25kHz Switching Transformer 600W 300V to 1200V PCB Mountable

1200W transformer to mount on a PCB?
How big is the board?

It will be small in size, the frequency is 25KHZ not 50/60 HZ.

right, Approximate size 2"x2"X1" for 1000W at 25kHz.

Ooopps, didn't spot the frequency, well I did but I missed the k. I'd got visions of you as He-Man sliding boards the size of a coffee table in to a rack!
There was a small company making specialised transformers near to where I live until recently. It's closed now, I don't know if they relocated or went to the wall. The owner was a friend of a friend I find out if he knows.

I don't know of any companies local to you but ferrite transformer manufacturers are your best bet for a special custom transformer you are asking for.

Have you tried Globalspec or an internet search and phone call of potentially suitable transformer manufacturers?

During my 15 years as an engineer and bottle-washer, running my 30-man oceanographic-instrument company, I (we) encountered such issues frequently. That's because 10 to 75kHz is the sweet-spot for long-distance undersea acoustic transmission, and we were using acoustics in all kinds of ways: data telemetry, bubble and particle backscatter sensors, integrated temperature-density profiling, doppler water velocity, iceberg depth, subsurface wave recording, seafloor-bottom motion, etc. Yep, and 1 to 3kV is the sweet spot for piezo transducers working at 0.5 to 5kW peak power.

We found the most sensible thing was to wind our own small ferrite transformers. It's not hard; at these frequencies you're usually talking under 100 turns, or not much more. Hah, at first we didn't even have a turns-counting setup. All by hand, sitting at the bench, adding layers of Kapton tape appropriately, and in just a few minutes the finished product emerges. With the right instructions, this would be a perfect exercise for students.

The fun part comes after its finished, and you crank it up and measure the kilovolts. If you're brave, you touch a wire as an experiment, youch! It may be high frequency and relatively harmless, but it hurts and leaves a small wound. Now that's an awesome demonstration.

I can see you have good experience about high voltage transformer winding, i have a question regarding Kapton tape,do you put it between each two layers or only over the final layer ?

Thanks for help.

First, I like the low-voltage primary on the inside, with a layer over it. The low-voltage end of the secondary starts next, that way the high-voltage end is on the outside. We slide a teflon sleeve over the HV wire for the last turn and bring it out from under the outside tape layer, away from the transformer winding.

We make a judgement as to the voltage difference building up between parts of the windings. For example, between two layers it's only the part that comes back over that builds up to the full amount of two layer's worth of voltage. So you could just do tape between halves of one side, and then the same on the other side, etc. I like to double-insulate (wire+tape=double) every 400 to 500V-worth of winding buildup. And always a full layer (or two) on the outside. But when you finish the bobbin has to fit into the ferrite cores, without insulation-damaging cramming. :-)

For truly high-voltage windings (over 5kV) multiple sections machined and separated are necessary. That's another story entirely. We're talking under 2.5kV here.

My technician used to like frequent tape layers to hold the finished windings in place. But sometimes he had to take it apart and rework to fit. His transformers almost never had breakdown failures.

Happy windings!

Yes, the primary side low voltage is preferred due to low voltage controls are much easier and voltage amplification by a factor of 3 to 4 is preferred at the output.

For 1000V output at 1W current of 1A plus will flow through secondary and part of it goes into high frequency harmonics filtering. About 80-90% efficiency forces primary to nearly 5A peak current level which is OK for MOSFETs and IGBT.

Do you have some idea about such high power transformer? Ferrite core of about 2" diameter with sufficient magnetic material and toroidal winding will be used.

One thing I will add is that the waveforms on the primary and secondary usually don't match. First, the secondary can be setup to be resonant - even tho the transducer is partially coupled to the metal or water (much less so in the case of air), it usually appears more capacitive than resistive. Second, the leakage inductance is often fairly high - given the need to isolate the secondary from the primary, no interleaving is practical, raising the leakage inductance.

This can mean you'll have a sine wave on the secondary, and a square wave on the primary. It also means some of the secondary current may be resonantly cycling, and the primary current may be less than the turns ratio implies. And it also means the secondary voltage may be higher than implied by the primary. In the extreme case (air) a full factor of two benefit is possible.

Leakage inductance is a significant issue. Spice modeling is very helpful, but you want to use real numbers in the model. It's easy to measure your finished transformer's leakage inductance, using the old shorted-winding trick. I like to short the primary while measuring the secondary, etc. It's also not too hard to calculate the leakage inductance in advance from the proposed winding dimensions.

It's a little harder getting the winding resistances right in the model, because the ac-resistance proximity effect is so large. I use an HP 4192A impedance meter to take measurements. There are graphs for estimating that as well. I rely on Snelling's excellent book, Soft Ferrites. Especially chapter 11, Properties of Windings. It's well worth the cost (oops! $440 at Amazon). Oh well, just measure the DC copper-winding resistances and multiply by 10 to 20. Or by 50 or 100, see graph. :-)

Load is about 330nF capacitive from ultrasonic transducer. As I am varying frequency from 20kHz to 25kHz, it is forced energy rather than tuned resonant frequency. Yes, the driver is push-pull type with some filtering for high frequency before drive power is injected into transformer input.

Actually, with that narrow frequency range, you can still rely on resonance, just with a low Q. Even a Q of two saves a considerable amount of power. You don't necessarily have to explicitly lower the Q, you can just accept less benefit as you move away from f₀, and the drive currents go up.

As for filtering your square-wave drive, I never bothered, that's what the transformer's leakage inductance does, or even add a small secondary inductor for the resonance "filter" if L_ell isn't high enough.

We worked to many kW, and were gratified at the high efficiency of it all.

All right. Assuming that leakage inductance will work as series inductance, I may not add any more of it.

For resonance frequency, i am measuring impedance dynamically, However, i am not measuring efficiency as outcome is not measured unless I use a probe to sense the power in mechanical form.

How, do I measure energy transfer efficiency? Heat? That will be slow. What is other faster method that I can use for dynamic tracking as resonance frequency will change as transducer gets heated up due to energy transfer.

Ultrasonic transducer is immersion type Weber made 500W and in the said frequency domain. It is yet to arrive. It will be used in water by immersion and will be driven by the push-pull 1000V output having 20kHz to 25kHz frequency.

Well, maybe I should say inefficiency. It's the portion of the power-supply current that ends up as heat. Switching mosfet heat, and transformer heat. It's hard to learn the acoustic efficiency. As far as off-resonance is concerned, you watch the supply current rise. Ouch!

Does that mean that low supply current meant effectively better transformer operation?

I measure the output voltage and primary current and keeping the output voltage same, primary current is the measure of efficiency?

I can then see how load responds to output voltage and by measuring its current at the output for normalized output voltage.

Could you please explain how to measure the leakage inductance using the shorted primary method.

Both of you guys are asking Mr.Winfield so many questions about the techniques of his own business but both of you and other business men are watching to pick up and never give ,i think not so many good people as him are left in our world to day.

Thank you, you're very kind. I'm just trying to get a G.A. -- haven't managed so far in this thread. :-) But it's always fun talking shop.

Dear Winfield Hill:

Discussing with you is of great pleasure.

I often procure parts from various supplier but purchasing parts like high frequency high voltage transformer is not all that easy.

Coilcraft supplied me 650VuS transformer with 3kV isolation, perhaps pulse type, which stores impulse energy in inductance and then deformed output appears at the secondary. This type transformers are OK for making DC-DC converters.

Now that I am looking for near sine wave for ultrasonic transducer, your current suggestions are good one. I am considering core types that are of high resistivity so less of Eddy current loss, and high switching frequency to derive proper pulse shape.

I have seen some TDK magnetic core and they look good up to 100kHz.

Discussing engineering matter is very good. In fact some people can pick the ideas and can even offer physical solution by giving commercial offer which is one good reason for all these exercises.

Another serious problem is the type of insulation people use in transformers. I have tested few transformers and they show 1uA dielectric leakage for1500V DC between isolated coils. That looks rather on high side. Then there is large capacitive coupling between primary and secondary coils. As I am having pulse applications, transfer of energy due to capacitive coupling is a serious problem.

I have designed isolated MOSFET driver for 1kV or +/-500V and perhaps it will also work for IGBT. Right now I am generating +500V to -500V pulsing to drive transducer directly having transducer as load connected to ground. Perhaps transformer can help me reduce voltage switching at MOSFETs or IGBT and can also do some pulse shaping. Efficiency is not critical and even 80% to 90% efficiency at 20kHz to 25kHz will do.

Finally you gave the Gust answer about a method to measure the mutual inductance,the timing is too late, where you been for long time sir ?