Resources for designing a kilovolt power supply?

[ajikacey]

What I'm trying to do:

In brief: design a 1000V+ power supply with a bit more than 10mA output. Input could be just about anything convenient (any one of these would be fine: 12VDC, 24VDC, or line voltage).

Context:

One of my ebay power supplies was way off in its power settings and it's made me a bit nuts about voltage metrology.  So currently I'm trying to calibrate my 34401A as well as I can at home and it wants 100mV, 1V, 10V, 100V, and kV references to calibrate its ranges.  Getting precision voltage standards for 100V and 1000V seems to get spendy, and this seems like a good excuse to learn more about designing power supplies.

What I've done so far:

The PentaRef should very nicely take care of the 100mV, 1V, and 10V ranges.

I saw a TI application report suggesting stacking REF5010s was a good way to create a decent high-voltage standard.  For 100V I would only need 10 of them and could feed them 100V by building a power supply based on the LT8304.

But finding dedicated flyback converter circuits for 1000V+ seems more challenging. Since the output is isolated I guess I could put 10 of the things in series but that seems a little silly?  I know I've seen Joule thiefs putting out over a kilovolt so there must be a way to do this with cheaper, simpler parts.

Current question:

Can anyone suggest resources for designing a power supply capable of over 1kV at 10mA out of reasonably priced components?

[bob91343]

You could go old fashioned and find a 60 Hz transformer to supply the needed ac, then rectify and filter it.  Or cannibalize an old oscilloscope or TV set.  Or not even cannibalize, just take power from the HV supply.  Will have to regulate, of course.

[mag_therm]

Yes, copy an old 'scope.

For a very low parts count, a "Ringing Choke" DC to DC converter could be used. I recently repaired a vintage HP141T crt display section.
It had a  one transistor Ringing Choke for the -2400 VDC crt levels. As far as I could see, the regulator's error amp varied the base current cut-off to vary the conduction angle.


Attached is a photo of a test transformer I use. It has not been used on a ringing choke, although it could be.
 (It is used on push~pull pwm).
https://app.box.com/s/zgrd6rslhr5q4wwoyx1qr09r1pu3066l
This transformer/ rectifier takes a 34 V sq wave at 30 to 70 kHz  and makes 220 V DC.
So with a bit of tedious winding, you could try for 1100V DC.
Or go for 550 V and use a voltage doubler on secondary, etc.
Happy winding!

[SmallCog]

When I was looking for a KV supply a member here put me onto "Electrophoresis Power Supplies"

These are used in biology labs and show up used on eBay quite cheaply

Perhaps they're a useful source of high voltage that you could precisely regulate?

Buying old stuff from eBay wasn't an option here at work so I bought a brand new Consort EV3150 that goes out to 1200VDC

[james_s]

I think for 1kV I'd try a flyback topology with a transformer wound on a small ferrite core. You might look at some of the small HeNe laser power supplies for ideas, I reverse engineered a few of them several years ago.

https://www.repairfaq.org/sam/laserhsc.htm

[bdunham7]

There's two steps--producing the raw HV and then regulating it to be stable, and adjustable or precise or both.

To produce HV one simple way is to use a transformer to step up 120VAC line voltage to 240VAC, then use a voltage multiplier, probably a quadrupler.  You can use a higher voltage transformer, but they are harder to find very cheaply, whereas a voltage quadrupler is 4 diodes plus 4 capacitors.

https://www.allaboutcircuits.com/textbook/semiconductors/chpt-3/voltage-multipliers/

Regulating it with enough stability to be worthy of checking your 34401A is another matter.  100 REF5010s?  Typically HV supplies that are expected to be stable and precise use a voltage divider, precision reference and feedback loop.  I have several devices that are in this category, one is the HP 6515A which you can find on eBay.  It is fairly stable after an overnight warmup.  I can't really think of a good way to easily replicate any of their designs easily with jelly bean parts.  I would suggest using the same method for the 100V and 1000V sections. 

If you are really only interested in checking the meter rather than wanting a power supply adventure, you can check the 1000V range with 100V and still be pretty confident it is working correctly.

[ajikacey]

100 REF5010s?

That's the idea. It sounds crazy but in that paper the TI engineers built a 100kV reference with 10,000 of them

And I appreciate the variety of suggestions, it looks like I've got lots of ideas to think about now!

[bdunham7]

That's the idea. It sounds crazy but in that paper the TI engineers built a 100kV reference with 10,000 of them

Yes, I saw that.  I have a 5500VAC R-transformer laying around that I took out of a Hi-Pot tester and I've thought about making a 10kV reference.  In that case, if you are willing to buy 100 REF5010s and put them on cards or something, then my simple transformer/quadrupler will work fine and be low noise to boot--and you would have a pretty accurate reference without needing voltage dividers or calibration.

[bob91343]

And let us not forget a microwave oven transformer.  They put out serious volts.

[james_s]

And let us not forget a microwave oven transformer.  They put out serious volts.

Not really, they're only about 2kV, what they do put out is serious current and they are thus extremely dangerous to work with. Not only that, anything reasonably modern is cost engineered to the max so they run at the brink of saturation even with a load. A MOT is an absolutely terrible way to accomplish what the OP is trying to do here.

[Teledog]

Look up HV inverter - transformer - supply on Aliexpress.
The black tubular things are fairly high voltage and only for intermittent use.

[David Hess]

Does it need to be variable output?  Can the output be positive and negative 500 volts centered on ground?

At that current level, you are into big fluorescent bulb power levels so I would use a Royer converter and CCFL transformer.  Check out figure H1 on page 37 of Linear Technology application note 55 to get an idea of what is required at that level of power and voltage, but voltage feedback will be required instead of current which is easy enough.

[Alex Nikitin]

1) Why do you need 10mA? For most meters the input resistance at 1kV is 10M, so only 0.1mA is required, and if the supply can produce 1mA it is more than enough. It also would be much safer to work with.

2) Do a search on Bay for "PMT supply". Photomultiplier supplies are not expensive, usually fairly stable and controllable. For example this one.

Cheers

Alex