RE: OL-DRSSTC 7 - It's Alive!

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "Jason Judd" <JJudd@xxxxxxxxxxxxxxxx>


Terry,

Thanks again ! I had neglected to take into account the switching losses
when I did my thermal calcs, thinking that the duty cycle was low enough
to ignore it. I had had noticed that my heatsinks temps increase
dramatically with non zero current switching so those switching losses
must really add up fast.

> Hi Jason,
>
>
> I see.  The "peak power" in that case might be enough to
> super heat the die over about say 200C for an instant.  That
> will explode the die.  When it is not switching, there is no
> switching loss in the equation.  But if the die has say 320
> amps 350VDC for 1uS, the peak power is over 100kW(!!) and 0.1J.
>
> Figure 6 in the data sheet puts the pulsed power thermal
> resistance at 0.007 c/w absolute minimum.  Assuming we can go
> from 25C to 175C and survive, the pulse power is 150C / 0.007
> = 21kW.  Thus, anytime the IGBT sees that, destruction is
> almost assured!!  There is no fudge or safety factor in that
> number and the destruction is "instant".

So from the data sheet, this is based on a 10us pulse ? If the pulse is
shorter then the IGBT should be able to cope with higher peak
dissipation levels ?

>
> Assuming the voltage drops by 1/2 and the current is 1/2 at
> the worst power peak in the switching and the bus voltage is
> say 320V...
>
> 1/2 x 320 x 1/2 x Ipeak = 21000    Ipeak = 262 amps.  Pretty close to
> your 320 amps!!

Given switch off would normally occur in less than 200ns is the 21KW
figure still correct ? I would have thought to work it out we would need
to calculate the Junction temp after each half cycle and add in the
switching loss for each switch cycle in the burst. The data sheets
switching loss graph (figure 8) only extends to about 60 amps though so
accuracy would be a problem.

>
> Without switching, at about 750 amps with 25VDC on the gate,
> the current is so high that the IGBT starts to fall out of
> saturation and the voltage across it start to dramatically
> rise.  More gate voltage does not help at that point.  21000
> / 750 = 28 volts.  So above 750 amps, the IGBT will turn into
> a resistor, and then a fire cracker real fast once Vce gets
> to about 28V.
>
>
>
> >The gate spikes I was seeing were going negative for about
> 50 ns, but I
> >never saw them exceed the gate voltage rating. Could those
> fast spikes
> >tear the gate apart without exceeding the voltage rating ?
>
> Could they have been scope probe glitches??  I see odd little
> glitches like that all the time.  Moving the probes around
> changes them all over the place.  Sometimes it is best just
> to switch the scope bandwidth down to 20MHz and be happy :o)
> Miller switching glitches go positive not negative.  The
> gates are pretty good 3.3nF capacitors at that point so a
> 50nS glitch across them would need a lot of current.
>
> V=1/C INT i Dt  15 = 1/(3.3e-9) I 50e-9    I = 1.0 amp
>
> A 1 amp gate current will not damage the IGBT.  They are
> using 10mil gate bond wires and all that.  I think it is a
> peak power thing instead.

OK, It seems the gates are not so much the problem.


> I think maybe fine tuning the numbers in the peak power thing above
> will answer the question without any more smoke ;-)
>

No smore smoke would be good, those IRG4PF50WDs are hard to get at a
decent price here in Australia. I'll keep working on the thermal
numbers...

Cheers,

Jason.