Re: OLTC II- advice on tank cap

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

Original poster: "Malcolm Watts by way of Terry Fritz <teslalist-at-qwest-dot-net>" <m.j.watts-at-massey.ac.nz>

Hi Stephen,

On 6 May 2003, at 8:02, Tesla list wrote:

 > Original poster: "Stephen Conner by way of Terry Fritz 
<teslalist-at-qwest-dot-net>" <steve-at-scopeboy-dot-com>
 >
 > Hi folks,
 >
 > I just received a nice pair of 1200V 600A IGBT bricks today. Thanks Terry!
 > It gives me an evil feeling of power just looking at them. Now comes the
 > hard part. I need to design a MMC cap that will store at least 2J,
 > preferably 5-10J, at 1000V. So we're talking between 2 and 10uF. You might
 > think this is easy enough, but it needs to have very low inductance, and
 > deliver a peak current that could be more than 10,000A.

It isn't too formidable.

 > Proper film and foil caps seem to be out, because I can't buy them in large
 > sizes. It would cost a fortune and be very bulky. So I'm thinking
 > metallized polypropylene. 10 to 50 of .22uF 1000V caps in parallel would be
 > needed. Can anyone comment on this choice or suggest any particular kinds
 > of cap? I used .47uF 500V LCR metallized film caps in my original OLTC with
 > no problems. Even though they were only rated 40V/us but I ran them nearer
 > 200. They got a bit warm though.

I'd use poly/foil caps myself. Many caps in parallel is a good choice
to minimize ESL and ESR and the caps will share current in accordance
with the matching of their capacitances.

 > Last of all, the construction, how should I arrange the caps for lowest
 > inductance and even current sharing? If I had axial caps (leads come out
 > the ends) I could drill a grid of holes in two copper-clad PCBs or thin
 > copper plates, sandwich the caps between the plates, and solder them in
 > place. Would this be a good layout, or would the caps in the middle pass
 > less current than the ones at the edges?

Because of differing distances between the ends of the caps and the
junction points of the whole structure, it is difficult to get exact
current matching. A cylindrical structure with all caps equi-distant
from the junction point will get you closest. The differences are
probably not worth worrying about unless the operating frequency is
so high that the leads are starting to look like transmission lines.

Malcolm