Re: OLTC II Initial calculations

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

Original poster: "Terry Fritz" <teslalist-at-qwest-dot-net>

Hi Steve,

At 11:38 AM 5/8/2003 +0100, you wrote:
>>Emitter inductance is indeed a big issue.  If we have 3150 amps at 58kHz and
>>want only say 2 volts of gate disturbance, we can calculate the inductance:
>>
>>L = 1.74nH
>>
>>Yipps!!  That is not very much inductance!!
>
>It's pretty scary isn't it... I suppose there are ways that Powerex could 
>have done it. If they laid out the wiring so the emitter inductance was 
>almost identical for each die, then made a Kelvin connection to one die, 
>the disturbance would be cancelled out for all the dies. That's if you 
>don't count mutual inductance between the gate wiring and the 
>emitter/collector wiring.

They can pull Kelvin connections right off dedicated pads on the IGBT die 
too.  When (If ;-)) we blown one up, we will have to tear it apart to see 
what's in there ;-)  I hope Powerex tried to keep things RF clean inside 
the package.  Lesser devices are like a rat's nest of messy bond wiring and 
traces...

>I'm assuming (hoping) that they did something like that. However I suspect 
>that crap current sharing is the main reason the ratio of pulse 
>rating/continuous rating is less than for single devices.

Yeap, but no one drives them like we do with very little power dissipation 
at 3000 amps :-))...  The fact that we can keep the dies "cold" really 
gives us a lot of room to hit them really hard with current.



>>My Pearson 101 goes up to 50,000 amps so maybe I can test it...  I think I
>>have all the stuff here.  Not sure how to drive the beast...  120nF gate and
>>24nF Cies is not easy...  How were you thinking of driving it?
>
>With a totem-pole driver made from two 60 amp MOSFETS =|>  I have a 
>circuit for this that switches in 0.5us but it's heckuva complex and I 
>should be able to simplify it somewhat. I'll post a schematic when I've 
>got something usable. It would be great if you could do some tests because 
>I just don't have the equipment to deal with this :(

Be careful!!!  A 500nS gate rise time is like a 500kHz gate signal and the 
shoot through current is like 75 amps!!!  We want to keep the dV/dT on the 
collector from going though the 24nF collector to gate capacitance and 
driving up the gate voltage destroying the device.  As we lower the rise 
time, the current shooting through to the gate lead is lowered.  Need BIG 
transorbs in any case ;-))  But transorbs will cook if hit over and over...



>>Maybe Emitter inductance is not such a big deal after all!!  It gets lost in
>>the reverse transfer current...  Not sure how to handle gate resistors
>>either...  Have to turn the gate on with a ramp instead to eliminate the
>>resistor...
>
>As far as I understand, gate resistors are used to damp out parasitic 
>oscillations. The IGBT capacitances, gain of the IGBT when in the linear 
>region, gate circuit inductance, etc, form an oscillator. If you can keep 
>the gate circuit inductance low enough, the risetime fast, and deliver 
>stupid levels of drive current to slam it through the linear region before 
>it knows what's hit it, you can use a very low value resistor 
>(non-inductive natch) or nothing at all.

Yep!  But be careful of how giant that gate current needs to be.  We need 
to slam it on, but we also need to keep the current through Cgc from 
driving up the gate voltage during turn on too.  Just imagine adding a 24 
NF cap between the gate and the collector.  That IS exactly the situation 
we have with these big puppies...  The coil's 58kHz may help limit that, 
but need to be real careful.  Lots of SSTC IGBTs have died from this.  The 
gates can probably take 50 volts before the gate oxide breaks.  That may 
not actually kill the device, but it wears them down real fast!!



>>Be careful about making the secondary too big with large numbers of turns 
>>of thin wire on cardboard or
>>the secondary "Q" my have trouble like mine did.  Normally, secondary Q is
>>not a big deal but it can catchup to you.
>
>I'm hoping that the "Stubby" secondary (tm) will have a higher Q than a 
>taller thinner type. According to my ham radio handbook the aspect ratio 
>for optimum Q is 2:1. I was surprised by how well the Mini OLTC secondary 
>worked even though it had 1500 turns of 40swg (whatever that is in awg) 
>and a 800 ohm DCR. Ringdown traces showed that it had much less losses 
>than the primary circuit. I did wind it on a very thin (<1mm) 
>polycarbonate former however. For this one I hope to snag some thin walled 
>10" PVC or ABS pipe. I know they make this for fume cupboard ducting.

Cool!  Just don't use cardboard like I did ;-)

Cheers,

         Terry




>Steve C.