From Newsgroup: sci.electronics.design

We're designing a little GaN-fet high-voltage pulse generator and it
needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.





John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Your time scale is in mS I think?
Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!



--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1
https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

Your time scale is in mS I think?

Yes, the sim runs for 300 ms. That's about 5 minutes in real time,
which leaves lots of time to pretty up the schematic. And make snacks.

Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!

We have all those parts in stock.

A single diode would have a kilovolt or so across it, which would be a
lot of stress on the PCB.


John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 3/01/2026 3:07 am, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Transformers should be wound for the job you want them to do. It's not
the kind of intellectual load you like to cope with, but it can be done.

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

Who cares.Jan can make things that work. He doesn't seem to put them
into production.

Your time scale is in mS I think?

Yes, the sim runs for 300 ms. That's about 5 minutes in real time,
which leaves lots of time to pretty up the schematic. And make snacks.

Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!

We have all those parts in stock.

But you don't have a transformer winding machine, reels of wire, stocks
of coil formers and the ferrite cores you plug into them.

A single diode would have a kilovolt or so across it, which would be a
lot of stress on the PCB.

There are Teflon insulated standoffs, that you can plug into a circular
hole in the PCB, which can stand-off a few kV. I used them from time to
time at Cambridge Instruments, mostly for photomultiplier dynode chains.
--
Bill Sloman, Sydney
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

john larkin <jl@glen--canyon.com>wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>

wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1
https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Oh, on the transfromer wire perhaps, seems to not mind

What diode is that?

No idea, SK11 or SK10? some thing that does HV


What's the big square thing?

OLED display LM096

Where is the oscillator?

Microchip PIC 18F14K22
It has build in PWM with cycle by cycle current limit (analog comparot in)

Does it regulate?

So, yes.
And you can set voltage and PWM and a lot of other stuff via USB
It has GPS too

Your time scale is in mS I think?

Yes, the sim runs for 300 ms. That's about 5 minutes in real time,
which leaves lots of time to pretty up the schematic. And make snacks.

Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!

We have all those parts in stock.

??

A single diode would have a kilovolt or so across it, which would be a
lot of stress on the PCB.

Na. never noticed :-)

This one is on 24/7 on my desk:
uses a voltage doubler, some old potcore I found.
Now runs on 12V via a mains adaptor, records radiation level 24/7 to a file, since.. 2010?
The detector is from an old army surplus set.

When I moved house:
https://panteltje.nl/pub/background_radiation_from_one_place_to_the_other.gif
radiation level went up!

Microchip PICs are nice, use hardly any power, have good build in hardware functions (ADC, PWM, analog comparator, DAC, what not), reliable and simple to program in asm.

There is a lot more to it, just ask.
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On Sat, 3 Jan 2026 03:42:13 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 3/01/2026 3:07 am, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

Obviously. That DRQ127 is surface-mount and costs 86 cents and
required no engineering design, no drawings, no quotations, no
prototypes, and has multiple sources. Leakage inductance is tiny. The
C-W multiplier lets the tranny work at low voltage stress.

The diodes are dual SOT23s which cost 5 cents per dual. The HV supply
is a minor side issue on the HV pulse generator project. [1]

It's easy for Bill Sloman to theorize when he doesn't actually design
anything to be built.


[1] last summer I had a great intern do a bunch of work with GaN fets
and gate drivers and txline transformers. Then he automated a bunch of
the floats in this years Rose Bowl Parade.

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Transformers should be wound for the job you want them to do. It's not
the kind of intellectual load you like to cope with, but it can be done.

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

Who cares.Jan can make things that work. He doesn't seem to put them
into production.

Your time scale is in mS I think?

Yes, the sim runs for 300 ms. That's about 5 minutes in real time,
which leaves lots of time to pretty up the schematic. And make snacks.

Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!

We have all those parts in stock.

But you don't have a transformer winding machine, reels of wire, stocks
of coil formers and the ferrite cores you plug into them.

A single diode would have a kilovolt or so across it, which would be a
lot of stress on the PCB.

There are Teflon insulated standoffs, that you can plug into a circular
hole in the PCB, which can stand-off a few kV. I used them from time to
time at Cambridge Instruments, mostly for photomultiplier dynode chains.

John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

Bill Sloman <bill.sloman@ieee.org>wrote:

On 3/01/2026 3:07 am, john larkin wrote:
On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Transformers should be wound for the job you want them to do. It's not
the kind of intellectual load you like to cope with, but it can be done.

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

Who cares.Jan can make things that work. He doesn't seem to put them
into production.

Your time scale is in mS I think?

Yes, the sim runs for 300 ms. That's about 5 minutes in real time,
which leaves lots of time to pretty up the schematic. And make snacks.

Top right HV generator:
https://panteltje.nl/panteltje/pic/gm_pic2/gm_pic2-0.8_circuit_diagram_IMG_4475.JPG

From:
https://panteltje.nl/panteltje/pic/gm_pic2/

Thing has been working fine 24/7 for > 10 years now!

We have all those parts in stock.

But you don't have a transformer winding machine, reels of wire, stocks
of coil formers and the ferrite cores you plug into them.

A single diode would have a kilovolt or so across it, which would be a
lot of stress on the PCB.

There are Teflon insulated standoffs, that you can plug into a circular
hole in the PCB, which can stand-off a few kV. I used them from time to
time at Cambridge Instruments, mostly for photomultiplier dynode chains.

https://panteltje.nl/pub/PMT_HV_generator_solder_side_img_3172.jpg https://panteltje.nl/pub/PMT_HV_supply_with_regulator_img_3175.jpg https://panteltje.nl/pub/gamma_spectrometer_plus_probe_plus_geiger_counter_2_IMG_4185.JPG
The PMT is in the carboard tube
Spectrometer runs on 2 AA batteries
Have no fear for HV, 25 kV color sets were more dangerous.



--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 3/01/2026 4:36 am, john larkin wrote:

On Sat, 3 Jan 2026 03:42:13 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 3/01/2026 3:07 am, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

Obviously. That DRQ127 is surface-mount and costs 86 cents and
required no engineering design, no drawings, no quotations, no
prototypes, and has multiple sources. Leakage inductance is tiny. The
C-W multiplier lets the tranny work at low voltage stress.

The diodes are dual SOT23s which cost 5 cents per dual. The HV supply
is a minor side issue on the HV pulse generator project. [1]

It's easy for Bill Sloman to theorize when he doesn't actually design anything to be built.

I'm not theorising - I'm just telling you what worked for me in the
past. The stuff got built and put into production, and if final test
didn't like it they would try to get me to change the design to one that
they liked better - mostly they'd got the wrong end of the stick. The
one design change I did make was when they couldn't get my set-up
procedure to work - I actually set up the offending unit in ten minutes,
but they'd struggled for hours, and switching to a laser-trimmed part
only cost (in cash terms) about twenty minutes of final test labour.

<snip>
--
Bill Sloman, Sydney
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

"Bill Sloman" <bill.sloman@ieee.org> wrote in message news:10j90or$j7t0$1@dont-email.me...

On 3/01/2026 4:36 am, john larkin wrote:

On Sat, 3 Jan 2026 03:42:13 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 3/01/2026 3:07 am, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

Obviously. That DRQ127 is surface-mount and costs 86 cents and
required no engineering design, no drawings, no quotations, no
prototypes, and has multiple sources. Leakage inductance is tiny. The
C-W multiplier lets the tranny work at low voltage stress.

The diodes are dual SOT23s which cost 5 cents per dual. The HV supply
is a minor side issue on the HV pulse generator project. [1]

It's easy for Bill Sloman to theorize when he doesn't actually design
anything to be built.

I'm not theorising - I'm just telling you what worked for me in the past.

No you're not. You're insulting him. Insulting others seems to be what you
do best. That and bragging about how you set the unit up in ten minutes after others had struggled for hours.

To avoid this becoming another silly mega thread I won't reply to any reply
to this.

The stuff got built and put into production, and if final test didn't like it they would try to get me to change the design to one
that they liked better - mostly they'd got the wrong end of the stick. The one design change I did make was when they couldn't get
my set-up procedure to work - I actually set up the offending unit in ten minutes, but they'd struggled for hours, and switching
to a laser-trimmed part only cost (in cash terms) about twenty minutes of final test labour.

<snip>

--
Bill Sloman, Sydney

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses,
something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com ====================================================== http://www.flickr.com/photos/didi_tgi/



--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it

needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses, >something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com >====================================================== >http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have
in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.




John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 3/01/2026 6:18 am, Edward Rawde wrote:

"Bill Sloman" <bill.sloman@ieee.org> wrote in message news:10j90or$j7t0$1@dont-email.me...

On 3/01/2026 4:36 am, john larkin wrote:

On Sat, 3 Jan 2026 03:42:13 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 3/01/2026 3:07 am, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid> >>>>> wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier >>>>>>>
https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to >>>>>>> mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

John Larkin admits the possiblity of getting special purpose
transformers wound, but he prefers not to have it done

Obviously. That DRQ127 is surface-mount and costs 86 cents and
required no engineering design, no drawings, no quotations, no
prototypes, and has multiple sources. Leakage inductance is tiny. The
C-W multiplier lets the tranny work at low voltage stress.

The diodes are dual SOT23s which cost 5 cents per dual. The HV supply
is a minor side issue on the HV pulse generator project. [1]

It's easy for Bill Sloman to theorize when he doesn't actually design
anything to be built.

I'm not theorising - I'm just telling you what worked for me in the past.

No you're not. You're insulting him. Insulting others seems to be what you
do best. That and bragging about how you set the unit up in ten minutes after others had struggled for hours.

You missed the point. I'd developed the circuit, and I knew that the
set-up procedure I'd described had worked for me and did take about ten minutes. The problem was that I hadn't described it well enough to let
the final test guys do it at all. There's no brag in admitting that you
hadn't been able to describe a fairly simple - if tedious and repetitive
- procedure well enough to let final test techicians (who were good at
that sort of stuff) make it work. My expectation had been that they
could do it faster than I could, and I was wrong about that.

The underlying point was John Larkin doesn't seem to realise that I've
done at least as much hand-on work as he has.

To avoid this becoming another silly mega thread I won't reply to any reply to this.

Why would you bother? You don't like being told that you have missed the point, and you always do.
--
Bill Sloman, Sydney


--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 1/3/2026 3:00, john larkin wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different
voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses,
something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We have used some of theirs and they work OK, but I still prefer to
have more control over the loop and do 1/2 74HC123, 1/2 LMC6482 plus
driver and FET (like on this board, where noise is critical and there
are what, 8 voltages in addition to the HV.

We try to surface-mount everything, and use parts that we already have
in stock. And avoid ebay or amazon or ali parts.

Same here. Throuhole come into play when they can save board area at
the expense of height, when available.
Ali got me by a nice surprise when I found there some coaxial connectors
which had disappeared at Mouser etc., at a much more reasonable cost.
So I tried again, this time for some mini-melf resistors we had run
out of; guess what, they track better than the previous we had, from
some reputable source... (tracking is crucial at a few points on this
board, we specify <30ppm/C gain drift, ship <20ppm, and *one* unit
came out around *0*, could not be measured, can you believe it!).
I also got some chips there (two types), yet to try these out though
(both are not to be found at Mouser/Digikey etc.).

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

Yes, these are ubiquitous and not usable. I was surprised to find the
fast ones some 15 years ago, we had run out of a supply of an older
type I had, Philips or something like that, were up to 3mA and lossy
but worked.
The new ones are much better, I think 10mA - no, 25! Located some
datasheet I must have found then, G10FS is the part name (or one of the neighbour ones), 100ns is pretty good.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.

No idea, I have only dismantled a dead microwave once to take the
magnets... :-)

--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

john larkin <jl@glen--canyon.com> wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM
out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >>voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses, >>something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com >>====================================================== >>http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have
in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.

Where you can find semiconductors that can work at 2.4 GHz at power
level of few hundred watts?
--
Waldek Hebisch
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On Sat, 3 Jan 2026 23:15:13 -0000 (UTC), antispam@fricas.org (Waldek
Hebisch) wrote:

john larkin <jl@glen--canyon.com> wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid>
wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier

https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to
mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG

Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >>>voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses, >>>something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com >>>====================================================== >>>http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have
in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.

Where you can find semiconductors that can work at 2.4 GHz at power
level of few hundred watts?

Apparently GaN.

We'll be driving GaN fets with the 700v power supply that I mentioned.

The GaNs are wonderful to drive. They only need 5 volts on the gate
and the drain-gate capacitance is essentially zero.


John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design






On 4/01/2026 12:09 pm, john larkin wrote:

On Sat, 3 Jan 2026 23:15:13 -0000 (UTC), antispam@fricas.org (Waldek
Hebisch) wrote:

john larkin <jl@glen--canyon.com> wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid> >>>>> wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier >>>>>>>
https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher
voltages in the future. They are crazy easy to drive compared to >>>>>>> mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG >>>>>>
Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work
for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >>>> voltages for different detectors, for Ge I have seen anything between
2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses,
something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have
in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.

Where you can find semiconductors that can work at 2.4 GHz at power
level of few hundred watts?

Apparently GaN.

We'll be driving GaN fets with the 700v power supply that I mentioned.

The GaNs are wonderful to drive. They only need 5 volts on the gate
and the drain-gate capacitance is essentially zero.

It won't be "essentially zero", but something small but finite that a
serious designer would know about, even if they didn't have to do
anything about it.

Joe Gwinn has identified a similar part - the 50V Qorvo T1G4020036_FL

https://au.mouser.com/datasheet/3/1081/1/T1G4020036_FL_Data_Sheet.pdf

and the data sheet doesn't a the drain to gate capacitance - it just
presents a series of Smith charts which encode much the same data.

Measuring it wouldn't be difficult. You could probably get a Spice model
out of the manufacturers if you asked, but it would need to include lead inductances and would be messy.
--
Bill Sloman, Sydney
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On Sun, 4 Jan 2026 14:25:30 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 4/01/2026 12:09 pm, john larkin wrote:

On Sat, 3 Jan 2026 23:15:13 -0000 (UTC), antispam@fricas.org (Waldek
Hebisch) wrote:

john larkin <jl@glen--canyon.com> wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid> >>>>>> wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier >>>>>>>>
https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher >>>>>>>> voltages in the future. They are crazy easy to drive compared to >>>>>>>> mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG >>>>>>>
Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work >>>>> for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >>>>> voltages for different detectors, for Ge I have seen anything between >>>>> 2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then
(ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses, >>>>> something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have >>>> in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would
have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap.

Where you can find semiconductors that can work at 2.4 GHz at power
level of few hundred watts?

Apparently GaN.

We'll be driving GaN fets with the 700v power supply that I mentioned.

The GaNs are wonderful to drive. They only need 5 volts on the gate
and the drain-gate capacitance is essentially zero.

It won't be "essentially zero", but something small but finite that a >serious designer would know about, even if they didn't have to do
anything about it.

It must be a drag to always dispute everything so literally. That
instantly kills any traces of creativity.

"Essentially zero" means "too small to matter" to us. The PCB has more
d-g capacitance that the fets we are using.

Joe Gwinn has identified a similar part - the 50V Qorvo T1G4020036_FL

https://au.mouser.com/datasheet/3/1081/1/T1G4020036_FL_Data_Sheet.pdf

and the data sheet doesn't a the drain to gate capacitance - it just >presents a series of Smith charts which encode much the same data.

My my, you made yet another spelling error!

That's an RF part and has RF specs, narrowband s-params and Smith
stuff, presumably to be used with slide rules. We use switch-type
parts that have proper specs, with c-v curves.

Many or maybe most RF parts have wirebonds that tune it for one
specific RF band, wherever they think the money is.

The bias procedure is typical for RF parts. Slowly adjust the gate
voltage until it works. I see no DC curves. It's probably a depletion
part.

Some day people will use Spice to simulate RF systems, and provide
proper device models.


John Larkin
Highland Tech Glen Canyon Design Center
Lunatic Fringe Electronics
--- Synchronet 3.21a-Linux NewsLink 1.2

From Newsgroup: sci.electronics.design

On 5/01/2026 3:11 am, john larkin wrote:

On Sun, 4 Jan 2026 14:25:30 +1100, Bill Sloman <bill.sloman@ieee.org>
wrote:

On 4/01/2026 12:09 pm, john larkin wrote:

On Sat, 3 Jan 2026 23:15:13 -0000 (UTC), antispam@fricas.org (Waldek
Hebisch) wrote:

john larkin <jl@glen--canyon.com> wrote:

On Fri, 2 Jan 2026 21:52:44 +0200, Dimiter_Popoff <dp@tgi-sci.com>
wrote:

On 1/2/2026 18:07, john larkin wrote:

On Fri, 02 Jan 2026 06:07:18 GMT, Jan Panteltje <alien@comet.invalid> >>>>>>> wrote:

john larkin <jl@glen--canyon.com>wrote:

We're designing a little GaN-fet high-voltage pulse generator and it >>>>>>>>> needs a high-voltage power supply.

This is a variation on the autotransformer flyback C-W multiplier >>>>>>>>>
https://www.dropbox.com/scl/fi/gnvz49s1cviissqrxtq4v/R250_HV_1.jpg?rlkey=lsp8l18ibxw90nxbfxfz4myl9&raw=1

https://www.dropbox.com/scl/fi/1nnurjxipvwz8zpaeulpy/R250_HV_1.asc?rlkey=w866yjcqrrqud8oikbpxr4ss5&dl=0

The HV is controlled by Vset, which will actually be a 3.3 volt PWM >>>>>>>>> out of an FPGA; saves a DAC.

We can get great 700 volt GaNfets in dpaks, and I expect higher >>>>>>>>> voltages in the future. They are crazy easy to drive compared to >>>>>>>>> mosfets or SiC.

Why not just put some more turns on L2 and use 1 diode?

https://panteltje.nl/pub/GM_tube_400V_HV_section_test_IMG_4195.JPG >>>>>>>>
Cheap 10 to 1 audio transformer in reverse :-)

Seems like a lot of stress on the poor tranny.

Hmmm, a lot of stress is when you do a straight flyback to 5kV,
like ours. Making the "tranny" is a real pain of course, so they work >>>>>> for years 24/7. (we do it on RM8 cores). The advantage of flyback
is that it regulates well down to well below 500V (users need different >>>>>> voltages for different detectors, for Ge I have seen anything between >>>>>> 2kV and 5kV, for NaI etc. with PMT-s go down into the hundreds).

http://tgi-sci.com/misc/DSCF3334.JPG (top right)

What diode is that? What's the big square thing? Where is the
oscillator? Does it regulate?

The diodes I found by sheer chance and I am not sure I know how
to get some more again - good thing I got a decent supply back then >>>>>> (ebay or aliexpress, not sure which).
Look like microwave HV diodes only they are fast, really low losses, >>>>>> something like 10kV rated IIRC. 10-12mm long, about 3mm diameter.

======================================================
Dimiter Popoff, TGI http://www.tgi-sci.com
======================================================
http://www.flickr.com/photos/didi_tgi/

That LTC3803 flyback controller is really nice.

We try to surface-mount everything, and use parts that we already have >>>>> in stock. And avoid ebay or amazon or ali parts.

I'd guess that a classic (line freq transformer type) microwave would >>>>> have slow-recovery diodes.

I guess the newer microwaves have switchers. I think some actually
have semiconductors instead of magnetrons... but magnetrons are cheap. >>>>

Where you can find semiconductors that can work at 2.4 GHz at power
level of few hundred watts?

Apparently GaN.

We'll be driving GaN fets with the 700v power supply that I mentioned.

The GaNs are wonderful to drive. They only need 5 volts on the gate
and the drain-gate capacitance is essentially zero.

It won't be "essentially zero", but something small but finite that a
serious designer would know about, even if they didn't have to do
anything about it.

It must be a drag to always dispute everything so literally. That
instantly kills any traces of creativity.

"Essentially zero" means "too small to matter" to us. The PCB has more
d-g capacitance that the fets we are using.

Joe Gwinn has identified a similar part - the 50V Qorvo T1G4020036_FL

https://au.mouser.com/datasheet/3/1081/1/T1G4020036_FL_Data_Sheet.pdf

and the data sheet doesn't give a drain to gate capacitance - it just
presents a series of Smith charts which encode much the same data.

My my, you made yet another spelling error!

Not a spelling error but a typo - an omission. We all do it. If it
creates a comic effect it can be worth commenting on.

That's an RF part and has RF specs, narrowband s-params and Smith
stuff, presumably to be used with slide rules.

Smith charts - or the tabulated data used to construct them - could be
used as inputs to a computer program. I've never seen such a program,
but I suspect that examples do exist.

We use switch-type
parts that have proper specs, with c-v curves.

But you didn't offer a link to a data sheet or even a part number to
back up your "essentially zero" nonsense.

Many or maybe most RF parts have wirebonds that tune it for one
specific RF band, wherever they think the money is.

Another unsupported speculation.

The bias procedure is typical for RF parts. Slowly adjust the gate
voltage until it works. I see no DC curves. It's probably a depletion
part.

Some day people will use Spice to simulate RF systems, and provide
proper device models.

It's actually possible to concoct your own Spice models - Gummel and
Poon aren't the only people who have done it. It's the sort of tedious
slog that gets inflicted on graduate students, and interns.

It's almost certain that some poor graduate student has been stuck with
doing that for an RF part to get a Ph.D. The part will probably be
obsolete before the thesis is submitted, but the student will be
remarkably employable after graduation. Not by you, of course.
--
Bill Sloman, Sydney

--- Synchronet 3.21a-Linux NewsLink 1.2