Re: OLTC - Primer

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

Original poster: "Marco Denicolai by way of Terry Fritz <twftesla-at-qwest-dot-net>" <marco.denicolai-at-tellabs-dot-com>

Hi Terry,

Tesla list wrote:
<SNIP>
> The thing is on 5 amps fuses, I am thinking the IGBTs will often win if
> something does not go right.  

Yeah... Sadly it really is true that the fastest fuse is the
semiconductor you are trying to protect itself :)
A 5A fuse that blows within 1 ms (and only when 50A or more  is passed
through it) is a FF rated one, that is a very fast one. The
short-circuit rated IGBTs usually will last for some tenths of
microseconds when short-circuited. More, you need a fuse rated for, say,
600 VDC, not just the usual ones. Otherwise the fuse will open but the
current will spark over it (internally) and keep on flowing for even a
longer time.
You just need a thick-film resistor and a inexpensive 2N2222 or similar
to protect each IGBT from overcurrent and leave time to the fuse to
blow. This for the OLTC final version, I mean.

>Another advantage of having the IGBTs on
> independent caps is that a failure will be fairly well isolated and easy to
> find.  Having the IGBTs in parallel would have made finding a bad one hard
> (of course, they would probably all be bad in that case).  

I must confess I had a 10-days vacation and I didn't dare to read
through the >600 list posts (just damped them, :( ). Would you mind to
explain once more this independent cap/IGBT connection scheme?

>I am trying to
> get enough silicon in the IGBTs so that they can absorb the full energy of
> the coil (until the fuses blow) if something goes bad.  

In all my IGBT damages (uncountables...) the mechanism is always been
the same:

1. the Vge rating (+/- 20V) is exceeded -> the gate insulation goes
breakdown
2. the IGBT goes into short circuit with <1 ohm impedance between G, C
and E
3. the short circuit generated blows what is capable of: another IGBT,
the PCB traces, the FF (yes FF!) fuse, etc.

I suggest you to tune the gate drive circuitry from the very beginning.
A high collector current is usually not one of the first problems. If
you have a lousy gate drive, collector current variations (e.g.
transients) will be reflected onto the gate, that will amplify them back
to the collector, that will (again) be seen on the gate. A kind of nice
loop, thanks to the Cge and Cgc capacitances internal to each IGBT.
You could start with a very slow turn on (i.e. a 100 ohm gate res) and a
rock solid, low impedance turn off (just a schottky in antiparallel with
the 100 ohm). Easier to deal with losses with more cooling than to
replace broken devices.

>If I redo it, I
> will go to isolated heatsinks so I don't have the sil pads. 

I use a long sil "slice" cut by myslef from a larger foil. While you are
replacing one IGBT, the others will keep the sil in place for you. 

Best Regards

-- 
_____________________________________________________________

 Marco Denicolai           Senior Design Engineer 
 Tellabs Oy                tel: +358 9 4131 2769
 DSL Products              mobile: +358 50 353 9468
 Sinikalliontie 7          fax: +358 9 4131 2410
 02630 Espoo  FINLAND      email: marco.denicolai-at-tellabs-dot-com
_____________________________________________________________
============================================================