Power FETs with low Vgs?

[Joerg]

Hello All,

Micro controllers and logic are moving to ever lower supply
voltages. This leads to a worsening problem: How to reliably
control large FETs.

An example would be a PWM switcher realized with a uC. If that uC
can't take more than 3 to 4 volts there are but a few options.
One, of course, is to provide an external driver/translator
circuit or chip. Another would be a little pulse transformer. The
latter burdens the controller chip quite a bit because the gate
capacitance translates with the voltage. Most of all, both
solutions add cost.

Any ideas? Any new devices out there that are reasonably cost
efficient?

Regards, Joerg
http://www.analogconsultants.com

 

[Tim Wescott]

Joerg wrote:

Hello All,

Micro controllers and logic are moving to ever lower supply
voltages. This leads to a worsening problem: How to reliably
control large FETs.

An example would be a PWM switcher realized with a uC. If that uC
can't take more than 3 to 4 volts there are but a few options.
One, of course, is to provide an external driver/translator
circuit or chip. Another would be a little pulse transformer. The
latter burdens the controller chip quite a bit because the gate
capacitance translates with the voltage. Most of all, both
solutions add cost.

Any ideas? Any new devices out there that are reasonably cost
efficient?

Regards, Joerg
http://www.analogconsultants.com

Vishay/Siliconix rates their "logic" devices with 3.3V gate drive. The
one I looked at had about a 2x difference in Rds between 3.3V and 5V,
but it _was_ specified.

I think the microcontrollers that could drive that big gate charge
reasonably fast would be few and far between. I'd check closely before
I tried to drive a gate quickly from a microcontroller pin.

--

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

 

[Joerg]

Hi Tim,

Yes, their Si4378DY is rated under 5 milliohms at 2.5 gate drive (at a
whopping 22 amps). It's a low voltage device (20V max), but it is a
beginning.

You are right, some really big loads are too much for a uC. Expecially in a
switcher where the GD capacitance causes the FET to sputter because the
driver can't hold it on. But in that case a pair of BSS123/BSS84 can
increase the oomph which won't add more than 10 or 12 cents. Still that
will not increase the voltage unless you spring for a few more transistors
to make an amp.

Thing is, many low cost apps can have a PWM controller or a uC. But not
both because that would blow the budget.

Regards, Joerg
http://www.analogconsultants.com

 

[Terry Given]

"Joerg" <notthisjoergsch@removethispacbell.net> wrote in message
news:40996D39.920336F@removethispacbell.net...

Hi Tim,

Yes, their Si4378DY is rated under 5 milliohms at 2.5 gate drive (at a
whopping 22 amps). It's a low voltage device (20V max), but it is a
beginning.

You are right, some really big loads are too much for a uC. Expecially in
a
switcher where the GD capacitance causes the FET to sputter because the
driver can't hold it on. But in that case a pair of BSS123/BSS84 can
increase the oomph which won't add more than 10 or 12 cents. Still that
will not increase the voltage unless you spring for a few more transistors
to make an amp.

Thing is, many low cost apps can have a PWM controller or a uC. But not
both because that would blow the budget.

Regards, Joerg

make sure the Vt of your FETs stays above Volmax, at maximum Tj. I once
fixed a problem just like this - LM339 couldnt turn the fet off at high
temperture. Some of the fets I looked at had Vt below 200mV...

 

[Kevin McMurtrie]

In article <40994E06.746394AF@removethispacbell.net>,
Joerg <notthisjoergsch@removethispacbell.net> wrote:

Hello All,

Micro controllers and logic are moving to ever lower supply
voltages. This leads to a worsening problem: How to reliably
control large FETs.

An example would be a PWM switcher realized with a uC. If that uC
can't take more than 3 to 4 volts there are but a few options.
One, of course, is to provide an external driver/translator
circuit or chip. Another would be a little pulse transformer. The
latter burdens the controller chip quite a bit because the gate
capacitance translates with the voltage. Most of all, both
solutions add cost.

Any ideas? Any new devices out there that are reasonably cost
efficient?

Regards, Joerg
http://www.analogconsultants.com

What happens when the MOSFET is off and the drain voltage suddenly drops
several volts? The gate-drain coupling will pull the microprocessor's
output below 0V. I've blown out plenty of CMOS 555 timer chips by
connecting them directly to power MOSFET.

Somebody must make a MOSFET with an integrated driver by now. Everybody
keeps asking for one.

 

[Winfield Hill]

Kevin McMurtrie wrote...

What happens when the MOSFET is off and the drain voltage suddenly
drops several volts? The gate-drain coupling will pull the
microprocessor's output below 0V. I've blown out plenty of CMOS
555 timer chips by connecting them directly to power MOSFET.

That problem is easily solved.

If there's a fast dip in the drain voltage, sufficient to drive
the gate driver's parasitic ground diode, the diode current is
i = Crss dV/dt. As an example, the IRFZ48V is a fairly-popular
and inexpensive ($0.58 qty 1k) 60V FET, rated at 72A. Its Crss
capacitance is 91pF at 25V, so the dV/dt must exceed 550V/us to
create a gate-current greater than 50mA (a common spec to stay
below to avoid CMOS SCR latchup). 550V/us is a pretty high
rate-of-change to see on a properly-bypassed power-supply rail,
so that's why many folk would not see the problem. (And if the
load was prone to such high dV/dt rates, the FET should probably
have a snubber affixed.)

But if such a fast-step possibility existed, or one was simply
playing it safe, adding a 100-ohm resistor in series with the
gate would probably solve the problem. And certainly adding a
Schottky diode to ground would solve it. These two items are
commonly seen.

Somebody must make a MOSFET with an integrated driver by now.
Everybody keeps asking for one.

There are many. Most include not only a FET driver (usually
including a charge pump and level-shifting circuitry), but
current-sensing, over-temperature limiting, fault indication,
etc. These are intelligent power switches, and are often called
high-side switches. For example, Infineon (formerly Siemens)
offers the BTS555, an impressive 44V 165A switch. They make a
whole family of these switches, and DigiKey stocks some of them.

For example, the BTS442E2 is a 42V 25A TO-220 switch selling for
$6.14 each, and the BTS409L1 is a 43V 2.3A switch in a TO-220
package selling for $2.82 each at DigiKey.

Thanks,
- Win

(email: use hill_at_rowland-dot-org for now)

 

[markp]

What happens when the MOSFET is off and the drain voltage suddenly drops
several volts? The gate-drain coupling will pull the microprocessor's
output below 0V. I've blown out plenty of CMOS 555 timer chips by
connecting them directly to power MOSFET.

Somebody must make a MOSFET with an integrated driver by now. Everybody
keeps asking for one.

OnSemi do their SmartFET series.

Mark.

 

[Fred Bartoli]

"Joerg" <notthisjoergsch@removethispacbell.net> a écrit dans le message
news: 40996D39.920336F@removethispacbell.net...

Hi Tim,

Yes, their Si4378DY is rated under 5 milliohms at 2.5 gate drive (at a
whopping 22 amps). It's a low voltage device (20V max), but it is a
beginning.

You are right, some really big loads are too much for a uC. Expecially in
a
switcher where the GD capacitance causes the FET to sputter because the
driver can't hold it on. But in that case a pair of BSS123/BSS84 can
increase the oomph which won't add more than 10 or 12 cents. Still that
will not increase the voltage unless you spring for a few more transistors
to make an amp.

Thing is, many low cost apps can have a PWM controller or a uC. But not
both because that would blow the budget.

Regards, Joerg
http://www.analogconsultants.com

If you want high voltage&current capability driving directly from a uc and
without a mosfet driver (the only 2 reasons I can see are 1)you can't afford
cost 2)you really don't want to stock a new item) and your VCC is made from
a highish 10 or more volts then you can go cascode.

A low voltage low rdson mos will add little to the high voltage mosfet
rdson, while the latter will nicely isolate the uc from the drain voltage.


Small quizz :
I once had to build a +/- 20V@1A supply from a +15V rail with as little
components as possible and with as standard components as possible (only
buck switchers in stock).
How would you all do this ?


Thanks,
Fred.

 

[John Larkin]

On Wed, 05 May 2004 20:22:06 GMT, Joerg
<notthisjoergsch@removethispacbell.net> wrote:

Hello All,

Micro controllers and logic are moving to ever lower supply
voltages. This leads to a worsening problem: How to reliably
control large FETs.

An example would be a PWM switcher realized with a uC. If that uC
can't take more than 3 to 4 volts there are but a few options.
One, of course, is to provide an external driver/translator
circuit or chip. Another would be a little pulse transformer. The
latter burdens the controller chip quite a bit because the gate
capacitance translates with the voltage. Most of all, both
solutions add cost.

Any ideas? Any new devices out there that are reasonably cost
efficient?

Regards, Joerg
http://www.analogconsultants.com

Don't tell anybody I suggested this, but I've been known to bias up a
fet a couple of volts - just below the spec'd threshold - so my logic
swing turns it on harder. Say, 2.5 to 7.5 swing instead of 0-5.

If you quote me on this, I'll deny it.

John

 

[Joerg]

Hi Fred,

Quiz: I'd probably sneak out and get an LM3478, it's small enough that it can be
hidden under an electrolytic so nobody sees... Then make a flyback converter
with it which wouldn't take more than a dozen parts or so. Maybe place but not
connect one of those buck converter atop the whole thing just as camouflage.

Regards, Joerg
http://www.analogconsultants.com

 

[Joerg]

Hi Kevin,

That sure is a problem. Schottly protection would work here but most FETs
don't have any so it would have to come as discretes. A FET driver pair
would fix that as well but now you are looking at two SOT-23 devices.

Regards, Joerg
http://www.analogconsultants.com

 

[Joerg]

Hi John,

Wow! Scary. But you are in good company. This is not a joke: I have seen
a properly released manufacturing instruction from a major appliance
manufacturer that contained a detailed procedure for aligning numerous
filters in an amp. It concluded with the words: An then work on it until
it worketh....

Regards, Joerg
http://www.analogconsultants.com

 

[Joerg]

Hi Winfield,

The only issue with the BTS versions with built in drivers is that their
cost is often out of range for a consumer or disposable application.
That cost pressure (and sometimes size) is really the main reason why
integrating PWM within the uC is considered. One Dollar is a whole lot
of money, in those designs I sometimes spend an hour just to replace a 6
cent MOSFET with a 3 cent bipolar. Every cent gets turned around and
around.

Regards, Joerg
http://www.analogconsultants.com

 

[Winfield Hill]

Joerg wrote...

The only issue with the BTS versions with built in drivers is that their
cost is often out of range for a consumer or disposable application.
That cost pressure (and sometimes size) is really the main reason why
integrating PWM within the uC is considered. One Dollar is a whole lot
of money, in those designs I sometimes spend an hour just to replace a 6
cent MOSFET with a 3 cent bipolar. Every cent gets turned around and
around.

I agree, a cell phone, etc., is going to be very cost driven.
But some of these parts aren't bad for other similar uses, e.g.,
the BTS428L2, a 7A 41V intelligent switch in a small Dpak-5 smt
package at $1.32 qty 5k from Digi-Key, less elsewhere no doubt.
http://rocky.digikey.com/WebLib/Infineon/Web%20Data/BTS428L2.pdf

Thanks,
- Win

(email: use hill_at_rowland-dot-org for now)

 

[Jan Panteltje]

On a sunny day (Wed, 05 May 2004 20:22:06 GMT) it happened Joerg
<notthisjoergsch@removethispacbell.net> wrote in
<40994E06.746394AF@removethispacbell.net>:

Hello All,

Micro controllers and logic are moving to ever lower supply
voltages. This leads to a worsening problem: How to reliably
control large FETs.

An example would be a PWM switcher realized with a uC. If that uC
can't take more than 3 to 4 volts there are but a few options.
One, of course, is to provide an external driver/translator
circuit or chip. Another would be a little pulse transformer. The
latter burdens the controller chip quite a bit because the gate
capacitance translates with the voltage. Most of all, both
solutions add cost.

Any ideas? Any new devices out there that are reasonably cost
efficient?

Regards, Joerg
http://www.analogconsultants.com
I am using some cheap ones, see irlz34n.pdf (google), these open

at 3 V (several amps), 3.3V 10 A (looking at curve).
Works fine on a PIC.
JP

 

[Joerg]

Hi Winfield,

$1.32 for a BTS428 isn't bad indeed. It's turn on threshold is 1.7V to 3.2V
(kinda close for a 3.3V driver) and might work but these are for really slow
stuff like turning on lamps or motors. They need a hundred usec or so to
switch state.

It's not just cell phones that need to be designed with every penny pinched.
There are lots of products even in the medical world that are one-time use, or
have to retail directly to consumers. Then there are all those household
gizmos such as automatic or remote appliance control. Usually the pain
threshold for a lot of this is the magic $19.95 number. If the store price
goes above that consumers tend to ponder a lot and often decide "nah, don't
need it that urgently". Realistically with all the distribution markups this
leaves $2 to $3 for materials.

Regards, Joerg
http://www.analogconsultants.com

 

[Joerg]

Hi John,

Thanks, just looked at it. Interestingly, the IRF server couldn't find
IRLZ34N but an Australian university had the data sheet.

The lowest Vgs where they have it spec'd is 4V and the Id/Vgs curve
looks awfully steep around 3V. I like stuff where they have a 3V hard
spec in the component data (min, max and all) because the curves
usually represent typical values. I just want to avoid the scenario
where a batch comes along where the process had hovered at the other
end, everything gets hot and then kablooouie.

Regards, Joerg
http://www.analogconsultants.com

 

[Ken Smith]

In article <409a58fe$0$17900$626a14ce@news.free.fr>,
Fred Bartoli <fred._canxxxel_this_bartoli@RemoveThatAlso_free.fr_AndThisToo> wrote:
[....]

Small quizz :
I once had to build a +/- 20V@1A supply from a +15V rail with as little
components as possible and with as standard components as possible (only
buck switchers in stock).
How would you all do this ?

There are some makers that make inductors with 2 windings. They intend
you to put them in series or parallel to get different inductances but you
don't have to. You can use them for the two outputs.

I hope your bucker chips are the ones that allow the switch voltage to
swing below ground or are rated at more than 35V.

If the switch voltage on the bucker can't go below ground, use the -V as
the ground of the switcher chip.

The simple (+) to (-) converter has 20V + a diode on the winding when it
is delivering power to the load side. The other winding can be delivering
the +20V

Is that what you did?


--
--
kensmith@rahul.net forging knowledge

 

[Winfield Hill]

Fred Bartoli wrote...

Small quizz :
I once had to build a +/- 20V@1A supply from a +15V rail with as
little components as possible and with as standard components as
possible (only buck switchers in stock).
How would you all do this ?

Generally to step up from say +15 to +20 we use boost switchers
to provide the required extra 5V from a flyback inductor. For
these switchers the voltage on the inductor is a 21V square wave,
so it's possible to add a capacitor and two diodes to obtain -20V.
Or one may use a second winding on the inductor, saving a diode
and one cap. Doing this job with a buck switcher (if that's what
you meant to say) would imply using a transformer rather than an
inductor anyway.

Thanks,
- Win

(email: use hill_at_rowland-dot-org for now)

 

[Fred Bartoli]

"Joerg" <notthisjoergsch@removethispacbell.net> a écrit dans le message
news: 409A7597.C6815791@removethispacbell.net...

Hi Fred,

Quiz: I'd probably sneak out and get an LM3478, it's small enough that it
can be
hidden under an electrolytic so nobody sees... Then make a flyback
converter
with it which wouldn't take more than a dozen parts or so. Maybe place but
not
connect one of those buck converter atop the whole thing just as
camouflage.

He he, well tried Joerg...
Unfortunately the client had heavy procedures for validation of components
to be inserted in their database and this one was for too few products. So
it was a matter of : use what we already have or die.
Plus it was 5 or 6 years ago this nice LM3478 is freshly designed.

So I'm afraid you'll have to do it with a buck converter...

Thanks,
Fred.