Super duper hype fast FET driver?

[Joerg]

Folks,

Have to drive around a hundred pF or so in parallel with maybe a few
hundred ohms blazingly fast. 10-12V amplitude, transitions ideally
sub-nanosecond from 10-90% both directions. Not too boutiquish or
unobtanium (which excludes certain companies ...) and not more than a
few Dollars. Shouldn't introduce noise when low. Low quiescent current,
preferably under 20mA.

Looked around and the fastest one I could see is this dude:

http://www.diodes.com/datasheets/ZXGD3003E6.pdf

Not too much data but it gets into the 2nsec range. Anyone aware of a
driver with even more testosterone?

Of course the ideal scenario would be a push-pull MMIC with 15V or more
supply voltage but that ain't going to happen. Same with RF switches,
even if they could take 12V they are surprisingly slow in switch action.

Ok, could roll my own, of course. But that gets to be involved and a bit
too big for this project.

 

[Joerg]

The rise/fall times are mediocre. A little series inductive peaking
would help.

Mediocre? At high capacitive load maybe but at low loads I think they
are amazing for a 25c part. Most others including its other ZXGD300x
brethren peter out towards 10nsec for one of both slopes towards zero
capacitance.

Gate driver chips are cool, though. They have lots of off-label uses.

Oh yeah

What's the load? Need DC coupling/long pulses?

Pulses are under 100nsec length and duty cycle is well under 5%. Getting
DC across would be great but I am afraid I'll have to clamp that ...
somehow ... because it's a high voltage I'll have to move with that. Or
I'll let this whole thing ride on the rail. Woe to dose who toucha wid
da fingahs

I know how to get 6 or 7 volts of brutal sub-ns swing cheap, at zero
Iq. Could be doubled, depending. Not for public release.

7V would be a bit low, and I'd need zippy in both directions.

 

[DonMack]

"Joerg" wrote in message
Folks,

Have to drive around a hundred pF or so in parallel with maybe a few
hundred ohms blazingly fast. 10-12V amplitude, transitions ideally
sub-nanosecond from 10-90% both directions. Not too boutiquish or
unobtanium (which excludes certain companies ...) and not more than a
few Dollars. Shouldn't introduce noise when low. Low quiescent current,
preferably under 20mA.

Looked around and the fastest one I could see is this dude:

http://www.diodes.com/datasheets/ZXGD3003E6.pdf

Not too much data but it gets into the 2nsec range. Anyone aware of a
driver with even more testosterone?

Of course the ideal scenario would be a push-pull MMIC with 15V or more
supply voltage but that ain't going to happen. Same with RF switches,
even if they could take 12V they are surprisingly slow in switch action.

Ok, could roll my own, of course. But that gets to be involved and a bit
too big for this project.

 

[Joerg]

If you need high-side drive, a pair of my drivers could feed opposite
ends of a 1:1 transmission line transformer (or maybe a balun) to
double the swing. 12 volts for 100 ns would be no problem, edges well
below a ns.

RF parts such as baluns or obscure directional couplers with a voodoo
coefficient never scared me.

At below 5% duty cycle, you probably don't need to DC restore.

Not sure yet if I do. This high voltages acts as kind of a reference
somewhere. It's one of those circuits that most people won't even touch
with a 10ft pole.

It'll cost you a burger and a couple of beers. Zeitgeist has about 40
on tap.

I am game. But I have no idea when I get to S.F. the next time. A lot of
my clients are no longer in California, for reasons we all know, and
Villaraigosa has just made that loud and clear to companies once more :-(

Now if you ever get into this area we have a very nice Japanese
restaurant. Blue Moon and Sapporo on tap.

If pulse width is fixed, there's another way to do it: a really fast
open-drain mosfet feeding a 1:1 transformer, 12 volts or whatever you
like. The magnetizing current powers the falling edge. Also cheap.

It must remain adjustable.

 

[Joerg]

"Joerg" wrote in message
Folks,

Have to drive around a hundred pF or so in parallel with maybe a few
hundred ohms blazingly fast. 10-12V amplitude, transitions ideally
sub-nanosecond from 10-90% both directions. Not too boutiquish or
unobtanium (which excludes certain companies ...) and not more than a
few Dollars. Shouldn't introduce noise when low. Low quiescent current,
preferably under 20mA.

Looked around and the fastest one I could see is this dude:

http://www.diodes.com/datasheets/ZXGD3003E6.pdf

Not too much data but it gets into the 2nsec range. Anyone aware of a
driver with even more testosterone?

Of course the ideal scenario would be a push-pull MMIC with 15V or more
supply voltage but that ain't going to happen. Same with RF switches,
even if they could take 12V they are surprisingly slow in switch action.

Ok, could roll my own, of course. But that gets to be involved and a bit
too big for this project.

This isn't a gate but another capacitive load. The goal is to make the
pulse waveform look as perfectly square as can be.

Wish I could find something like the NE3509 that can take 15-20V instead
of just 4V.

 

[Joerg]

They are actually happy to run around 8 volts. Lovely part. But heck,
a 2N7002 can switch 15 or 20 volts in under a ns... if you bang the
gate hard enough.

The 2N7002 or BSS123 plus some magnetics would be kind of my last resort
if none of the ICs can do that. And it looks like none can. No market I
guess.

Some of the mesfets can be run at 12 or even 15 volts. But the bees
knees these days is GaN fets, which work in the 40-200 volt range,
need tiny gate drives, and switch insanely fast. But not cheap.

Doesn't have to be cheap, but my definition of expensive is everything
north of $5

I'll have to see what CEL has to offer. Paralleling European dual-gates
for UHF tuners might also work but that gets old.

We blew up about a kilobuck worth of Nitronex fets one afternoon. They
claim they have the process down now.

Ouch!

 

[Joerg]

Never turn down an offer from Larkin. He's promised you a half-fast solution,
and I'm sure he can deliver ;-)

I am sure he can. Fast stuff is what his company mostly does and if they
didn't earn megabucks per year with this they would hardly be able to
buy a large building smack dab in the middle of a metropolis.

 

[Joerg]

It was a working fortune cookie factory. 1930's (ie, terrible)
concrete side walls, wooden floors and roof, a perfect candidate for
pancaking in the next good earthquake. It's sitting on sand, at the
edge of a liquefaction zone. We added a bunch of concrete footings and
steel frames and plywood and bolts.

You can see our place in the default Google Earth view of San
Francisco. For some odd reason, Google thinks "San Francisco" is in
the middle of the interesction of Market and Van Ness, one block away.

And we haven't had a sandstorm for as long as we've been here.

http://www.panoramio.com/photo/57596123

Those cherubim were on there before you bought it? They are amazingly
well preserved. I can't imagine the salty air down there to be too
friendly to concrete and plaster. The almond below probably has some
cultural meaning as well.

 

[Joerg]

They are original, and we just cleaned them up and repainted. No
architect would dare do anything this cool any more.

The salty fog is out west, near the ocean. Everything rusts out there.

The bummer is that it's a wooden building that looks up at Sutro
Tower, 22 megawatts of AM/FM/TV/whatever. The EMI is ghastly.

Look at the bright side: If you'd need a sensor somewhere out of reach
of power you can just add a loop, some diodes, and it'll be supplied
with power until Sutro Tower keels over

If I'd needed a quiet lab place I'd find something in the outbacks of
Alabama or similar states. Then you are neither bothered by RF fields
nor by biz-hostile politicians.

 

[Joerg]

On the other hand, the chances of being able to hire locally resident
expert help wouldn't be that great.

Near Huntsville? You've got to be kidding ...

... And they are Baptists - perhaps not
as sincerely Baptist as the inhabitants of Urk are Calvinist, but still
pretty inflexible - so you'd run the risk of being rejected as a
schismatic Lutheran.

Baptists and Lutherans get along quite well, and I love their choirs.
They can make the rafters shake.

 

[Joerg]

Jim Thompson wrote:

John Larkin wrote:
John Larkin wrote:
Folks,

Have to drive around a hundred pF or so in parallel with maybe a few
hundred ohms blazingly fast. 10-12V amplitude, transitions ideally
sub-nanosecond from 10-90% both directions. Not too boutiquish or
unobtanium (which excludes certain companies ...) and not more than a
few Dollars. Shouldn't introduce noise when low. Low quiescent current,
preferably under 20mA.

Looked around and the fastest one I could see is this dude:

http://www.diodes.com/datasheets/ZXGD3003E6.pdf

Not too much data but it gets into the 2nsec range. Anyone aware of a
driver with even more testosterone?
The rise/fall times are mediocre. A little series inductive peaking
would help.

Mediocre? At high capacitive load maybe but at low loads I think they
are amazing for a 25c part. Most others including its other ZXGD300x
brethren peter out towards 10nsec for one of both slopes towards zero
capacitance.


Gate driver chips are cool, though. They have lots of off-label uses.

Oh yeah


Of course the ideal scenario would be a push-pull MMIC with 15V or more
supply voltage but that ain't going to happen. Same with RF switches,
even if they could take 12V they are surprisingly slow in switch action.

Ok, could roll my own, of course. But that gets to be involved and a bit
too big for this project.
What's the load? Need DC coupling/long pulses?

Pulses are under 100nsec length and duty cycle is well under 5%. Getting
DC across would be great but I am afraid I'll have to clamp that ...
somehow ... because it's a high voltage I'll have to move with that. Or
I'll let this whole thing ride on the rail. Woe to dose who toucha wid
da fingahs


I know how to get 6 or 7 volts of brutal sub-ns swing cheap, at zero
Iq. Could be doubled, depending. Not for public release.

7V would be a bit low, and I'd need zippy in both directions.
If you need high-side drive, a pair of my drivers could feed opposite
ends of a 1:1 transmission line transformer (or maybe a balun) to
double the swing. 12 volts for 100 ns would be no problem, edges well
below a ns.

RF parts such as baluns or obscure directional couplers with a voodoo
coefficient never scared me.


At below 5% duty cycle, you probably don't need to DC restore.

Not sure yet if I do. This high voltages acts as kind of a reference
somewhere. It's one of those circuits that most people won't even touch
with a 10ft pole.


It'll cost you a burger and a couple of beers. Zeitgeist has about 40
on tap.

I am game. But I have no idea when I get to S.F. the next time. A lot of
my clients are no longer in California, for reasons we all know, and
Villaraigosa has just made that loud and clear to companies once more :-(

Now if you ever get into this area we have a very nice Japanese
restaurant. Blue Moon and Sapporo on tap.


If pulse width is fixed, there's another way to do it: a really fast
open-drain mosfet feeding a 1:1 transformer, 12 volts or whatever you
like. The magnetizing current powers the falling edge. Also cheap.

It must remain adjustable.
Never turn down an offer from Larkin. He's promised you a half-fast solution,
and I'm sure he can deliver ;-)

I am sure he can. Fast stuff is what his company mostly does and if they
didn't earn megabucks per year with this they would hardly be able to
buy a large building smack dab in the middle of a metropolis.

It must really stick in his craw when he learned that a good many of the
ECL/PECL parts he buys from AZMicrotek were designed by yours truly ;-)

...Jim Thompson

[On the Road, in New York]

Why would that bother me? You seem to be OK at linear IC design.

I think we buy one part from AZM, a PECL/TTL converter, and few of
those. The rest of their stuff is too slow. And we don't do much ECL
any more; the gates inside an FPGA are a lot faster at 1/1000 the
cost.

And I wouldn't use that AZM part in a new design; there are faster and
cheaper ways to do it.

I think I am going the old 2N7002 plus transformer route, still hoping
to find faster ones along the way. Maybe some dual-gate FETs.

But: Now I need complementary drives with gusto, to do the ping-pong
drive. 5V min, sub-nsec transitions, source impedance each 20ohms or
less. Ok, I could concoct something around 5GHz+ BJTs but when the guys
see all that they are going to get sick.

What's the brute force logic du jour these days if one wants to stay
clear of FPGA and in particular BGAs? LVC? The datasheets are flimsy, no
rise and fall times, no info what it looks when loading the output to
the hilt, only prop delays.

 

[Spehro Pefhany]

Why would that bother me? You seem to be OK at linear IC design.

I think we buy one part from AZM, a PECL/TTL converter, and few of
those. The rest of their stuff is too slow. And we don't do much ECL
any more; the gates inside an FPGA are a lot faster at 1/1000 the
cost.

And I wouldn't use that AZM part in a new design; there are faster and
cheaper ways to do it.

John

Have you used any of the SiGe logic parts? eg.

http://www.onsemi.com/pub_link/Collateral/NBSG86A-D.PDF

40ps rise/fall.




Best regards,
Spehro Pefhany

 

[[email protected]]

Almond? I thought it was a tamale.

Wouldn't a fortune cookie factory be more likely to have a grain of rice on
the front wall?

 

[[email protected]]

Near Huntsville? You've got to be kidding ...

Slowman is never shy about his ignorance.

Baptists and Lutherans get along quite well, and I love their choirs.
They can make the rafters shake.

Yup (wife belongs to a Baptist church here). She's not in the choir, though.
She doesn't like the director.

 

[Joerg]

Wouldn't a fortune cookie factory be more likely to have a grain of rice on
the front wall?

At least it'll be super-duper feng shui'd

 

[Joerg]

Have you used any of the SiGe logic parts? eg.

http://www.onsemi.com/pub_link/Collateral/NBSG86A-D.PDF

40ps rise/fall.

And the rise/fall times are in the datasheet, in contrast to regular logic.

Now if they had 5V swing I could drive my FETs with these ... <drool>

 

[[email protected]]

The building was plastered with the arguably racist signs like

ftp://jjlarkin.lmi.net/Lotus.jpg

but the Louie family put them up, so I guess it was an inside joke or
something.

Looks like a fortune cookie with eyes.

Edward Louie invented the fortune-cookie making machine in
1974.

ftp://jjlarkin.lmi.net/Fortune_Cookie_Machine.jpg

Is that your SMT line now?

 

[Joerg]

The building was plastered with the arguably racist signs like

ftp://jjlarkin.lmi.net/Lotus.jpg

Yikes. Did anyone ever dare to stand on that balcony? I'd put up a sign
"Max load 0.1 ton".

but the Louie family put them up, so I guess it was an inside joke or
something. Edward Louie invented the fortune-cookie making machine in
1974.

ftp://jjlarkin.lmi.net/Fortune_Cookie_Machine.jpg

Is it still there somewhere?

Good book: The Fortune Cookie Chronicles, all about Chinese food and
the Chinese restaurant industry.

We had lunch at our Japanese place. Lagunitas IPA on tap, yumm. Recently
they started handing out fortune cookies, too. But I prefer the Korean
ice cream.

http://kobesushiandgrill.com/

 

[Joerg]

isn't that what the ibis models are for?

Real datasheets should have that in there. Just like they do for really
fast ECL stuff.

 

[Joerg]

Can I interest you in 6 volts, hard 560 ps rise/fall, 12 cents?

ftp://jjlarkin.lmi.net/ABneg.jpg

ftp://jjlarkin.lmi.net/Rise.jpg

Shazam! That does look impressive.