200V N-channel FET 200V with nsec speed?

[Robert Baer]

Fred Bartoli wrote...



I'm not sure cascode would be useful for fast turnoff.

We need to know more to better answer Joerg's question, for example,
is he thinking of completely switching the MOSFET in 4ns, or is he
thinking of obtaining a say 180V swing in 5ns, say not including the
last 20V of a 200V load? Is he doing this once every now and then,
or is he seeking to repeat this every 5 to 10ns, as Henry would like
to do? Joerg, tell us, what are you working on?

I'll venture the observation that many, if not most, high-voltage
MOSFETs can be made to switch ON in 5 to 10ns if enough gate current
is applied. At least the near portion of the die will switch; there
may be issues getting all of the MOSFET's area to rapidly turn on,
and it may not achieve its low long-term Rds(on) value until many
more ns have passed. If you look at a typical spice MOSFET model,
you'll usually see a series gate "spreading" resistor, often 50 ohms.
I have found these apparently arbitrarily-picked values to be far too
high when compared to the performance of an actual part on the bench.
A better model would divide the MOSFET into several portions, with a
very low gate spreading resistor for one portion, etc.

With respect to fast turn-on, for small output load-current swings
the gate scene described above is the issue, but for high currents,
driving capacitive loads, etc, where a high dV/dt means a high I/C,
the raw current capability of the MOSFET can become the determining
issue. For example, I just completed a fast 1.2kV cable pulser in
which I used a single FET switch to drive a 50-ohm coax through a
50-ohm source resistor. My MOSFET needed to sink I = 1kV/100 = 10A
in driving the 50-ohm resistor in series with the 50-ohm coax. I
observed that even though it could sink 5A to make a 500V pulse in
10ns, when making a 1kV 10A pulse it slowed to about 15ns. There's
a set of electrodes at the far end of the coax, which sees a nice
-1kV pulse with 15ns risetime. Happily that was good enough for us.

Joerg mentioned fast rise and fall time. While it's easy to rapidly
turn off a portion of the area in a typical power MOSFET, portions
that have high series gate resistance will stay on until later, so
that a gradual rather than abrupt complete turnoff is experienced.
This can last as long as 50ns or even 100ns in severe cases. This
is assuming the turn-on gate drive lasted long enough to reach the
far parts of the MOSFET. I've found wide variations in this effect
from one manufacturer's part to another, and in one type to another.
Joerg, you'll need to experiment with many different MOSFETs to see
how they fare. Sadly, these issues are not covered in the datasheet.

Then again, for fast turnoff, aggressively pull current *out* of the
gate using a negative-going gate drive...

 

[Joerg]

Hello Tom,

...

FWIW...

Back during the early days of power mosfets, I used one in essentially
a forward-mode switching supply to drive a load that varied with time
(well, really with heating...a filament whose resistance increased
considerably as it got hot). I blew up quite a few mosfets figuring
out what was going on right at turn-on: that is, when the power to the
filament was turned on, not when the FET turned on each cycle. This
was before the time of digital scopes, and the fastest available
storage scope had some trouble keeping up. I was getting about 200V in
WELL under 10 nanoseconds on turnoff: it turned out to be because the
load, a transformer that drove the filament, was storing quite a bit of
inductive energy that wanted to come back out quickly. It was NOT
because of inductive kickback overvoltage driving the FET into
avalanche.

Might it be possible to use some inductance in the net load, to help
out?

There is already some inductance ;-)

Warning: at least with older power mosfet designs, too high a dv/dt
could cause failure of the mosfet.

Beware g-d capacitance: 50pF*100V/nsec = 5 amps, and that has to come
out of the gate. There may be some advantage to cascode. Also: if
you can account for delay time externally and only need a fast rise
time, it seems like that should be easier.

I'd hate to add capacitance to the output node you're driving, but
could a P-channel (or N-channel) pullup help? How about a current
from a higher voltage, with a clamp diode to keep it limited to
whatever you want?

I did already suggest push-pull which would remedy a lot of problems.
Hopefully the engineers there will warm up to that idea because it makes
life easier.

Have you looked into power mosfets used in RF amplifiers? That's not
an area I'm very familiar with, but I think you can find them with 200V
ratings. Clearly a mosfet driving a load at 150MHz must be able to
deal with cycles shorter than your application. Normal power mosfets
as you'd use in a 100kHz to 1MHz switching power supply don't work well
for VHF power amplification, as compared with RF power mosfets designed
for the job. They'll be pricier.

Those are generally too low in breakdown voltage. The option is, of
course, push-pull with a step-up transformer. For that scenario we'd
already have the right FETs.

 

[Joerg]

Hello Robert,

Then again, for fast turnoff, aggressively pull current *out* of the
gate using a negative-going gate drive...

Or employ a fast clamp that charges up a capacitor, then use the cap as
a "kicker".

 

[Joerg]

Hello Robert,

What about the BSS131 by Infineon? rather small die size, rated to
240V and N-channel SOT-23 package).

It has poor tdoff and tf (>>10nsec), plus we'd need to parallel about
ten of them which drives the capacitances to uncomfortable levels.
Infineon does have some newer nice devices but I am a bit hesitant with
that manufacturer after they weren't able to get samples of their BSP297
to us. Despite the fact that I told them we'd pay whatever it takes to
get them across the ocean and almost two hours on the phone :-(

 

[Robert Baer]

Hello Robert,




It has poor tdoff and tf (>>10nsec), plus we'd need to parallel about
ten of them which drives the capacitances to uncomfortable levels.
Infineon does have some newer nice devices but I am a bit hesitant with
that manufacturer after they weren't able to get samples of their BSP297
to us. Despite the fact that I told them we'd pay whatever it takes to
get them across the ocean and almost two hours on the phone :-(

Look at ST Micro STN1N20 or STQ1NC45R-AP.

 

[Boris Mohar]

Hello Robert,



It has poor tdoff and tf (>>10nsec), plus we'd need to parallel about
ten of them which drives the capacitances to uncomfortable levels.
Infineon does have some newer nice devices but I am a bit hesitant with
that manufacturer after they weren't able to get samples of their BSP297
to us. Despite the fact that I told them we'd pay whatever it takes to
get them across the ocean and almost two hours on the phone :-(

When I was working on ground penetrating radar we discovered several things
most of which are still true today. The on and off times are as per test
circuit and not absolute limits of the device. Many devices could sustain
larger than specified gate voltage overdrive. Caveat here, our duty cycle was
low. Often unspecified Internal Source Inductance was the unwelcome wall.

 

[Joerg]

Hello Robert,

Look at ST Micro STN1N20 ...

Thanks. Whatever they mean with crossover time is a bit long but we'll
order some.

... or STQ1NC45R-AP.

This one is a bit slow but as Win hinted one has to take that data with
a grain of salt. So, we'll order some of these as well.

 

[Steve Kavanagh]

I am a bit hesitant with
that manufacturer after they weren't able to get samples of their BSP297
to us.

I've been casting about for fast MOSFETs too (though I can live with
100V but need 10W or so dissipation rating) and found that Infineon's
OptiMOS series looks quite promising. But nobody seems to stock much
of what is supposedly available !

Steve

 

[Joerg]

Hello Steve,

I've been casting about for fast MOSFETs too (though I can live with
100V but need 10W or so dissipation rating) and found that Infineon's
OptiMOS series looks quite promising. But nobody seems to stock much
of what is supposedly available !

That's just the problem. They do not seem to understand how electronic
parts are marketed. By now I believe it would take some high-level
management changes over there to correct that problem. Then their sales
will probably increase, big time. They need some execs who pound the
pavement and regularly do spot-check visits on clients large and small.
Just to feel the pulse. That's what we did in our med biz and the CEO
expected it from us.