Re: OLTC Update - Primary circuit resistance.

["Tesla list" <tesla-at-pupman-dot-com>]

Original poster: "Darren Freeman by way of Terry Fritz <twftesla-at-qwest-dot-net>" <free0076-at-flinders.edu.au>


Comments interspersed:


At 11:56 PM 19/08/2002, you wrote:
>Original poster: "Eddie Burwell by way of Terry Fritz 
><twftesla-at-qwest-dot-net>" <eburwell-at-columbus.rr-dot-com>
>
>At 11:14 PM 8/18/02 -0600, you wrote:
> >Original poster: "Terry Fritz" <twftesla-at-qwest-dot-net>
>
> >However, the current is "high"!  In my 50 volt testing tonight, I must have
> >been hitting 500 amps peak!  It may turn into a 120VAC coil instead of a
> >240VAC coil...  Or, get rid of that charging reactor...  Definitely must be
> >very careful of running without a secondary or a miss-tuned secondary.
> >MicroSim says I could hit 5000 amps in a nice slow ringdown.  That would
> >destroy the IGBTs...


Something I've been worried about for a while =)


>   One thing I have contemplated but have not tested is I believe that if you
>push an IGBT past it's maximum current ratings it may not necessarily die. I
>think the reason the manufacturers place the maximum peak current rating
>where they do is that if the current exceeds a certain point the IGBT
>latches on like a SCR. In other words the ratings are spec'ed to maintain
>gate control.
>   The important thing for semiconductor health is die temperature. We could
>start out by assuming that the pulse is so brief and intense that the
>thermal conductivity of the package can't remove the heat generated during
>the actual pulse. If that is the case, any energy dissipated in the device
>during the pulse will go directly to heating the die. We can measure the die
>and approximate it's mass. Using the specific heat of silicon we could then
>calculate the energy necessary to elevate the die temperature by some
>amount. After the pulse the die will cool down until it is hit with the next
>pulse.


With BJTs and I suppose IGBTs as well, this doesn't work. What happens is 
the most conductive part of the die takes a little more current than the 
rest, so heating itself up a little hotter. With MOSFETs that would cause 
its conductivity to drop and reduce its heating, so the process is 
stabilising. But with a BJT the locally heated region becomes more 
conductive and the current rises, causing more heating and so on until the 
device is destroyed.

At least it's what I've read, I never pushed insane currents through them 
before =)

Check the Safe Operating Area of the devices and don't expect to push much 
past it for too many pulses - there should be a limit to what you can do 
even in a once off pulse. Don't count on thermal effects allowing stupidly 
high currents =)


>   I have a feeling this approach should be valid especially since IGBTs
>don't have di/dt failure modes(at least for halfway sane levels of di/dt)
>like SCRs do. The main question is if there is some other failure mode that
>kicks in above Icm. If the IGBT latched on while the current was above Icm
>and the regained gate control below Icm the OLTC may seriously be in
>business. The other snag I can think of is that it's not just one pulse that
>the IGBT is dealing with, it's a sequence of decaying pulses. The first
>pulse will bump the die temp up the most. The last of the pulses as
>quenching is approached will probably be of little consequence as far as die
>heating is concerned.
>
>Perhaps it's time to destructively test some single devices under the heavy
>scrutiny of test equipment?


Probably a good idea if the plan is to go ahead!!


>Eddie Burwell


Have fun,
Darren Freeman