From: Winfield Hill
Subject: Re: Gate drive transformer ferrite?
Date: 19 Oct 2002 06:39:24 -0700
Organization: Rowland Institute
References: <3DB09F8E.4070509@BOGUS.earthlink.net> <3DB0A0B8.DCC38BFD@rica.net>
X-Newsreader: Direct Read News 2.98
John Larkin wrote...
> John Popelish wrote:
>> I think that for low power pulse forming transformers, permeability
>> is king. The 78 material should do fine, but other high permeability,
>> lower frequency ferrites are often used as well. This works, because
>> low power transformers often use only a fraction of the core capability
>> in terms of saturation. Loss per cycle is often unimportant if the
>> excitation is low. If you look at the permeability versus frequency
>> for several ferrites, you will find that some of the low frequency
>> types have a high frequency permeability that equals the permeability
>> of a lower frequency lower permeability type over some range of
>> frequencies that exceed the flat permeability part of the spectrum.
> sounds good. We're using a Magnetics W-41206-TC torroid (u=10,000) for
> gate drive. It's wound with 4 turns of micro-coax, with the braid used
> as the primary and the inner as the secondary. You can get sub-ns
> risetimes like this. The core mainly supports the big flat parts of
> the pulse, and the transmission-line winding handles the fast stuff.
John, more detail please on what FET you were driving, and with what
drive ckt, etc. Considering for example a small MOSFET, such as the
BSS138, with Ciss = 30pF. Oops, that's at 25V, at Vgs = 1V it's 60pF.
What's the model to use for your unconventional pulse transformer? If
we consider it to be a 50-ohm transmission line, we have a time constant
of 50 * 60pF = 3ns. That's fast but it's not sub-ns. Of course, it's
not a transmission line, not just an inch or so of coax.
We can model the transformer as some inductance and capacitance, driven
by a sub-ns drive pulse, capacitively coupled to the gate capacitance,
. ,-||-, |--'
. __|\_________| |_____||--, FET
. |/ driver >||< | |
. >||< | Ciss|
. >||< '-||--+
. gnd ____ >||< _________|__
Hmm, say we have 2 inches of coax, could be up to 5pF, but not enough
to charge the FET gate's 60pF. Of course, on a ns time scale, FET gate
spreading resistance means the leading-edge of our sub-ns pulse sees only
a small portion of the gate capacitance, say 10%, which can be quickly
charged, enough to start that 10% portion of the FET conducting. With a
big enough drive voltage, that may be enough to create a say 50-ohm FET
resistance, which in some applications could have a useful effect.
So, with specious hand waving, we've rationalized an isolated sub-ns FET
gate driver. In the process we've ignored the FET's Crss 8pF (Vds=5V)
feedback capacitance. We've also ignore the smt FET's few nH of lead
inductance (2nH 60pF resonates at 460MHz and has a 0.5ns T/4 risetime).
But, ignoring all that, we press on. As ns by ns ticks by, the pulse
spreads further into the gate capacitance, requiring more drive charge.
But now traditional transformer action kicks in, completing FET turnon.
OK, John, set us straight, give us the straight dope. Lay it it for us.