Re: Recent s.s.t.c work

["Tesla list" <tesla@xxxxxxxxxx>]

Original poster: "K. C. Herrick" <kchdlh@xxxxxxx>

Steve (& all)-

As to the current-sense resistor, I find that I have a 100 mV/50 A current 
shunt (a Janco #8632)  that looks as if it has very low inductance.  I 
suppose a Rogowski coil might be better; I'm planning to use one--or 
perhaps such a coil but with a ferrite core; don't know right now--for 
sensing the primary current for oscillation-feedback.  However, it's pulsed 
dc that I'm to sense for this overload purpose: the current from the 
bridge's power-source capacitors.  If that (excessive) current should 
become steady before the circuit could trip, then the Rogowski coil would 
no longer perform.  So perhaps I'll attempt to use the shunt.  I'll have 
very short leads from it to the trip circuit; no more than a few inches, 
and twisted.  Also, I could add a capacitor to slow down the response a tad.

Now that I have thought on this a bit more, I conclude that I should cut 
off the drive rather than dump the capacitor charge as I described.  The 
UCC37321/-2 drivers I will use conveniently have enable inputs that, when 
false, cause the outputs to go to 0.  That should cut off the IGBTs a lot 
faster than discharging 22 uF capacitors from 25V.  In order to do that 
I'll need another opto-isolator--but those still turn on within a us or so.

I hear you on the philosophical matter...but I've toasted so many MOSFETs 
in my day--tho not bricks and certainly due mostly to my design 
errors--that I'm gun-shy.

And by the way, those buckets I use are 18-quart size, not gallon.  Large 
but not humungous.

KCH

Tesla list wrote:
> Original poster: Steve Conner <mailto:steve@xxxxxxxxxxxx><steve@xxxxxxxxxxxx>
>
>
>
> > I invite comments from the List
>
> Yeeha! Go for it KC!
>
>
> > Perhaps someone has an opinion as to whether that's fast enough to save a 
> > brick in the event of full upper:lower conduction.
>
> I think you are mighty brave to actually try and incorporate shoot-through 
> protection. Any protection is probably better than what we have just now.
>
> The problem you'll run up against with this design is stray inductance of 
> the current sense resistor. You wrote that it had a resistance of 0.001 
> ohm. So the reactance at the operating frequency ought to be less than 
> that or you'll get spurious triggering due to L*di/dt voltages.
>
> If you're running at 50kHz then Xl=0.001=2pi*50000*L so L=3.2nH, good luck!
>
> And the same goes for any mutual inductance linking the power circuit to 
> the wires that connect the shunt to the sensing circuit. This is one 
> reason why we tend to use CTs and Rogowski coils rather than shunts. In 
> your existing circuit you could use a Rogowski coil in series with the 
> sensing wires and adjust it to "buck out" the stray self/mutual inductance.
>
> Finally, there's a philosophical question here. If your driver circuit is 
> reliable it should never cause shoot-through by misdriving. So the only 
> situation that will cause serious shoot-through is if one IGBT in the arm 
> tries to turn on into the other one that has failed short circuit. Now, 
> those IGBTs come two to a package and if one has failed the whole package 
> has to be thrown away. So who cares about saving the other one? Just my 
> opinion of course. I could probably be persuaded otherwise.
>
>
> > The photo shows my primary: 6 turns of paralleled 1/4" Cu tubing wound 
> > inside an 18-gallon plastic bucket.
>
> Neat!
>
>
> Steve Conner