Build a Push-Pull HV power transformer

[Neukyhm]

Hi forum. I have to design&build a DC-AC converter, Vin=48V 300W, Vout=5KV desired power=150W-200W. I have some ETD59 cores, with 1.5mm gap and without gap.

I first thought about a modified ZVS, but I want to avoid it since this is a college project and I have to give it some theorical basis (and there is a lack of it related to Mazzilli drivers, almost like if people are happy enough if it works xD). So here I am looking for any book or website where I can find information about Push-Pull transformers.

Thank you in advance

PS: how are the two MOSFET in a Push-Pull transformer protected against back emf coming from the primary anyway?

[Zero999]

A snubber network is used to protect the driver transistors.

The greatest challenge is insulating the high voltage secondary winding, which will arc over and destroy itself, together with the driver transistors, if it's not properly insulated.

Does the output over the converter need to be AC or will DC do? If DC is fine, build a transformer with a lower output voltage <1kV, which will be much easer to insulate and add a diode and capacitor voltage multiplier to the secondary, to get 5kVDC. If you need AC, then it's much more difficult.

[Benta]

Hmmm...

You say DC-AC. So what AC frequency are you aiming for?
5 kV - really? In that case, the ETD59 is not a good choice. I'd say a U-U core with separate primary and stacked secondary would be more appropriate.
At 200...300 W and this voltage ratio I'm uncertain as to what I'd suggest. We're talking a 1:100 Pri/Sec ratio here.

I've worked with "Joule thieves", ZVS, Mazzilli or whatever you like to call them and I hate them. Not because of the concept, but because driving a power transformer into saturation to ensure oscillation creates an EMI monster that is basically untameable. People using TV or radio in a half-kilometer radius will hate you.

Concerning books: look for Siemens/Philips/Ferroxcube/Epcos/TDK databooks with "Soft Ferrites" in the title.

[David Hess]

PS: how are the two MOSFET in a Push-Pull transformer protected against back emf coming from the primary anyway?

They are not.  The switch which is off sees twice the input voltage plus whatever comes from the leakage inductance.  That will not be a burden when operating off of 48 volts.

How much regulation do you need?  If a push-pull inverter is not good enough, then given the high output voltage, I would regulate the primary side voltage.

[Neukyhm]

Does the output over the converter need to be AC or will DC do? If DC is fine, build a transformer with a lower output voltage <1kV, which will be much easer to insulate and add a diode and capacitor voltage multiplier to the secondary, to get 5kVDC. If you need AC, then it's much more difficult.

The output must be AC because there will be a voltage multiplier there.

Hmmm...

You say DC-AC. So what AC frequency are you aiming for?
5 kV - really? In that case, the ETD59 is not a good choice. I'd say a U-U core with separate primary and stacked secondary would be more appropriate.

I expect frequency to be below 80kHz.

How much regulation do you need?  If a push-pull inverter is not good enough, then given the high output voltage, I would regulate the primary side voltage.

I don't think I will need regulation. The load of the transformer is a voltage multiplier feeding a resistive load.

[Neukyhm]

I've found this web page: https://www.maximintegrated.com/en/app-notes/index.mvp/id/3835

It uses a RC circuit as snubber, well I could use that. What do you guys think?

[Zero999]

Does the output over the converter need to be AC or will DC do? If DC is fine, build a transformer with a lower output voltage <1kV, which will be much easer to insulate and add a diode and capacitor voltage multiplier to the secondary, to get 5kVDC. If you need AC, then it's much more difficult.

The output must be AC because there will be a voltage multiplier there.

Then why not use a lower AC voltage and more stages on the multiplier?

Hand winding a small 5kV transformer is not easy.

[Neukyhm]

Does the output over the converter need to be AC or will DC do? If DC is fine, build a transformer with a lower output voltage <1kV, which will be much easer to insulate and add a diode and capacitor voltage multiplier to the secondary, to get 5kVDC. If you need AC, then it's much more difficult.

The output must be AC because there will be a voltage multiplier there.

Then why not use a lower AC voltage and more stages on the multiplier?

Hand winding a small 5kV transformer is not easy.

Because I'm so restricted of space in my machine that I don't have space for a bigger multiplier   

[jbb]

I first thought about a modified ZVS, but I want to avoid it since this is a college project and I have to give it some theorical basis (and there is a lack of it related to Mazzilli drivers, almost like if people are happy enough if it works xD). So here I am looking for any book or website where I can find information about Push-Pull transformers.

Sorry, it seems we ran away from your topic a bit.  If you want more detailed info on 'a ZVS,' you would do better to use a real name for it.  ZVS = Zero Voltage Switching, which is a property of a converter, not a specific type of converter.  Try looking for a Royer oscillator/converter. Or a 'self-commutated push-pull converter.'  Sadly, much of the power electronics information that's readily available online (especially on Youtube) is woefully incomplete 'suck it and see' work, without any theoretical treatment.

Also, 5kV / 150W = 33mA.  That would definitely burn you and could kill you, especially if there are large capacitors anywhere.  And you might find that none of the test gear in the lab is rated for it.  If you're actually going to build it, you need to get some safety training, which should be provided by your college.

The output must be AC because there will be a voltage multiplier there.

Ah, so that's why you want AC output from the transformer.  So what you really want is a DC-DC converter consisting of a high frequency transformer stage and a voltage multiplying rectifier.  That's extremely relevant.  You can't just design 'a transformer' and connect it to 'a voltage multiplier' and expect everything to be fine; you have to design all of the converter at once.

I don't think I will need regulation. The load of the transformer is a voltage multiplier feeding a resistive load.

You need to pin this point down.  I'm 99% sure that the 'resistive load' is something more complicated than just a resistor.  Typical guesses would be photomultiplier tubes, X-ray tubes, assorted RF power amplifier tubes, LINAC stages, electrostatic precipitators, capacitor bank chargers etc.  Pretty much everything in that list will actually need some regulation to give consistent performance, so make sure you're getting a good specification for the project.  Remember your 48V supply will have some tolerances too (e.g. if battery powered you're looking at a range of 40 - 55V).  You may have to quietly determine who is designing the equipment that connects to your converter and sidle up to their desk informally.

If space is critical, you'll need to keep an eye on the following:

• Choosing the right switching frequency
• Using a symmetrical topology which applies both positive and negative flux to the transformer and transfers power to the output on the positive and negative cycles
• How much creepage clearance you will need on the transformer secondary winding (hint: LOTS)
• Getting as much primary voltage as possible to reduce the turns ratio (so use a full-bridge inverter).

Based on these criteria, I think that a classic H Bridge, Phase Shifted Full Bridge, Series Resonant Converter, Parallel Resonant Converter or LLC (with full bridge driver) might be a good start.

When you design the transformer, you'll see that the creepage & clearance requirements are critical.  In addition to the primary-secondary insulation, even the end-to-end insulation of the secondary will be critical too.  You may well end up using a sectioned winding approach.  This takes extra space and adds significant leakage inductance.  Benta is right that ETD59s are not an ideal geometry for this situation.

5 kV - really? In that case, the ETD59 is not a good choice. I'd say a U-U core with separate primary and stacked secondary would be more appropriate.

Given these conditions, an LLC with an H-bridge primary side driver might be a good option.  Here are the qualities that could help:

• Uses transformer core symmetrically - improves density.
• Actually uses transformer leakage inductance.
• Generally doesn't need snubbers.
• Can provide voltage gain and regulation by frequency modulation, in addition to the transformer turns ratio.
• • Achieves soft switching of primary switches (subject to constraints on load and operating frequency).  Good for EMI and efficiency.
  • Achieves soft switching of secondary diodes (subject to constraints on operating frequency).  This is really useful because your output multiplayer will have lots of diodes in it.
  
  The downsides:
  
  • Usually controlled by varying the switching frequency.
  • Likely to experience hard switching during initial charge of all the capacitors on the rectifier side.
  • Efficiency isn't fantastic at light load (but improvements are available).

[Neukyhm]

Thank you very much for your reply, jbb. I was planning to make arduino send a pwm signal to a TC4469 driver to run the mosfets.

And you are right lol, it's a x-ray tube
And the multiplier is already built with 12 stages using 1nF caps and 2CL2FM diodes, I will immerse it in epoxy.

So you say the winding will be difficult, I know that. I have design and 3d printed an ultra simple manual winder just for this. What about if I use 1mm width wire? If the ETD59 coil former is 41.2 long, I will have to wind those 41mm a few times but I will decrease the voltage between spires, avoiding breakdown.

Is it better a full-bridge inverter than push-pull?

[jbb]

Thank you very much for your reply, jbb. I was planning to make arduino send a pwm signal to a TC4469 driver to run the mosfets.

And you are right lol, it's a x-ray tube
And the multiplier is already built with 12 stages using 1nF caps and 2CL2FM diodes, I will immerse it in epoxy.

So you say the winding will be difficult, I know that. I have design and 3d printed an ultra simple manual winder just for this. What about if I use 1mm width wire? If the ETD59 coil former is 41.2 long, I will have to wind those 41mm a few times but I will decrease the voltage between spires, avoiding breakdown.

Is it better a full-bridge inverter than push-pull?

OK, so you know what the rectifier stage will be. You should consider that when you design the rest of the converter.

For the final control scheme, I recommend you use some kind of standard control chip. It will be more reliable than Arduino - it won't blow up because of a coding problem. Also you can find some with the gate drivers built in. You will also need some kind of overcurrent protection.

On potting: I haven't potted with epoxy before, but I hear that it's not compatible with ferrite due to different thermal expansion constants (it can crack the ferrite). There are soft, clear, silicone potting compounds which might be more suitable. Finally, there is a big risk of air bubbles which are bad news because they cause big jumps in the electric field and can be the starting point of age-related failures.

On transformer: you need to find out how much creepage and clearance you need. There should be some standard tables for this but I don't know which ones. You'll find that the creepage can be the biggest challenge. I expect your transformer will have quite a bit of air in it.

On push-pull vs full bridge: push pull subjects MOSFETs to 2x DC input (plus spike) and definitely needs some kind of snubber. Gate drive is easier.

A full bridge needs 4 MOSFETs, but each will only be subjected to 1x DC input and may not need a snubber. Gate drive is harder. Transformer primary is a little easier.