High voltage PSU

[IanB]

You have sharp step changes in the square wave on the primary side. Step changes will induce ringing in any circuit that has any resonant properties. Is there a way you can soften the square wave so it has softer rising and falling edges, getting it closer to a sine wave or possibly a sawtooth wave?

[Maister]

Yes, I cant get the ringing in simulation as well, when I add some capacitors and reduce the coupling factor.


I am experimenting with some higher-value gate resistors to slow down the rise time, but it gave me a huge current peak. Since my FET driving IC has a fixed dead-time, I realized that there is a time when both MOSFETs are turned on or half way turned on and since the primary windings are wound in opposite direction with same amount of turns, there is no inductance any more and the current is limited by nothing any more, except a little leakage inductance.

I will try to slow down the rise time while not having them overlap now. I also used logic level MOSFETs and turned on their gates with 10V, which might also causes a quick rise time. I will change them to some IRF840 or so.

[Maister]

I changed the MOSFETs from IRL3803 to IRF830 and now it looks way better (see the screenshot as attachment).
Gate resistor is 110Ohms to charge the gate and 55 Ohms to discharge it (with fast diode).

So there is no ringing any more, but I can't explain the huge current peak, which remains. This peak does not vanish, when I use only one of the primary windings.

By the way: I have a SB 340 Shottky diode across each MOSFET to compensate for the bad body diodes.

[Maister]

I forgot to attach the attachment

[Neilm]

Have you looked at adding snubbers across the primary windings?

[jimmc]

How about using vacuum rather than silicon?
The PD500 (http://www.mif.pg.gda.pl/homepages/frank/sheets/010/p/PD500.pdf) or PD510 (http://www.mif.pg.gda.pl/homepages/frank/sheets/010/p/PD510.pdf) were designed as EHT shunt regulators at up to  25kV for colour TVs and will easily sink 0 - 1mA.
Long obsolete of course but still available on ebay (http://www.ebay.co.uk/itm/1-PC-PD510-Philips-Shunt-Stabilizer-Triode-Tube-Valve-/220502123346) etc.

Jim

[Maister]

I measured inductance, leakage inductance and capacitance of the transformer, like    rbola35618 showed in his video, which he postet on page 2.
Capacitance turns out to be about 3uF and coupling factor is 0.997. With these values I am getting the same current peak in LTSpice, as in the real curciut. So that made it clear: such a high capacitance is charged when the MOSFET turns on and therefore the high initial current.

Even with a very slow gate charging (I used 1kOhm gate resistance and added 10nF to the gate capacity) the current peak is still there.

A snubber is a capacitor and a resistor in series, right ? That will increase the capacitance even more. Or am I wrong?

Thanks jimmc, thats a great idea. But I am not allowed to use obsolete parts.

[Maister]

I read about this current peak. It seems to be a common problem of transformers with a high with a high winding ratio like mine has (7 turns primary and 1470 secondary), because the winding capacity is very high.
There are winding techniques to minimize the interwinding capacity, like back winding or z-winding. I combined both of them back winding and z-winding to build a new transformer and this one has also 1.8uF, which is better than 3uF, but still too high. The current peak occurs with the same height.

Maybe I will put a transistor in a current-limiting configuration in series to cut the peak at some current and live with that loss.

[rbola35618]

You may be Abe to use the capacitance to your advantage and design a resonant flyback. That is what we do at work with high turn ratio transformers. You might to do a google and if there any design that you can use.

[Neilm]

A snubber is a capacitor and a resistor in series, right ? That will increase the capacitance even more. Or am I wrong?

The most basic snubber would be a diode, cap and resistor. They are placed in parallel with the primary. When the voltage starts to rise, the diode conducts and charges the capacitor. The resistor is there to discharge the cap ready for the next one.

There are several app notes on the web that show how to calculate the values.

[Marco]

A snubber does him no good ... he doesn't have a flyback converter.

[Maister]

Thank you all for your great comments and ideas!

My Thesis is now done.

I attached an image of the final circuit used in my design, just for your interest.
"Greinacher-Kaskade" is german for "Cockcroft–Walton generator"

The design is able to deliver an output voltage between 1kV and 20 kV, adjustable. Input is 24V. All DC. Output ripple is about 8V pk-pk, measured @ 15kV output with a  100Meg load.

Thank you all for your good advise.

[oldway]

I did not read all the posts of this topic, but i think that what you are looking for is something who does allready exist.
In color TV, we needed a stable 25KV voltage for feeding de CRT anode.
This was done by shunt regulation of a flyback HV. (a flyback power supply is a current source.)
Whe had than to sink part of the current in the shunt regulator.
For this application, it has been developped a high voltage triode with high anode dissipation: that is the valve/tube : PD500.
Specifications of HV triode PD500:
Heater: 7.3V 0.3A
Anode supply: 25KV
Max anode power: 30W
Max anode current: 1.6mA
Base: B9A
Using this triode as a shunt regulator, you can sink up to 1.6mA at 20kV, as needed.
On the marked, there are a lot of high voltage power supplies (20kV or more), but they are not able to sink current.
With one of those commercial HV supplies + a shunt regulator with PD500, you do what you wanted to do of a very simple and clever manner.

[T3sl4co1l]

I don't think this was mentioned before....... why not use an off-the-shelf or salvaged flyback transformer?  Preferably from a portable TV, which will have a 12V primary.  Or the kind with an accessible core, wind your own primary, and do the dumb 2N3055 driver (single or push-pull) that everyone does.  Integral diode, fully potted construction, no worries about high voltage breakdown ever.

As for sinking, an HV tube would do exactly that (over here, we had 6BK4C and etc., octals), though that would require an additional ~200V bias supply.  Just as well draw enough current through the voltage sense resistor divider to account for it, and double the supply capacity.  Another two watts, in the lab, who cares?

Tim