[TCML] Re: IGBT paralleling

[Greg Leyh <lod@xxxxxxxxxxx>]

Finn Hammer skrev:


> Holy cow, I just snipped 8 replies deep......

Cheers!  It boggles the mind as to why people attach those huge rambling 
multi-100-line replies to their posts; perhaps there should be an award 
for the most convoluted ones?  ;)

You're correct that the switching speed of bricks is largely limited by 
their internal bus inductance.  Some bricks have internal discrete gate 
resistors at each die as well, to prevent fast criss-cross oscillations. 
 Moreover, the ability of a brick to survive fast dI/dt faults is 
limited by the *imbalance* of bus inductances between dies, and even 
between the sides of a given die.

Back in 2002 I explored this in greater detail by cracking open some 
3300V bricks, and tried my hand at laying out some 3300V and 6500V 
bricks with bus structures suitable for sub-microsecond switching at 
2000A.  Here's a link to the paper:

http://www.lightninglab.org/papers/leyh/PAC2003.pdf

Typically, industrial applications using these IGBT bricks run them with 
safety factors slightly greater than two;  the 3300V bricks are used in 
bridge configurations for 1500V DC link voltages, and the 6500V IGBTs 
are used on 3kV systems.

Currently, I use two 4500V IGBT bricks in parallel on a 3200V DC link 
for my 1:12 scale twin coil.  Not nearly the same safety factor as 
above, but then again the primary ckt is a pure resonant switcher, where 
the IGBTs used in a 3kV electric train are hard switchers.  -GL



> Scott,
>
> I cannot say about parallelling, and then....
>
> IGBT modules are what is also referred to as Bricks.
> They are the large.... arh - well - erh  -  you know   -   Bricks or  
> modules      we use in the large coils, like Steve's and my Thumper.
>
> It is interesting that these Modules internally consist of parallelled   
> IGBT and Freewheling diode dies.
>
> These dies are bussed in a fasion that allow them to operate within their  
> specksheet ratings.
> I seem to recall one very large full bridge which didn't last long, due to  
> the inductive limitation of the internal bussing, which lead to  
> catastriphic failure of the dies in just one end of the module.
>
> What this tells me is, that with proper attention to layout, you can  
> parallell IGBT's, but that doing so is a pain, and that most of us would  
> just go get a brick and be done with it.
>
>
> Cheers, Finn Hammer


_______________________________________________
Tesla mailing list
Tesla@xxxxxxxxxx
http://www.pupman.com/mailman/listinfo/tesla