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

[John Larkin]

Now I'm challenged with a creating a moderately-precise (5%),
programmable, fairly-fast 10kV ramp, of 1 to 50us in duration.
It's floating on 3kV. The 1us-long, 3kV to 13kV ramp spec is
tough, but may have a relaxed 5%-precision spec, provided it's
repeatable.

You can't fool us. This is just a homework assignment.

John

 

[John Larkin]

Hello John,



Is that available online? I could only see literature on scholar.google
that quoted RadLab.

Anyway, I was thinking about those ballast triodes in earlier color TVs.
If there is any chance to find one, that is, without having to rent a
backhoe and dig through a landfill.

6BK4, 25KV but can sink only 1.5 mA, not enough to drive Win's load.

John

 

[Winfield Hill]

John Larkin wrote...

MIT RadLab, volume 5, "Pulse Generators."

We have that at work in the library. I bought the entire
set years ago from a flea-market guy for $700, IIRC.

 

[Winfield Hill]

John Larkin wrote...

6BK4, 25kV but can sink only 1.5 mA, not enough to drive
Win's load.

Guess it's going to be a stack of FETs...

 

[Frank Bemelman]

Hello John,



Is that available online? I could only see literature on scholar.google
that quoted RadLab.

Anyway, I was thinking about those ballast triodes in earlier color TVs.
If there is any chance to find one, that is, without having to rent a
backhoe and dig through a landfill.

You're not going to tell me that you replaced it?

 

[Joerg]

Hello Win,

Are there any practical 12 to 15kV triodes, or pentodes?

Yes, even 30kV. 6BK4 or Telefunken ED500. Possibly Svetlana still makes
the GP-5:
http://216.87.144.102/svetlana/pdf/GP5.pdf

They may be hard to obtain unless you settle for a few used ones. Such
tubes were employed as regulated ballasts in early color TVs to keep the
HV from varying. IOW huge shunt regulators. That was back in the days
when electricity was considered to be almost free and nobody was too
concerned about x-rays. Later they went to a separate HV generator
instead of using the deflection flyback.

Regards, Joerg

 

[John Larkin]

John Larkin wrote...

Guess it's going to be a stack of FETs...

Sounds like it.

Instead of a stack of fets and a 10KV supply, might it be possible to
sort of stack stages, and run each one from a charged cap (sort of a
controlled Marx?) or maybe individual dc/dc converters, maybe a dozen
stages at 800 volts each?

John

 

[Ian Du Rieu]

Guess it's going to be a stack of FETs...

Hi Win,

What about a triggered spark gap?
It'll handle the volts & amps.
Good fault tolerance.
Dielectric is self healing.
Good optical & acoustic entertainment side effects


Cheers,
Ian Du Rieu
The Leon Audio Company

Add .au to email address to reply.

 

[John Larkin]

I was thinking about a classic glass jug like an 833A or something.
Anybody know how much plate current one of these can manage? My
transmitting tube books are at home.

Hey, cool:

Max plate voltage 4000, but if you look at the amount of vacuum in
there, I'd go for lots more.

4 amps Id (er... Ip) with about +350 on the grid *and* the plate.

Cout probably in the 15 pF range.

And needs a mere 10 volts, 10 amps for the filament.

I have a couple, and they're beautiful hunks of glass.

John

 

[John Larkin]

Hi Win,

What about a triggered spark gap?
It'll handle the volts & amps.
Good fault tolerance.
Dielectric is self healing.
Good optical & acoustic entertainment side effects


Cheers,
Ian Du Rieu
The Leon Audio Company

Add .au to email address to reply.

Or a nice Russian hydrogen thyratron, except that Win has to control
the slew rate somehow.

John

 

[Fred Bloggs]

Over the years I've completed a number of projects involving
roughly 10kV voltages: fixed sources, fast-shutoff, reversible
programmable DC sources, precision-ramped AC sources at 300kHz
to 1MHz with 10kV maximum amplitude, etc.

Now I'm challenged with a creating a moderately-precise (5%),
programmable, fairly-fast 10kV ramp, of 1 to 50us in duration.
It's floating on 3kV. The 1us-long, 3kV to 13kV ramp spec is
tough, but may have a relaxed 5%-precision spec, provided it's
repeatable.

What has to be 5%, maximum deviation of best fit ramp to actual, or
total ramp time? And then when you attain 10KV what happens?- discharge
any old way?- hold it? Sounds like you're going to be doing some serious
study of transient thermal response curves, something is going to be
dissipating 10KW peak.
View in a fixed-width font such as Courier.

. 13kV
. |
. .
. .

 

[John Larkin]

What has to be 5%, maximum deviation of best fit ramp to actual, or
total ramp time? And then when you attain 10KV what happens?- discharge
any old way?- hold it? Sounds like you're going to be doing some serious
study of transient thermal response curves, something is going to be
dissipating 10KW peak.

If you look at the soar curves of a typical TO-247 fet, most of them
can dissipate kilowatts in the microsecond time range. But I'd still
expect a number of interesting blow-all-the-fets modes.

An IRFPS37N50A can dissipate about 40 KW for 10 usec.

John

 

[Fred Bloggs]

If you look at the soar curves of a typical TO-247 fet, most of them
can dissipate kilowatts in the microsecond time range. But I'd still
expect a number of interesting blow-all-the-fets modes.

An IRFPS37N50A can dissipate about 40 KW for 10 usec.

John

At these relatively modest currents but high voltage he should look to
SuperTex. They have more than a few patents on virtually blow-out proof
inventions in this area.

 

[Fred Bloggs]

If you look at the soar curves of a typical TO-247 fet, most of them
can dissipate kilowatts in the microsecond time range. But I'd still
expect a number of interesting blow-all-the-fets modes.

An IRFPS37N50A can dissipate about 40 KW for 10 usec.

John

Hmmm...at 10V/ns and 100p Cdg , he will need 1A constantly pumping into
those gate circuits.

 

[Fred Bloggs]

Hmmm...at 10V/ns and 100p Cdg , he will need 1A constantly pumping into
those gate circuits.

That would be divided by the number of FETs...phew.

 

[Ben Bradley]

You can't fool us. This is just a homework assignment.

I want to know what class he's taking that gets this for a homework
assignment!

 

[John Larkin]

That would be divided by the number of FETs...phew.

Well, it will be an interesting control problem. With his numbers, the
usual resistive or even r-c gate drive stack sounds nasty. A separate
servoed gate drive for each stage sounds good to me, but that will
need some isolated mechanism for selecting the slew rates and a
(perhaps separate) way to fire them all. Fiberoptics? Batteries?

I'm sure looking forward to hearing how he solves this one.

John

 

[Fred Bloggs]

Well, it will be an interesting control problem. With his numbers, the
usual resistive or even r-c gate drive stack sounds nasty. A separate
servoed gate drive for each stage sounds good to me, but that will
need some isolated mechanism for selecting the slew rates and a
(perhaps separate) way to fire them all. Fiberoptics? Batteries?

I'm sure looking forward to hearing how he solves this one.

John

Well- if he keeps that load capacitance down and uses a boot capacitor
across the whole current source of about 100x this- back biasing an HV
diode to the 10KV supply , then the slew rate problem is eliminated or
at least vastly reduced.

 

[Fred Bloggs]

Well- if he keeps that load capacitance down and uses a boot capacitor
across the whole current source of about 100x this- back biasing an HV
diode to the 10KV supply , then the slew rate problem is eliminated or
at least vastly reduced.

That's not right- he needs to boot the whole 10kV supply...

 

[John Larkin]

That's not right- he needs to boot the whole 10kV supply...

I still sort of like the idea of a linear Marx generator. Have, say, a
dozen mosfet ramp generators, all sitting on ground. Charge their
supply caps to 800 volts apiece, then trigger them to start ramping
together, with steering diodes to "erect" the string.

Transformers might be interesting, except for the 50:1 time spread.

John