Re: Small Curtis as transistor driver?

[Jean-Claude Touzin <toujc@sympatico.ca>]

Lee,
Thanks for this eyes opener info.
I am in the "never done but want to" category!
So...
For my lectric Honda 3wheelers to be (I have scrounged some more parts in
addition to the 200A Aircraft generator, progressing ssslowly), since there is
not lot of space, I am thinking of a 36V system.
- are the problems described below more or less (or same!) manageable at lower voltage?
- since lower voltage mosfets are cheap, I was thinking to put lot of it (for
ex from digikey I can have 10 IRF1310N-ND for $35 (55V-72A-.012RDSon), giving
inefficiencies of my design do you think that I can start somehow the above
Aircraft thing with 720A of mosfets?
- one amp per gate (ouch!): will it be necessary also? I knew that they suck
amps like caps but how do you calculate when you have enough amp so that you
do not loose too much time in between off to fully on?
- Also since this is a generator, during off duty cycle it will continue to
"generate", so I will have to "kill it"? How is this done?
- maybe that I will try to control the fan heater motor of my car as a first
project after all!
- do some of you have a ready made diagram for a clean triangle wave at
15-20KHz? Just asking before going to the design board :-)
I'm silently reading the list every morning and learning a lot. Thanks to all
of you for sharing hard to get at info...
 Hope that my FrenchGlish did not get in the way...
JCT
=========
from LEE A HART:
In theory, neither MOSFETs nor IGBTs are hard to parallel. At, least, they are
easier than bipolar devices like transistors and SCRs.

Then reality sets in.

First, they need a lot of gate drive. Forget the "infinite input impedance" of
an insulated gate MOSFET or IGBT. They look like a capacitor, so there is no
DC current, but there is a substantial AC current. To get fast switching, you
need a driver that can deliver an amp or more per gate. Put dozens of devices
in parallel, and you need dozens of amps of gate drivers.

Second, the Miller capacitance between gate and drain or collector couples a
huge noise spike into the gate as the device switches on/off. If all gates are
in parallel, the first device to turn on pulls down the gate voltage for all
the rest. So most designers use multiple gate drivers; not one big one.

Third, devices are not matched. The ones with the lowest Rds(on) or Vce(sat)
will hog the current. Put five 100amp devices in parallel, and you'll probably
only get 400 amps through them before one of the devices hits its 100 amps. So
you have to match devices, or use extras.

Fourth, how do you consistently heatsink them all? If one part has a little
worse thermal connection to its heatsink, it overheats. While it may not go
into thermal runaway like bipolars, it will overheat and so fail to deliver
full power and have a shorter life.

Fifth, the parts will tend to have different lead lengths because they all are
mounted in physically different positions. Ones closer to the filter
capacitors may get more peak current, for example. So you have to arrange the
mechanical layout to minimize resistance and inductance differences between devices.

Now you can see why designers try to use the minimum number of devices!

Lee Hart                     If you would not be forgotten
4209 France Ave. N.          Soon as you are dead and rotten
Robbinsdale, MN 55422 USA    Either write things worth the reading
phone (612) 533-3226         Or do things worthy of the writing
e-mail XURQ03A@prodigy.com   (Ben Franklin, Poor Richard's Almanac)