Win's next 10kV project, a 1us ramp

[Tony Williams]

However, the tube approach is still rather complicated, because
the "grounded grid" tube with its programmable current-sinking
MOSFET cathode driver (500V, 1A - not too bad) and its 50W
filament power, will all be sitting at -16kV. Hmm, that better
all be inside a big Faraday cage with rounded corners. Can't
have corona discharge off of our sharp-edged components! Maybe
we can use a clear housing, or add some windows.

snip a few minor problems that would loosen my bowels.....

Since it is a capacitive load could you run everything
up at 0v+positive, and capacitively-couple the negative
voltage ramp to the load?

-+- +12 to 13KV
|
R
|
+-----------+
___|___ |
|
-------+ ? |
0V-+------- |
_______ |
/ |
| Heater |
|--+ |
|MOSFET |
-----||--+ |
| |
R |
| |
-+--0v ===C
| |
[-3KV supply] |
| D | Electrode
+--------+--|<|--+-------> 50pF total
| |
+---R---+

 

[Winfield Hill]

Tony Williams wrote...

However, the tube approach is still rather complicated, because
the "grounded grid" tube with its programmable current-sinking
MOSFET cathode driver (500V, 1A - not too bad) and its 50W
filament power, will all be sitting at -16kV. Hmm, that better
all be inside a big Faraday cage with rounded corners. Can't
have corona discharge off of our sharp-edged components! Maybe
we can use a clear housing, or add some windows.

snip a few minor problems that would loosen my bowels.....

Since it is a capacitive load could you run everything
up at 0v+positive, and capacitively-couple the negative
voltage ramp to the load?

-+- +12 to 13KV
|
R
|
+-----------+
___|___ |
|
-------+ ? |
0V-+------- |
_______ |
/ |
| Heater |
|--+ |
|MOSFET |
-----||--+ |
| |
R |
| |
-+--0v ===C
| |
[-3KV supply] |
| D | Electrode
+--------+--|<|--+-------> 50pF total
| |
+---R---+

Hey, Tony, great idea, why didn't I think of that? We'll have a 0.1%
to 5% duty cycle at most (1kHz rep rate), so that should be just fine.
I can use the DC restoration you show. It's a huge simplification,
and limits the extreme HV to the plate and a few other items.

Good show!

 

[John Larkin]

Tony Williams wrote...

However, the tube approach is still rather complicated, because
the "grounded grid" tube with its programmable current-sinking
MOSFET cathode driver (500V, 1A - not too bad) and its 50W
filament power, will all be sitting at -16kV. Hmm, that better
all be inside a big Faraday cage with rounded corners. Can't
have corona discharge off of our sharp-edged components! Maybe
we can use a clear housing, or add some windows.

snip a few minor problems that would loosen my bowels.....

Since it is a capacitive load could you run everything
up at 0v+positive, and capacitively-couple the negative
voltage ramp to the load?

-+- +12 to 13KV
|
R
|
+-----------+
___|___ |
|
-------+ ? |
0V-+------- |
_______ |
/ |
| Heater |
|--+ |
|MOSFET |
-----||--+ |
| |
R |
| |
-+--0v ===C
| |
[-3KV supply] |
| D | Electrode
+--------+--|<|--+-------> 50pF total
| |
+---R---+

Hey, Tony, great idea, why didn't I think of that? We'll have a 0.1%
to 5% duty cycle at most (1kHz rep rate), so that should be just fine.
I can use the DC restoration you show. It's a huge simplification,
and limits the extreme HV to the plate and a few other items.

Good show!

So now you run the fet source at ground, through a sense resistor
maybe, and the tube grid somewhat positive, whatever it takes to get
your amp plus some fet headroom.

Run the tube filament through a dc/dc (or dc/ac!) converter to keep
from having a lot of capacitance in a filament transformer.

Or, build my pseudo-Marx thing with hv fets. But the fet-tube cascode
is quaint, so that's more fun.

John

 

[John Larkin]

Winfield Hill wrote...

Here's one approach to extending the circuit. You can see the
complexity grows rapidly as cascode stages are added, because
a new current sink is required for each gate-biasing capacitor,
and these also have to be cascoded.

. -3kV
. |
. 1mA
. | out
. -4.5kV ______________ +--------------------,
. | | | |
. _|_ | | ___ -3.0 |
. --- 0.5mA | \ |
. Co | | | \ _|_
. -6.0kV ______ |____|_______||-' \___-7.5kV --- Co
. | | | | ||-, |
. _|_ | | | 10V | -4.5kV ramp |
. --- 0.3mA | '--|<|--+ gnd
. Co | | | |
. |____|_______||-' __________||-'
. | | ||-, | ||-,
. __ | | 10V | | 10V |
. __| |__ | '--|<|--+ '--|<|--+
. --+---------- |---+---------- | --, |
. | _ | | _ | | _ |
. '--| \___||-' '--| \___||-' '--| \___||-'
. ,--|_/ ||-, ,--|_/ ||-, ,--|_/ ||-,
. | | | | | |
. '-----------+ '-----------+ '-----------+
. | | |
. 3 R 2 R 2 R
. | | |
. --------------+---------------+---------------+--- -7.5kV

At this point, with only a 4.5kV capability, the MOSFET circuit
may still be more simple than the tube circuit, but extending
the scheme to 6kV would probably tip the balance.

If I could figure out a way to stack the gate-bias capacitors...

Do the Larkin-Marx circuit. Ten stages, all initially at ground, each
a mosfet ramp generator on a small pc board. Charge up their reservoir
caps to 1KV each. At trigger time, each ramps. Diodes stack them in
series during the ramp. No HV supply, no floating gates, very rugged.

John

 

[Joseph2k]

Tony Williams wrote...

However, the tube approach is still rather complicated, because
the "grounded grid" tube with its programmable current-sinking
MOSFET cathode driver (500V, 1A - not too bad) and its 50W
filament power, will all be sitting at -16kV. Hmm, that better
all be inside a big Faraday cage with rounded corners. Can't
have corona discharge off of our sharp-edged components! Maybe
we can use a clear housing, or add some windows.

snip a few minor problems that would loosen my bowels.....

Since it is a capacitive load could you run everything
up at 0v+positive, and capacitively-couple the negative
voltage ramp to the load?

-+- +12 to 13KV
|
R
|
+-----------+
___|___ |
|
-------+ ? |
0V-+------- |
_______ |
/ |
| Heater |
|--+ |
|MOSFET |
-----||--+ |
| |
R |
| |
-+--0v ===C
| |
[-3KV supply] |
| D | Electrode
+--------+--|<|--+-------> 50pF total
| |
+---R---+

Hey, Tony, great idea, why didn't I think of that? We'll have a 0.1%
to 5% duty cycle at most (1kHz rep rate), so that should be just fine.
I can use the DC restoration you show. It's a huge simplification,
and limits the extreme HV to the plate and a few other items.

Good show!

I find it interesting that my "toob" suggestion is now being seriously
considered. Please be very careful in considering the tube voltage -
current curves, my suggested 4CV35000's would need about 1200 volts cathode
above "grounded grid" to reach cutoff (<1 mA>). Fortunately, the swing
necessary between cutoff and 3 A is no more than 600 V for 4CV35000's.

 

[Winfield Hill]

Joseph2k wrote...

I find it interesting that my "toob" suggestion is now being seriously
considered. Please be very careful in considering the tube voltage -
current curves, my suggested 4CV35000's would need about 1200 volts cathode
above "grounded grid" to reach cutoff (<1 mA>). Fortunately, the swing
necessary between cutoff and 3 A is no more than 600 V for 4CV35000's.

In fairness to Ancient_Hacker (who was that comcast grg2 masked man?),
I have to point out your suggestion was a non-starter. Large modern
tubes are wholly unsuited, with an unattractive amount of capacitance,
and with a devastating level of filament power required, they'd be an
unnecessary cooling nightmare. That's in part because one has to go
all the way up to the giant-size monsters to obtain a 15kV or higher
voltage rating. So those are out, and never were in.

Likewise, the 1mA wimp suggested by Jim and others was a nonstarter,
because we need currents closer to an amp. However, I may employ one
for the low 0.1 to 5mA end of our needs. Sort of a range-switch idea.

No, it was "Ancient_Hacker" who thankfully got the tube ball rolling,
with his 715-C suggestion. I wasn't able to easily find a 715, but it
did cross to the 5D21, a 1944-vintage tube that comes up on eBay all
the time (I bought one for $6.50), and to the Eimac 4PR60C, which is
still in production from time to time (I got a one for $175). It was
these 60W pulse types that made a tube not only viable, but attractive.
Now we can consider 5 to 10A currents, that give us the 0.1 to 0.25us
region as well, very nice!! So I say again, thanks "Ancient_Hacker"!

 

[Joseph2k]

Joseph2k wrote...

In fairness to Ancient_Hacker (who was that comcast grg2 masked man?),
I have to point out your suggestion was a non-starter. Large modern
tubes are wholly unsuited, with an unattractive amount of capacitance,
and with a devastating level of filament power required, they'd be an
unnecessary cooling nightmare. That's in part because one has to go
all the way up to the giant-size monsters to obtain a 15kV or higher
voltage rating. So those are out, and never were in.

Likewise, the 1mA wimp suggested by Jim and others was a nonstarter,
because we need currents closer to an amp. However, I may employ one
for the low 0.1 to 5mA end of our needs. Sort of a range-switch idea.

No, it was "Ancient_Hacker" who thankfully got the tube ball rolling,
with his 715-C suggestion. I wasn't able to easily find a 715, but it
did cross to the 5D21, a 1944-vintage tube that comes up on eBay all
the time (I bought one for $6.50), and to the Eimac 4PR60C, which is
still in production from time to time (I got a one for $175). It was
these 60W pulse types that made a tube not only viable, but attractive.
Now we can consider 5 to 10A currents, that give us the 0.1 to 0.25us
region as well, very nice!! So I say again, thanks "Ancient_Hacker"!

Very clear, and the cost of the new Eimac tubes is non-trivial. At least i
may have helped jog the discussion onto a more productive line. Thank you
for your honest answer.

 

[Winfield Hill]

John Larkin wrote...

Do the Larkin-Marx circuit. Ten stages, all initially at ground, each
a mosfet ramp generator on a small pc board. Charge up their reservoir
caps to 1KV each. At trigger time, each ramps. Diodes stack them in
series during the ramp. No HV supply, no floating gates, very rugged.

Cute idea, but 10 programmable MOSFET ramp generators? Ackk! Yes,
I know they're all identical, but there's got to an easier approach.

 

[Winfield Hill]

Winfield Hill wrote...

Winfield Hill wrote...

Here's one approach to extending the circuit. You can see the
complexity grows rapidly as cascode stages are added, because
a new current sink is required for each gate-biasing capacitor,
and these also have to be cascoded.
. -3kV
. |
. 1mA
. | out
. -4.5kV ______________ +--------------------,
. | | | |
. _|_ | | ___ -3.0 |
. --- 0.5mA | \ |
. Co | | | \ _|_
. -6.0kV ______ |____|_______||-' \___-7.5kV --- Co
. | | | | ||-, |
. _|_ | | | 10V | -4.5kV ramp |
. --- 0.3mA | '--|<|--+ gnd
. Co | | | |
. |____|_______||-' __________||-'
. | | ||-, | ||-,
. __ | | 10V | | 10V |
. __| |__ | '--|<|--+ '--|<|--+
. --+---------- |---+---------- | --, |
. | _ | | _ | | _ |
. '--| \___||-' '--| \___||-' '--| \___||-'
. ,--|_/ ||-, ,--|_/ ||-, ,--|_/ ||-,
. | | | | | |
. '-----------+ '-----------+ '-----------+
. | | |
. 3 R 2 R 2 R
. | | |
. --------------+---------------+---------------+--- -7.5kV

At this point, with only a 4.5kV capability, the MOSFET circuit
may still be more simple than the tube circuit, but extending
the scheme to 6kV would probably tip the balance.

If I could figure out a way to stack the gate-bias capacitors...

Well, duh, go ahead and stack them! Why not try simply making
the capacitive-divider stack part of the capacitive load? This
would allow for a compact circuit that can be placed closer to
the chamber, saving on coax capacitance. Simple!

(And there's no 50W filament heat-removal problem.)

.. 10kV -10kV 1us ramp generator
.. |
.. 1mA CS 0.01
.. | 15kV out
.. ,--------+--+----------+-----------||---------,
.. | | | |
.. | 330pF_|_ | ___ 0 _|_ 35pF
.. 10M 3kV --- | \ ---
.. | | | \ |
.. |________|__________||-' \ GND
.. | | | ||-, \__-10kV
.. | etc _|_ | 10V |
.. 10M --- '--|<|--+
.. | | |
.. |________|__________||-'
.. | | | ||-, 1kV MOSFETs
.. | etc _|_ | 10V | (10 places)
.. 10M --- '--|<|--+
.. | | |
.. GND GND :
.. _ etc
.. __| |__ 5V :
.. -----------, |
.. | _ |
.. '--| \___||-'
.. ,--|_/ ||-, fast pulsed
.. | | constant-
.. '-----------+ current sink
.. |
.. 5.0
.. |
.. ----+--GND

That's not bad, only a 33pF net divider load. A quick calculation
shows under 10V of change in the capacitive-divider bias voltage
resulting from providing the MOSFET's initial Ciss gate charge.

 

[John Larkin]

John Larkin wrote...

Cute idea, but 10 programmable MOSFET ramp generators? Ackk!

Ackk? It doesn't sound that bad to me. A single fiberoptic data stream
could trigger the module and load the slew rate for the next shot into
a serial dac, all in the same burst. The micropower logic could be
powered any number of simple ways. Better yet, couple the power,
trigger, and data through a single, very simple high-voltage
transformer: biphase data/power waveform into a Coolrunner, and a big
spike as the trigger.

John

 

[Ken Smith]

Ackk? It doesn't sound that bad to me. A single fiberoptic data stream
could trigger the module and load the slew rate for the next shot into
a serial dac, all in the same burst.

Better yet: Leave out the fiber. Just point very strong LEDs at good
photocells. This way you can get power to the electronics as well.

If I was doing it, I might invest in some 9V batteries and battery
holders. They should be good for quite a large number of shots.

 

[Ancient_Hacker]

You're welcome.

If you want to build a scaled down version just to get confidence, you
could look at using the 6HV5 tube. It's a common CRT shunt regulator,
not the 6BK4 type, shunting the huge high voltage, but shunting the
boosted B+. So it's a 5KV/300ma tube. And with its 65,000 gm
transconductance, you could even drive it with a 2N404.

And there's bushels of them available on eBay.

 

[Winfield Hill]

Ancient_Hacker wrote...

You're welcome.

If you want to build a scaled down version just to get confidence,
you could look at using the 6HV5 tube. It's a common CRT shunt
regulator, not the 6BK4 type, shunting the huge high voltage, but
shunting the boosted B+. So it's a 5kV/300mA tube. And with its
65,000 gm transconductance, you could even drive it with a 2N404.

2n404, hah!

That's another nice lead, thanks! However, I think I'll hold off
from experimenting with 6HV5 for now, because they'd never handle
the roughly 15kV I'm looking for. They do approach the required
peak current territory, and their 11W heater power isn't too bad.

One issue that bothered me with the 715 family was removing the
50W filament heat, while keeping most of the airflow away from the
12kV wiring. Not too bad, but along with the size and shielding,
a pain. So I'm setting that option aside for now.

Anyway, I think I'll pursue my new MOSFET idea for the moment.
If it works I'll have a nice small structure that can be placed
right next to the vacuum-chamber flanges.

And there's bushels of them available on eBay.

That's always attractive to me. OK, I'll go grab a few of
those beauties, just in case. Where do I find 12-pin sockets?

 

[Tim Williams]

That's always attractive to me. OK, I'll go grab a few of
those beauties, just in case. Where do I find 12-pin sockets?

AES (www.tubesandmore.com) might have them, although I don't remember if
they have just PC mount, or chassis mount as well.

FYI, I have five of those sorts of tubes, a pair 6HV5, 6HS5, and a single
6JH5. All pretty similar, IIRC, at least more similar than your MOSFETs.
;-)

Other tube geeks might have more on hand, rec.audio.radio+phono guys might
have some. Be worth checking vs. e-Bay. Probably not a problem finding
them for cheap though... not really any use for them!

Tim

 

[Winfield Hill]

Tim Williams wrote...

Winfield Hill wrote ...

FYI, I have five of those sorts of tubes, a pair 6HV5, 6HS5,
and a single 6JH5. All pretty similar, IIRC, at least more
similar than your MOSFETs. ;-)

Listen up, you young whipper-snapper, have some respect for
your elders!! All this talk about new-fangled tubes being
better than semiconductors. All heresy, I'll tell ya, and
no respect! What's the world coming to? Those of us who've
been around a while know that it's the important parameters
that count. Like transconductance. This is mathematically
predictable for MOSFETs, within perhaps 20-30% for a given
type, yet is not only all over the map for tubes (whatever
those identical "target" specs of yours may imply), but badly
degrades with time, as well! Dry that spot behind your ears
and listen up, *that* doesn't happen with MOSFETs! Did you
ever hear of us silicon guys ever needing the equivalent of
a tube tester? "Ya better test those MOSFETs, they might be
getting a bit weak." Nope. That's right, you won't find
any FET testers in our supermarkets! No call for 'em.

Another thing, you kids may be proud of running your plates
red hot, but those of us who are more knowledgeable realize
that it's better to put your heat into a slab of silicon and
transfer it from that to a slab of copper, and then to a huge
molded aluminum heat sink. That's because we know thermal
conductance beats thermal radiation, any day. And don't you
forget it kiddo!

Probably not a problem finding them for cheap though...
not really any use for them!

Agreeed.

 

[Ancient_Hacker]

One issue that bothered me with the 715 family was removing the 50W filament heat.

Re the heat, remember that toobs are not temperature sensitive, not
even slightly, not until the glass melts or the seals overheat. maybe
not even then. I've seen 6L6's with the glass melted and sucked in,
shrink-wrapping the plates, and the tubes still worked fine.

50 watts can be carried away by radiation if you leave a few inches of
free space around the tube. You don't have to shoehorn this into a
cell-phone case, do you?

The 715's have a hard-glass bulb and is rated to dissipate the
filament heat and another 60 watts from the plate with no problem.

Just eyeballing it, I suspect you don't have a problem as all the old
TV's had much smaller soft-glass sweep tubes dissipating up to 40 watts
with no forced-air cooling at all. Many mil-spec transmitters ran
813's, a tube only slightly taller than the 715 at over 100 watts with
no airflow at all. The tubes did run very warm, but within the
conservative mil deratings.

Here's a 813 running at 150 watts output, horizontal, convection
cooled:

http://tinyurl.com/dykct

Just a hint. Toobs can take the heat.



The only problem then is the radiated heat warming up any solid-state
stuff you have around the tube!



Regards,

A_H

 

[Winfield Hill]

Winfield Hill wrote...

Winfield Hill wrote...

Well, duh, go ahead and stack them! Why not try simply making
the capacitive-divider stack part of the capacitive load? This
would allow for a compact circuit that can be placed closer to
the chamber, saving on coax capacitance. Simple!

. 10kV -10kV 1us ramp generator
. |
. 1mA CS 0.01
. | 15kV out
. ,--------+--+----------+-----------||---------,
. | | | |
. | 330pF_|_ | ___ 0 _|_ 35pF
. 10M 3kV --- | \ ---
. | | | \ |
. |________|__________||-' \ GND
. | | | ||-, \__-10kV
. | etc _|_ | 10V |
. 10M --- '--|<|--+
. | | |
. |________|__________||-'
. | | | ||-, 1kV MOSFETs
. | etc _|_ | 10V | (10 places)
. 10M --- '--|<|--+
. | | |
. GND GND :
. _ etc
. __| |__ 5V :
. -----------, |
. | _ |
. '--| \___||-'
. ,--|_/ ||-, fast pulsed
. | | constant-
. '-----------+ current sink
. |
. 5.0
. |
. ----+--GND

That's not bad, only a 33pF net divider load. A quick calculation
shows under 10V of change in the capacitive-divider bias voltage
resulting from providing the MOSFET's initial Ciss gate charge.

Ooohhhh, seems to work great in a quick-look spice model. A linear
ramp in about 1us. I used Fairchild's FQD2N100, a 1kV 1.6A MOSFET in
a DPAK smd package, http://www.fairchildsemi.com/pf/FQ/FQD2N100.html
I used the modified spice model I previously developed and vetted
for my 2.5kV precision amplifier project. Each cascode stage has a
damping resistor not shown in the drawing above. Damn, it's looking
good so far!

 

[Winfield Hill]

Winfield Hill wrote...

Winfield Hill wrote...

Ooohhhh, seems to work great in a quick-look spice model. A linear
ramp in about 1us. I used Fairchild's FQD2N100, a 1kV 1.6A MOSFET in
a DPAK smd package, http://www.fairchildsemi.com/pf/FQ/FQD2N100.html
I used the modified spice model I previously developed and vetted
for my 2.5kV precision amplifier project. Each cascode stage has a
damping resistor not shown in the drawing above. Damn, it's looking
good so far!

http://www.picovolt.com/win/elec/ckts/spice/RIS-583/4kV-1A_FQD2N100.pdf

I'd like some comments on this circuit, and so have posted an Acrobat
file on the simulation to my webpage. I started with a simple circuit
and four cascode MOSFET stages. Spice shows a roughly 5ns propagation
delay for each stage, or about 20ns for the four. If this is correct,
it would correspond 75ns for 15 stages, a bit slower than I'd like.

.. ___________________
.. ________/ \______
..
.. --------,,
.. \
.. \
.. \
.. \
.. \
.. \ -10kV/us
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. \
.. '-__________+10V/us

My next step had better be FQD2N100 bench-testing used this way.

 

[Ancient_Hacker]

Wear safety glasses.

But seriously, how about some diodes to clip the gate excursions? I
know, Spice says the voltages will track perfectly. IMHO asking for
2% tracking shows an awful lot of faith in the matching of dynamic
transconductances.

BTW Tektronix tried something similar for the HV regulator board on one
of their early 43x scopes. A nice sedate application, with lots of
zener diodes maintaining order.

A year later they sent out a new, completely redesigned PCB board, with
a completely different, no series transistors, design. Hmmmm........

 

[Winfield Hill]

Winfield Hill wrote...

http://www.picovolt.com/win/elec/ckts/spice/RIS-583/4kV-1A_FQD2N100.pdf

I'd like some comments on this circuit, and so have posted an Acrobat
file on the simulation to my webpage. I started with a simple circuit
and four cascode MOSFET stages. Spice shows a roughly 5ns propagation
delay for each stage, or about 20ns for the four. If this is correct,
it would correspond 75ns for 15 stages, a bit slower than I'd like.

. ___________________
. ________/ \______
.
. --------,,
. \
. \
. \
. \
. \
. \ -10kV/us
. \
. \
. \
. \
. \
. \
. \
. \
. \
. \
. \
. \
. '-__________+10V/us

My next step had better be bench-testing FQD2N100 used this way.

From the .pdf, "The Iout current (bold blue trace) starts in about
20ns, but shows a 1.8% drop during the pulse, from 270mA to 265mA..."

Make that a 4% drop, and, from 275 to 265mA. Still not too bad.

Differentiating the output-voltage ramp, we get a -8.3GV/s slope
from 36ns until 300ns, afterwhich the slope drops gradually to
-8.0GV/s at 400ns, then faster to -7.5GV/s at 435ns, where the
output voltage is only 250V, and pulse is nearly done.

And yes, that's right, an 8 GV/s voltage-slewing rate! Well, OK,
that's 8kV/us in more common terms. But it still impresses me.
I wonder if it's real.