Maker Pro
Maker Pro

IGBT in a MOSFET application ?

M

Mike Diack

Jan 1, 1970
0
I need a 1Kv 200ma regulated variable supply, and thought that the
reference design in Win Hill's "A of E" was a good place to start. Trouble
is that MOSFETs hit the wall at 1Kv, wheras IGBTs are readily and cheaply
available up to 1200v (more than enough, the raw supply will be around
1160v). I cannot see any reason why the IGBT should not make a suitable
series pass element in a linear regulator like this - unless I missed
something.
Any Gottchas ?
cheers
M
 
J

James Meyer

Jan 1, 1970
0
I need a 1Kv 200ma regulated variable supply, and thought that the
reference design in Win Hill's "A of E" was a good place to start. Trouble
is that MOSFETs hit the wall at 1Kv, wheras IGBTs are readily and cheaply
available up to 1200v (more than enough, the raw supply will be around
1160v). I cannot see any reason why the IGBT should not make a suitable
series pass element in a linear regulator like this - unless I missed
something.
Any Gottchas ?
cheers
M

I think there's one big "gottcha" up front. Using a 1200 Volt rated
device with "around" 1160 Volts applied is asking for trouble. You're only 40
Volts away from disaster.

Jim
 
J

John Larkin

Jan 1, 1970
0
I think there's one big "gottcha" up front. Using a 1200 Volt rated
device with "around" 1160 Volts applied is asking for trouble. You're only 40
Volts away from disaster.

Jim


If they didn't think it was reliable at 1200 volts, they wouldn't rate
it such. Discretes usually have lots of margin.

John
 
W

Winfield Hill

Jan 1, 1970
0
Mike Diack wrote...
I need a 1Kv 200ma regulated variable supply, and thought that the
reference design in Win Hill's "A of E" was a good place to start. Trouble
is that MOSFETs hit the wall at 1Kv, wheras IGBTs are readily and cheaply
available up to 1200v (more than enough, the raw supply will be around
1160v). I cannot see any reason why the IGBT should not make a suitable
series pass element in a linear regulator like this - unless I missed
something. Any Gottchas?

I have a nice stock of 1200V MOSFETs, so wouldn't consider IGBTs for
the task. The problem I've faced in making fast amplifiers has been
getting small enough transistors. In the old days 1n120 parts (small
1.2kV fets) were available, but these were supplanted by 1n100 (1kV)
parts, and by 3n120 parts (1.2kV, but larger die size). The problem
with big FETs is their higher capacitance, which forces one to run
at higher operating currents. Now when adding IGBTs into the mix, we
observe that for a given current rating they have smaller dies, and
hence lower capacitance than similar MOSFETs. But sadly, the IGBTs
generally aren't offered in truly low-current versions, so the "small"
IGBTs may end up less appealing the larger MOSFETs they might replace.
Another factor, important for higher-power linear amplifiers, is that
good heat removal capability => good power dissipation capability, is
in direct proportion to the die size => higher capacitance. Another
serious aspect is which parts can you get from a distributor somehow?
They may be "active" parts, but non-stocked. In addition, there are
the 1700 volt IGBTs to consider. Finally, tests on individual FETs
will reveal that many entire runs work well to several hundred volts
above their rated values. And that some in fact fail to reach their
rated values. (Motorola discontinued all FETs above 900V for a time
until they got their fab process back on track, and meantime shipped
non-compliant parts. Later they discontinued 1.2kV parts, but their
1kV parts in fact were the same and as good as the older 1.2kV parts.
Finally ON Semi discontinued the entire high-voltage line.)

Alternately one can make cascode high-voltage MOSFET amplifiers.

Yep, Mike, it's all a giant trade-off game.

So good luck, have lots of fun, and let us know how it works out.
 
W

Winfield Hill

Jan 1, 1970
0
John Larkin wrote...
If they didn't think it was reliable at 1200 volts, they wouldn't rate
it such. Discretes usually have lots of margin.

That's right, but as I mentioned in my other post, they can be all over
the place, and it's worthwhile when making high-voltage linear mosfet
amplifiers to check all the FETs. E.g., with a curve tracer. It's a
useful fact that the "failure" mode is a innocuous avalanche leakage.
 
M

Mike Diack

Jan 1, 1970
0
John Larkin wrote...

That's right, but as I mentioned in my other post, they can be all
over the place, and it's worthwhile when making high-voltage linear
mosfet amplifiers to check all the FETs. E.g., with a curve tracer.
It's a useful fact that the "failure" mode is a innocuous avalanche
leakage.
Ah ha - the curve tracer would indeed be a nice thing - such a nice
thing that it is what, in fact, I'm making :) Wasn't it Confucius that
said "To make a pair of tongs, you need a pair of tongs"
cheers
M
 
J

John Larkin

Jan 1, 1970
0
Ah ha - the curve tracer would indeed be a nice thing - such a nice
thing that it is what, in fact, I'm making :) Wasn't it Confucius that
said "To make a pair of tongs, you need a pair of tongs"
cheers
M

Aren't there some 1400-volt fets somewhere? I know there are 1400 volt
bipolars, horizontal output thingies.

John
 
P

Pooh Bear

Jan 1, 1970
0
Winfield said:
In the old days 1n120 parts (small
1.2kV fets) were available, but these were supplanted by 1n100 (1kV)
parts,

1N parts are diodes.

1N100 & 120 are germanium to boot !


Graham
 
P

Pooh Bear

Jan 1, 1970
0
Mike said:
I need a 1Kv 200ma regulated variable supply, and thought that the
reference design in Win Hill's "A of E" was a good place to start. Trouble
is that MOSFETs hit the wall at 1Kv, wheras IGBTs are readily and cheaply
available up to 1200v (more than enough, the raw supply will be around
1160v). I cannot see any reason why the IGBT should not make a suitable
series pass element in a linear regulator like this - unless I missed
something.

Simple answer.

I don't think you missed anything. They should be fine.

Graham
 
S

Spehro Pefhany

Jan 1, 1970
0
1N parts are diodes.

1N100 & 120 are germanium to boot !
Graham

Some MOSFETs follow a numbering scheme that consists of some
characters followed by a number indicating the Id current rating in
amperes, followed by an 'N' and then a number indicating the voltage
rating.

So, for example, a IXTP3N120 ('3N120') is a 3A 1200V MOSFET made by
IXYS. A '1N120' is a short form way to refer to a 1A rated 1200V
MOSFET- it's not the entire part number.

Best regards,
Spehro Pefhany
 
P

Pooh Bear

Jan 1, 1970
0
Spehro said:
Some MOSFETs follow a numbering scheme that consists of some
characters followed by a number indicating the Id current rating in
amperes, followed by an 'N' and then a number indicating the voltage
rating.

So, for example, a IXTP3N120 ('3N120') is a 3A 1200V MOSFET made by
IXYS. A '1N120' is a short form way to refer to a 1A rated 1200V
MOSFET- it's not the entire part number.

I've never seen that usage. Sounds kinda sloppy to me. You could drop the MTP
off the front of On Semi parts too if you wanted.

Part numbers beginning 1N, 2N, 3N etc... are JEDEC authorised. Any manufacturer
can use their own letter prefixed part designation ( excepting Pro-electron part
number designations of course ).


Graham
 
F

Frithiof Andreas Jensen

Jan 1, 1970
0
If they didn't think it was reliable at 1200 volts, they wouldn't rate
it such. Discretes usually have lots of margin.

Depends on who "they" are - the brochurepeople; i.e. Sales clearly, would.
As confirmed by the availablity of 1200V devices in TO220 packages!!

However, There is nothing wrong with sticking two 800V devices in series -
and supply the gate of the upper device with a resistive voltage divider
backed up with capacitors to supply the gate current. The drive is via the
lower device. A'la Cascode.

IGBT's have a "buglet" in that it is possible to make them latch-up on a
rapid overcurrent, which can happen easily in an HV supply, making is safer
to series-connect MOSFET's.
 
J

James Meyer

Jan 1, 1970
0
Ah ha - the curve tracer would indeed be a nice thing - such a nice
thing that it is what, in fact, I'm making :) Wasn't it Confucius that
said "To make a pair of tongs, you need a pair of tongs"
cheers
M

But.. but... but.... Aren't curve tracers easier to make if you don't
have to have regulated and/or programmable power supplies? Wouldn't the old
standard half-wave rectified transformer output along with voltage and current
measurements and a base or gate step generator be better?

Jim
 
W

Winfield Hill

Jan 1, 1970
0
Pooh Bear wrote...
I've never seen that usage. Sounds kinda sloppy to me. You could drop
the MTP off the front of On Semi parts too if you wanted.

If there was any sloppiness, it was leaving off a leading mark, would
'1n100 have satisfied you? As Spehro says, there're many suppliers of
nearly identical high-voltage FETs and IGBTs, with different prefixes.
A quick search in my computer shows these: ixtp, ixta, ixfh, fqp, fqd,
fqa, mtp, mtb, fqpf, ssw, sth, ixcp, ixrh, stp, mty, mtw, mty, ixgp,
irfp, irfb, spn, spw, skw, spd, spp, mgp, mgw, and mgy. There's more
in there, all followed by xNxx specific identifiers, you get the idea.
So when talking about power fets, it's clearly convenient to ignore the
prefix! And few folks will think we're talking about germanium diodes!
 
S

Spehro Pefhany

Jan 1, 1970
0
If there was any sloppiness, it was leaving off a leading mark, would
'1n100 have satisfied you? As Spehro says, there're many suppliers of
nearly identical high-voltage FETs and IGBTs, with different prefixes.
A quick search in my computer shows these: ixtp, ixta, ixfh, fqp, fqd,
fqa, mtp, mtb, fqpf, ssw, sth, ixcp, ixrh, stp, mty, mtw, mty, ixgp,
irfp, irfb, spn, spw, skw, spd, spp, mgp, mgw, and mgy. There's more
in there, all followed by xNxx specific identifiers, you get the idea.
So when talking about power fets, it's clearly convenient to ignore the
prefix! And few folks will think we're talking about germanium diodes!

Sometimes I really wish there was a way to search with regular
expressions on Google..

Best regards,
Spehro Pefhany
 
K

Ken Smith

Jan 1, 1970
0
I need a 1Kv 200ma regulated variable supply, and thought that the
reference design in Win Hill's "A of E" was a good place to start. Trouble
is that MOSFETs hit the wall at 1Kv, wheras IGBTs are readily and cheaply
available up to 1200v (more than enough, the raw supply will be around
1160v). I cannot see any reason why the IGBT should not make a suitable
series pass element in a linear regulator like this - unless I missed
something.
Any Gottchas ?

Some IGBTs are not rated for linear service. Check the data sheet of the
one you use.

Also IGBTs are slow to turn off. If the load is suddenly removed the
voltage will spring up above the set point.
 
W

Winfield Hill

Jan 1, 1970
0
Spehro Pefhany wrote...
Sometimes I really wish there was a way to search with regular
expressions on Google..

Hear! Hear!
 
T

Terry Given

Jan 1, 1970
0
Frithiof said:

ROTFLMAO!

lies, damn lies, and things salesmen put in brochures. cf rating Rdson
at 25V Tj.....or TO220 packages with 180W power ratings (and 1C/W Rthetajc)

It is rare for ANY manufacturer to tell the whole awful truth. Normally
they display only favourable characteristics, in whatever light makes
them look best, regardless of how the device will actually be used.
Depends on who "they" are - the brochurepeople; i.e. Sales clearly, would.
As confirmed by the availablity of 1200V devices in TO220 packages!!

However, There is nothing wrong with sticking two 800V devices in series -
and supply the gate of the upper device with a resistive voltage divider
backed up with capacitors to supply the gate current. The drive is via the
lower device. A'la Cascode.

IGBT's have a "buglet" in that it is possible to make them latch-up on a
rapid overcurrent, which can happen easily in an HV supply, making is safer
to series-connect MOSFET's.

why not use a 4kV IGBT? (actually, because the "little" ones tend to be
about 1,000 amps :)

1700V IGBTs are as common as all hell.

a 1200V IGBT is good in a switching app up to about 900Vdc, depending on
stray inductance and dI/dt. Many IGBT-based products self-destruct when
turning off under fault conditions, as dI/dt is often about 10x larger
than normal - although a gatedrive that uses a much larger Roff under
fault (ie desat) conditions helps a LOT (Semikron do this in their gate
driver modules). In a linear app, dI/dt should be a hell of a lot lower,
so you ought to be able to push it well over 1kV, but dont forget fault
scenarios - shorts followed by opens etc.

My tech once built me a 0-400V linear supply that ran from 900Vdc using
a 1200V 600A IGBT - for doing splat tests on very large chokes (it
allowed us to suck almost all the energy out of the HVDC bus caps whilst
supplying constant voltage across the DUT). It worked quite well, BUT
igbts are designed for switching rapidly, so can be prone to oscillation
when used in a linear mode.


cheers
Terry
 
Top