup\$ – user16307 Nov 10 '16 at 10:26
  • \$\begingroup\$ 2-) if i take 100ms average around the peak it will be 30W less than 88W so i can conclude it is in safe area. but if i take 1 sec average for the whole bellshaped power pulse the average becomes 16W and it is in the limit. thats why im not sure which interval should i take. what do you think about it? \$\endgroup\$ – user16307 Nov 10 '16 at 10:26
  • 1
    \$\begingroup\$ If you look at the 10ms line, it is at around 250W. That means if the case (Tc) is at 25C, it implies that a 10ms 250W pulse would not heat the junction (Tj) past 175C. So if extrapolating a 100ms line, the limit would be around 100W, or call it 88W. Therefore for 100ms or longer period, there is no extension of the range beyond the 88W. So for any 100ms period, the power should be less than 88W. \$\endgroup\$ – rioraxe Nov 10 '16 at 11:25

Your Answer

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

Browse other questions tagged or ask your own question.