From: John Larkin
Subject: Re: Gate drive transformer ferrite?
Date: Sat, 19 Oct 2002 11:31:49 -0700
Organization: Posted via Supernews, http://www.supernews.com
References: <3DB09F8E.4070509@BOGUS.earthlink.net> <3DB0A0B8.DCC38BFD@rica.net>
X-Newsreader: Forte Agent 1.91/32.564
my knowledge of this is almost all experimental. This is because 1) in
most real situations, the transformer is not the speed limiter and 2)
the standard Spice transmision lines can't be used to model this, so
it would be a serious nuisance to simulate. So apart from the obvious
low-freq stuff related to magnetization/saturation, I have pretty much
treated these as ideal transformers with maybe a tiny bit of leakage
When driving a big honkin' mosfet, the problem is finding a driver to
give the speed and charge needed, and then the fet parasitics. PCB
layout inductance typically swamps transformer leakage here. The
standard 'ultra fast' fet drivers, to use some technical terms, suck
bigtime. The MAX4420 is about the best of that sort, and it is still
slow and has some very weird duty cycle effects.
Where this kind of transformer is really useful is with gaasfets. A
big fet (like a CLY-10 or 15) will go from off to an Idss of many amps
in about 3 volts, and the input c is very small (Crss ~~ 0!) so this
is a good gate driver. A gaasfet is also an excellent way to drive the
A few of the goofy things I've done here are...
CLY2 gaasfet driving 1:1 transformer. This puts 5 volts into
an isolated 50-ohm load in well under 1 ns. Scoping this, the
transformer looks ideal. Used to fire a largish laser.
CLY2 driving two paralleled 2N7002s, driving a 2:1 stepup. This
is bizarre, a gaasfet driving the gates of mosfets. With 24
volt supply available, output is about 40 volts into 50 ohms,
in maybe 2-3 ns. Used to fire detonators, I think.
The aforementioned sucky Maxim part driving two paralleled
NTP15N40 fets, through the transformer, switching 400 volts/40
amps. The problem wasn't speed (we used surprisingly large
series gate r's to keep the edge from trashing everything else
in the box) but rather the nasty pulse width variations vs
reprate. Not the transformer's fault.
So what I need now is a good gate driver circuit to replace the Maxim
stuff. The Ixys things look better, but I suspect a nice discrete
driver would be better and cheaper. We need maybe 20 volts gate drive
into the NTPs, 100 ns or thereabouts, fast edges, but mostly very
stable pulse widths vs reprate. Any ideas?
(One other interesting configuration would be the same core and coax,
but rewired as a common-mode isolator. This is a lot easier to
analyze, and might be useful in a few special situations.)
The bane of my life (kids excepted) seems to be translating very fast,
small-voltage-swing signals into very fast, large-swing thingies.
Hey, the sun came out! That's novel around here. Maybe I'll go out for