Custom transformer

[Laval]

A while ago I made a post concerning a DC "calibrator" project

https://www.eevblog.com/forum/projects/relatively-high-voltage-precision-dac-amplification/

and the biggest problem I had was the design of the 1 KV module. I was then going the amplification way but it was not a good idea for many reasons: requiring HV transistor or cascode circuit, stability and compensation not to mention I still need a 1 KV power supply.

I'm considering a boost converter using a PWN controller like SG3525 (or similar) driving a pair of power MOSFETs in a push-pull configuration. These MOSFETs in turn would drive the primary winding of a step-up transformer. The secondary would then be rectified and filtered then passed to a doubler to avoid insane turn ratio. There would also be a feedback loop using a precision divider and a precision op amp like OPA 189. The output of the op amp would drive the comparator of the controller directly bypassing the integrated error amp. The non-inverting input of the precision op amp would allow the output to be programmable with a DAC.

My problem with this design is the transformer. There is obviously no such transformer so I would have to wind it myself. After looking at transformer parts on DigiKey and Mouser (cores and bobbins), it seems they are intended for low voltage, step-down designs with limited winding window area and insulation margins. Insulation will be an issue and I'm not sure how to deal with it.

Does anyone have a solution to this problem ?

[moffy]

Considering the low power requirement, 20mA@1kV, it might be easier to use a flyback converter and not use a doubler at all. A split bobbin design is the best for isolation, but 1kV is modest for insulation purposes, the yellow mylar tape is commonly used in such transformers. What is your DC input voltage? Do you also have ripple requirements because of the precision aspect?

[Laval]

Considering the low power requirement, 20mA@1kV, it might be easier to use a flyback converter and not use a doubler at all. A split bobbin design is the best for isolation, but 1kV is modest for insulation purposes, the yellow mylar tape is commonly used in such transformers. What is your DC input voltage? Do you also have ripple requirements because of the precision aspect?

As a matter of fact, 20 mA is much higher then I really need. 2-5 mA is way enough. This is to be used on high impedance devices and should not have to source much current.

It is still in the design phase so the input voltage is not decided yet but I had 50 VDC in mind to power the MOSFETs. That makes a big turn ratio (1:22-1:24 assuming a doubler).

And yes, since accuracy is the important end goal, ripple should be as low as possible as well as interference which is a reason why I was leaning toward push-pull.

Quantifying the amount of isolation I need is something I'm looking into. I'm not sure how to evaluate this.

[moffy]

Flybacks are generally a bit noisy, but if you could put a post regulator on the output then the noise can be dealt with. Another way to get low noise is to control the switching edges of the push/pull drivers, slower edges means less noise, e.g. a low noise supply by Jim Williams: https://www.analog.com/en/resources/design-notes/off-line-low-noise-power-supply.html

[moffy]

Another Jim Williams design, a 15W/1000V amp using an off the shelf transformer powered from 28V, see AN18-9: https://www.analog.com/media/en/technical-documentation/application-notes/an18f.pdf
P.S. Transformer is no longer available.

[Doctorandus_P]

The secondary would then be rectified and filtered then passed to a doubler to avoid insane turn ratio

You'd better put it though the doubler before you filter it

Is this for a one-off or do you have commercial (serial production) in mind?

For one-off, you can get plenty of variation of boards form china that have a SG3525 with step up.

And for the rest, there are plenty of commercially available transformers for high voltage applications. CRT transformers are obsolete, but there is stuff from electric fly swatters to cold cathode tube drivers and car bobines.

That old AN18 is a good start, for the rest 1kV was (is?) a pretty common standard for metrology stuff. Once I read an application note that put a lot of shunt regulators in series, and it really was a lot. I think that was for a 20kV reference, an the string of shunt regulators was wound in a spiral around a plastic pipe.

Do you want a fixed or variable voltage?
And at what sort of accuracy and stability?

The little gadget in the thread below goes up to 3.6kV.

https://www.eevblog.com/forum/beginners/how-dangerous-is-4000v/?all

[Laval]

Is this for a one-off or do you have commercial (serial production) in mind?

This is a on-off but if it ever works well enough, it could become open source.

For one-off, you can get plenty of variation of boards form china that have a SG3525 with step up.

I have a few SG3525 handy, they are very unexpensive. I have made a basic circuit on a breadboard to try it (no MOSFET and no transformer, only the SG3525 and a few necessary components) and it works well. It is very stable even with all the parasitic capacitance of the breadboard.

And for the rest, there are plenty of commercially available transformers for high voltage applications. CRT transformers are obsolete, but there is stuff from electric fly swatters to cold cathode tube drivers and car bobines.

That old AN18 is a good start, for the rest 1kV was (is?) a pretty common standard for metrology stuff. Once I read an application note that put a lot of shunt regulators in series, and it really was a lot. I think that was for a 20kV reference, an the string of shunt regulators was wound in a spiral around a plastic pipe.

Using a transformer from another device, at least for tinkering, is a good idea but it is difficult to know the spec of these transformers. Making the transformer myself would give me a lot of flexibility as well as allow for some prototyping.

Do you want a fixed or variable voltage?
And at what sort of accuracy and stability?

As accurate as I can make it    this is the primary goal. I'm aiming for 0.05% (which I probably won't reach) and good stability. Voltage should be variable.

[Berni]

You can get ready made high frequency high voltage transformers for CCFL drivers. These transformers can produce something up to around 10kV AC 100KHz while being driven with 10s of volts. Can be scavenged out of existing CCFL drivers or even bought new.

For for your requirement of only 1000V you don't need to go for any specialty transformers. Most mains voltage ferrite transformers will survive 1kV just fine, here is a selection of brand new transformers for 400VAC:
https://www.we-online.com/en/components/products/WE-OLLT

If you go for a voltage doubler on the output then you don't even need to go out of spec for these transformers since 400VAC is 560V peak

[Someone]

Is this for a one-off or do you have commercial (serial production) in mind?

This is a on-off but if it ever works well enough, it could become open source.

Two ways to get enough voltage rating: use a transformer designed for that range already (coilcraft FL Series or similar), or design in a planar core to the PCB with buried via and then you get the insulation built in cheaply.

Trying to think if there is a way to use coupled inductors to boost a voltage multiplier without exceeding their limited voltage ranges.

[Laval]

You can get ready made high frequency high voltage transformers for CCFL drivers. These transformers can produce something up to around 10kV AC 100KHz while being driven with 10s of volts. Can be scavenged out of existing CCFL drivers or even bought new.

For for your requirement of only 1000V you don't need to go for any specialty transformers. Most mains voltage ferrite transformers will survive 1kV just fine, here is a selection of brand new transformers for 400VAC:
https://www.we-online.com/en/components/products/WE-OLLT

If you go for a voltage doubler on the output then you don't even need to go out of spec for these transformers since 400VAC is 560V peak

These Würth Electronik transformers are great and these CCFL transformers pretty much have the spec I need

https://magnetic-components.mpsind.com/Asset/P9121_Series_Datasheet.pdf

Thanks!

[Someone]

If you go for a voltage doubler on the output then you don't even need to go out of spec for these transformers since 400VAC is 560V peak

For this design sticking within rated voltages is quite the pain with off the shelf magnetics, most certainly with enough volume the manufacturers could uprate the parts.

Sticking to lower voltage parts I arrived at a mid fed voltage multiplier driven by capacitors to standoff the voltage (driven by a resonant stage). Can use 250V diodes, caps, and inductors, except for the final output cap and those standoff caps feeding the autotransformers (film probably for lower loss?). Looks similar in cost except for the beefy inverter required to drive it. Two of 4 stages shown for brevity.

[GigaJoe]

I'm figure out that 1000 V with mV precision would be difficult for me ... I made 2 x 350V fixed and one 95..350 . connecting all 3 1000v achievable

[Berni]

In any case if you need precision you need to have a voltage feedback from the output to regulate it.

You can easily regulate all of these high voltage boost stages by simply adjusting the input voltage. If higher powers are involved this can be done using a buck DC/DC that you tap into the FB pin to give it an input for a extra 'control voltage' from some opamp.