Please, is there any practical tip / caveat I should be aware, when winding a high-frequency transformer? I'm used to wind 50/60Hz transformers and they do work ok, but now I'm trying to make a high frequency (50kHz) step-up transformer, and I'm facing problems, probably related to losses. My final goal is to get 1500V on the secondary, but I did not get even close to that.

The core I'm using is ferrite, model NEER-28/17/12-2200-IP12R (datasheet: http://www.thornton.com.br/pdf/ner_28_17_12.pdf), and a matching reel (http://www.eletrodex.com.br/media/catalog/product/cache/1/image/9df78eab33525d08d6e5fb8d27136e95/1/_/1_60.jpg), made of bakelite. I think the reel has a bit thick walls, keeping the windings about 1.5mm distant from the ferrite, I don't know if this is an issue.

The PWM signal I'm using is 50% duty cycle, 50kHz, 0V to 12V, and fed to the transformer using a mosfet. I can see a clean and strong PWM signal on mosfet output, I'm sure it's switching completely on and off.

Here's the circuit I'm using in my tests:

enter image description here

Q2 acts as a level shifter, since the PWM signal comes as 0V - 5V, and the mosfet must oscillate on 0V - 12V. R3 and R4 act as a high impedance voltage divider, (1/100), so I can safely measure higher voltages. My goal is to make the transformer produce 1500V in the secondary.

A test with identical primary and secondary, with 3 turns on each, with AWG 20 wire, works good: if I apply 12V to the primary, I can see the same wave on the secondary. A bit distorted, but I guess it's normal, since I'm using a PWM (it's not a sine wave).

However, when I try a secondary with thinner wire (AWG 38) and a few hundred turns, the wave I see on the secondary is horrible, completely distorted, and resulting RMS voltage is way below what it should be.

So, are there any tips on how to lay the turns? I'm pretty confident this is where the problem lies.

  • Should I lay them side by side, covering the core lenght, or is it better to stack them up and use a shorter lenght of core?

  • Should I stack primary and secondary? Or keep them separate, with no overlap?

  • Does wire thickness have any influence on noise and distortion?

  • When the winding reaches the core end, should I go back slowly in the opposite direction, doing that as many times as needed, or should I bring the wire perpendicularly to the winding start, and start over where I began, making all layers winded in the same direction?

Any pointers will be greatly appreciated! Thank you all very much.


  • 1
    \$\begingroup\$ Pretty sure you're experiencing core saturation, the ferrite core has a practical finite bound on magnetic field flux. Since it's working with three loops and not with hundreds. Might be able to calculate the saturation flux knowing your material. Also, are you running without any load on the secondary? \$\endgroup\$ Commented May 24, 2016 at 15:54
  • 2
    \$\begingroup\$ There are many second-order parasitic effects that come into play at high frequencies that are negligible at power-line frequencies. Leakage inductance and self-capacitance become very significant, as do their interactions with the series resistance of the wires. Also, what you connect to the windings (source and load impedances) becomes much more significant. This topic is far to broad to address here unless you supply a lot more details about your specific application. \$\endgroup\$
    – Dave Tweed
    Commented May 24, 2016 at 15:54
  • \$\begingroup\$ @while1, are you sure? Bmax on this core is 2200, and at such high frequencies, my calculations led me to two turns! Smaller frequencies would demand more turns, but at 50kHz, the number of turns drops dramatically. \$\endgroup\$ Commented May 24, 2016 at 16:29
  • 1
    \$\begingroup\$ @Marcovecchio It was something to look into. And it sounds confirmed? Since you calculated saturation at two turns, at three turns you saw slight distortion, and at hundred turns you saw teh ugly. Seems like the data fits the hypothesis. \$\endgroup\$ Commented May 24, 2016 at 19:31
  • 1
    \$\begingroup\$ The secondary is wound first, spreading the wire evenly around the core. The primary should have 6 turns minimum even at 12 volts spread around the core, however you will find this circuit is MUCH more efficient with a higher primary voltage, as it means fewer turns on the secondary, less capacitive leakage and an improved Q. At 24 volts your secondary turns are cut in half. \$\endgroup\$
    – user105652
    Commented May 24, 2016 at 20:35

3 Answers 3


No matter how you look at it you are pumping DC into the transformer primary. The FET's source is going to be wanting to produce a square wave from 0V to +12V and the transformer will want a primary waveform that has an average value of zero volts. Somewhere along the line you might be getting saturation because you are not "resetting" the flux in the core. In other words residual flux left from one switching cycle gets built on by the next cycle and the transformer is said to be "walking into saturation".

Get this bit right then start to worry about the secondary because unless you drive the primary correctly you will be fighting a losing battle.

  • \$\begingroup\$ Thanks for the reply, but you mean I should somehow pump -6V - +6V, instead of 0V - 12V? This makes a lot of sense, but how do you suggest I should do it? Perhaps something like an H-bridge? \$\endgroup\$ Commented May 24, 2016 at 17:49
  • 2
    \$\begingroup\$ I certainly did and a H bridge is a really good idea because you then get a 24Vp-p drive voltage and you can reduce secondary turns to 50%. Also the data sheet for the core is insufficinet - AL is 2200 not Bmax. With AL at 2200 nH your total induictance for 3 turns is about 20 uH and I think you'll see massive saturation. \$\endgroup\$
    – Andy aka
    Commented May 24, 2016 at 17:51
  • \$\begingroup\$ Thanks once again, I thought AL was equivalent to Bmax. I will email the factory ASAP to ask about the true Bmax for this core. I'm not sure the H-bridge will solve the whole problem, but it certainly is the solution for a big mistake I made, so I will accept yours as the correct answer. \$\endgroup\$ Commented May 24, 2016 at 18:01
  • 1
    \$\begingroup\$ You should be able to get 1500 volts with a little bit of care BTW. \$\endgroup\$
    – Andy aka
    Commented May 24, 2016 at 18:08

Interesting. You're driving the transformer primary with a single-ended source follower configuration, which means that the voltage is swinging between +12V - VGS (when Q2 is cut off) and VCE(SAT) - VGS (when Q2 is saturated). This is nearly a 12V swing, and that's what you see with a 1:1 secondary that is lightly loaded.

However, note that the current in the primary only flows in one direction. It ramps up at one rate when the voltage is high, and ramps down at a different rate when the voltage is low. During the low part of the waveform, Q1 is kept conducting until the current reaches zero, at which point its VGS drops and it cuts off.

When you use the 100:1 secondary, you now have an impedance transformation of 10,000:1, which means that your 1.1 MΩ divider now looks like just 110 Ω in parallel with the primary. When the output of Q1 goes low, the current cuts off much more quickly, which is why you see such a highly-distorted waveform.

As Andy says, you need to drive the transformer with a bipolar signal so that you can get the current flowing symmetrically in both directions. There are a number of ways to do this:

  • Use a full H-bridge driver with a single power supply
  • Use a half-bridge driver with bipolar power supplies
  • Use a center-tapped primary with a single power supply
  • \$\begingroup\$ Thanks for the reply, Dave. About the distortion with the 1MΩ load, I would like to have a way to measure it with no load, but it seems to be impossible, since any scope probe I use to measure it will have impedances in the order of mega-ohms anyway. I have a feeling that my winding style still have something to do with the distortion, but I need to test more to be sure. By the way, thanks a lot for the center tap idea. I read about it before, but I didn't get the idea before, now it's very clear! Thanks again! \$\endgroup\$ Commented May 24, 2016 at 18:50
  • \$\begingroup\$ To add to your list of drive configurations there is a possibility of using a DC blocking capacitor and a loading resistor to ground to help with the decay current. \$\endgroup\$
    – KalleMP
    Commented Aug 30, 2016 at 9:00

I've tried for a pretty long time to do PWM on transformers, and the best you can do is shoot to avoid transformer saturation. The first thing I noticed about your circuit is that there are no storage capacitors. You really can't build up any charge from the voltage spike unless you store it. It takes at least one cap and one diode to hang onto that charge.

You don't particularly need Schottky diodes and I think you can get away with high-voltage diodes and caps. Don't use resistors to measure the output - use a capacitor voltage-divider circuit. The output of the capacitive voltage-divider should be a MOSFET, which is also non-resistive.

Resistors will kill a high-voltage circuit.

If you really want super-high-voltage circuits, you can use a Cockcroft–Walton multiplier ladder.

I'm also winding my own transformers, with the intent of making 20kV output, so I'm looking more at ways to bring the output voltage down without the use of resistors - which work, but at the expense of time and energy.

I have a youtube channel in which I discuss all of this, and I'll be posting a video on this particular exercise - winding a high-voltage transformer for use in a I've tried for a pretty long time to do PWM on transformers, and the best you can do is shoot to avoid transformer saturation.

  • \$\begingroup\$ Please use some formating to make it easier to read. \$\endgroup\$
    – kruemi
    Commented Apr 29, 2022 at 12:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.