Series mosfets for high voltage?

[Mike Poulton]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser. The power supply for a copper
halide laser is a bit tough -- a stream of 50ns 15kV pulses at 20kHz.
Hydrogen thyratrons are great for this sort of thing, except that they
can't easily be made to repeat their performance 20,000 times per second.
I'm leaning towards a string of series mosfets or IGBTs. How can I make a
string of 15 switches (1200V each) commutate rapidly, without destroying
half the stack due to avalanche energy? How can I deal with 15 separate
isolated gate drives that all must switch simultaneously? Is this even
practical? Any advice would be appreciated.

--
Mike Poulton
MTP Technologies

Live free or die! http://www.indefenseoffreedom.org/

Unless the government has a really excellent reason, anyone should be
allowed to possess, own, purchase, store, use, publish, say, or do
anything that does not cause demonstrable harm to another person without
that person's consent. "To fight terrorism" in the vague sense is not
even close to sufficient reason.

 

[Heindorf]

Hello Mike
I've only scanned your message and know of this site.www.behlke.de
Perhaps it's what you're looking for.
Regards
Rolf

 

[Fritz Schlunder]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser. The power supply for a copper
halide laser is a bit tough -- a stream of 50ns 15kV pulses at 20kHz.
Hydrogen thyratrons are great for this sort of thing, except that they
can't easily be made to repeat their performance 20,000 times per second.
I'm leaning towards a string of series mosfets or IGBTs. How can I make a
string of 15 switches (1200V each) commutate rapidly, without destroying
half the stack due to avalanche energy? How can I deal with 15 separate
isolated gate drives that all must switch simultaneously? Is this even
practical? Any advice would be appreciated.

This document here may be of interest:

http://www.eece.ksu.edu/~gjohnson/tcchap7.pdf

It regards a solid state tesla coil using series IGBT devices.

 

[Rene Tschaggelar]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser. The power supply for a copper
halide laser is a bit tough -- a stream of 50ns 15kV pulses at 20kHz.
Hydrogen thyratrons are great for this sort of thing, except that they
can't easily be made to repeat their performance 20,000 times per second.
I'm leaning towards a string of series mosfets or IGBTs. How can I make a
string of 15 switches (1200V each) commutate rapidly, without destroying
half the stack due to avalanche energy? How can I deal with 15 separate
isolated gate drives that all must switch simultaneously? Is this even
practical? Any advice would be appreciated.

There are indeed circuits that use stacks of MOSFETs.
A reprate of 20kHz is way beyond though.
You'll have to be content with 1kHz or even lower.
Best get a switch from behlke as mentioned by Heindorf.

Rene

 

[Frithiof Andreas Jensen]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser. The power supply for a copper
halide laser is a bit tough -- a stream of 50ns 15kV pulses at 20kHz.
Hydrogen thyratrons are great for this sort of thing, except that they
can't easily be made to repeat their performance 20,000 times per second.
I'm leaning towards a string of series mosfets or IGBTs. How can I make a
string of 15 switches (1200V each) commutate rapidly, without destroying
half the stack due to avalanche energy? How can I deal with 15 separate
isolated gate drives that all must switch simultaneously? Is this even
practical? Any advice would be appreciated.

It can be done - the 20 kHz +15 kV being the hard part because of capacitive
losses - but it is *really hard*; I burned about 200++ MOSFET's doing this
for a Magnetron power supply.

This is serious tech. and the web will not tell you much, but there are
several tricks+techniques described in the patent litterature which is a
good reference to search and also the IEEE has some papers which are useful.

What I did was to build a stack of switch modules, each module mounted on a
small pcb with "tabs" so that the stack could be soldered into a PCB. The
drive was via a ring core mounted on each swich PCB and threaded with a
single turn of EHT coax cable with the shield connected to half the EHT
supply to cut down on capacitive coupling.

You will need to place transorbs on each FET/IGBT; two 400 V devices will be
much cheaper than a single 800 V d.o. You also need to provide circuitry to
discharge the gate on the end of the pulse as well as a way to control the
rise rate of the current to something the device will handle. The little
bonding wire inside the HV FET/IGBT *will easily* have enough inductance to
overvolt the gate and kill the switch - this was what killed the majority
of my FET's before I figured that out.

The drive of the stack is somewhat critical too. Basically you will want to
switch on rapidly but not extremely fast.

Finally, you will need to measure the current sampled on a fixed position
within each pulse - this is used to set the voltage so that the current
remains constant despite that your load changes over time and the resistance
of the switches vary with
temperature.

Based on the above hints - which are also in the litterature - I think you
can work it out.

I cannot be more specific though: if I venture too much detail I would
probably violate a previous employers IPR.

 

[Paul Burridge]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser.

What's this type of laser capable of in the powe output you envisage?
It's not a type I've encountered before...

 

[Jeroen Vriesman]

http://www.divtecs.com/pmtech_03.htm

you can put them in series and they can make pulses of 50ns. they do have a
leakage of 1mA, I don't know if that is a problem.

cool thingy your building, is it for home use?

Cheers,
Jeroen

joenix at gmx dot net

 

[Mike Harrison]

http://www.divtecs.com/pmtech_03.htm

you can put them in series and they can make pulses of 50ns. they do have a
leakage of 1mA, I don't know if that is a problem.

cool thingy your building, is it for home use?

Cheers,
Jeroen

joenix at gmx dot net

I'm sure it's practical, but will be fraught with unexpected problems... you will certainly learn a
lot in the process.
I'd advise using the cheapest parts you can get as you will probably kill many....
I think 1KV Mosfets and IGBTs are available at reasonable cost.

For gate drive, fibre optic would probably be ideal, although getting power for gate drive may be
interesting - maybe have the light normally on, and blink to fire, or use a second fibre into a
photovoltaic cell for power.

It might also be worth exploring a transformer/voltage multiplier approach, maybe a marx generator
plus pulse-forming network would work - maybe using HV thyristors or gas-discharge tubes instead of
sparkgaps ?

 

[Robert Baer]

I imagine this topic has probably been covered before, but I was unable to
find any good info using google/google groups. I am considering (yet
again) building a copper halide laser. The power supply for a copper
halide laser is a bit tough -- a stream of 50ns 15kV pulses at 20kHz.
Hydrogen thyratrons are great for this sort of thing, except that they
can't easily be made to repeat their performance 20,000 times per second.
I'm leaning towards a string of series mosfets or IGBTs. How can I make a
string of 15 switches (1200V each) commutate rapidly, without destroying
half the stack due to avalanche energy? How can I deal with 15 separate
isolated gate drives that all must switch simultaneously? Is this even
practical? Any advice would be appreciated.

--
Mike Poulton
MTP Technologies

Live free or die! http://www.indefenseoffreedom.org/

Unless the government has a really excellent reason, anyone should be
allowed to possess, own, purchase, store, use, publish, say, or do
anything that does not cause demonstrable harm to another person without
that person's consent. "To fight terrorism" in the vague sense is not
even close to sufficient reason.

I suggest you use *one* 1000V FET and a transformer to step up the
voltage.
It is not easy to find 1200V FETs, but 1000V FETs are fairly common
(in TO-220).
You did not say anything about the current or power needed for the
laser.
When one gets into the high power pulse region, you need to know the
I*I*T rating
of the switcher, and no datasheet these daze even hints of that.
About 20 or so years ago, "eye squared tee" ratings were published or
were easily available if not in datasheet.
Ignitrons always had that rating, some thyratrons, and a number of
SCRs designed for PFN discharge work also were speced.
High power FETs did not exist then (at least commercially).

 

[JeffM]

transformer/voltage multiplier approach, maybe a marx generator

We did this (100ns, 6kHz, +/-15kV, 300A) for plasma generation.
We found it much easier to use an inductive adder
(Lots of paralleled primaries, series secondaries).
The available ferrites defined our limit.

Advanced Power Technology 1000V 40A FETs ($38 each). 30 iterations.

|
|
||_|_||____ || __________
| || ) (
---|| ) || (
| ) (
||_________) || (__
| |
| || |
| | |
||_|_||____ || __|
| || ) (
---|| | ) || (
| | ) (
||_________) || (__
|
| || |
| | |
||_|_||____ || __|
| || ) (
---|| | ) || (
| | ) (
||_________) || (_________
|

 

[Mike Poulton]

We did this (100ns, 6kHz, +/-15kV, 300A) for plasma generation.
We found it much easier to use an inductive adder
(Lots of paralleled primaries, series secondaries).
The available ferrites defined our limit.

I suppose that could work. It's not the typical way of building a copper
halide laser PSU, but it seems reasonable. The pulses would need to be
50ns max width, 20kHz min rep rate, with an average power of about 2kW.
That works out to 150A or so. The ferrite would definitely be the tough
part.

--
Mike Poulton
MTP Technologies

Live free or die! http://www.indefenseoffreedom.org/

Unless the government has a really excellent reason, anyone should be
allowed to possess, own, purchase, store, use, publish, say, or do
anything that does not cause demonstrable harm to another person without
that person's consent. "To fight terrorism" in the vague sense is not
even close to sufficient reason.

 

[Frithiof Andreas Jensen]

I suppose that could work. It's not the typical way of building a copper
halide laser PSU, but it seems reasonable. The pulses would need to be
50ns max width, 20kHz min rep rate, with an average power of about 2kW.
That works out to 150A or so. The ferrite would definitely be the tough
part.

The capacitance of the upper stages will kill the efficiency - you may be
better off by using a transformer with layered secondaries wound in the same
direction with the voltage increasing outwards away from the core. This
transformer is the power supply; the secondaries are wound in the same
direction to remove inter-winding coupling in the power supply.

Each swich stage is powered off a secondary so that the first stage, using
the innermost secondary, is at ground potential and the outermost stage is
at the 15kV level. This is a good solution for high powers - the tranformer
*needs* to be large anyway because of insulation distances and the need to
keep the capacity loading down. It is similar to a Marx Generator in
topology - hence I do not think there are any patents to violate ;-).

Look up a patent by Thomcast CSF for a "Pulse-Step Modulator" for an
incarnation - essentially a series connection of switched 1 kV power
supplies, the patent covering mainly the implementation of the open-loop
voltage control algorithm.

 

[Rich Grise]

How about a surplus radar set?

Good Luck!
Rich

 

[Robert Baer]

How about a surplus radar set?

Good Luck!
Rich

If you still need to run a number of isolated FETs, use transformers.
If you need fast rise and fall times at the gates, bifilar is one way
to get low mutual inductance; another way is to use coax cable with the
shield as the primary and the inner conductor as the secondary.
And with a bit of ingenuity, turn ratios of 1:5 to 5:1 are possible.