High Frequency Low Voltage Push-Pull Transformer Design

Hi,

The application is:
Push-Pull topology, running at 1MHz.
Output approx. 5W, 14V.
Input 24V
Controller: LT3999

In this circumstance I need to design a custom transformer with dual primaries and dual secondaries. Reading between the lines of what LT have done with their demo board, minimizing leakage inductance seams important, I guess to avoid having to use lots of lossy snubbers.

Given this, what is the best layer stack, of the typical makup of a push-pull transformer, assuming that 1 wining - 1 layer, eg I don't think its practical to have more than one layer per winding.

Do I need to put the primaries on top of each other, of do I do a pri, sec, pri, sec split. Would it make any difference?

Low leakage transformers are a challenge.
The decision to make it a pushpull increases the difficulty level. But it can be done.

With pushpull transformers it is very important that each of the two halves in both the primary and secondary are balanced, to minimize staircase saturation.
This means bifilar winding, in addition o interleaving.
In an effort to explain myself better, I've divided the winding segments in colors as shown below.



I would wind:
first layer; red and green
second layer: blue and light blue
third layer: pink and lime green.

Of course, you would have to keep tabs on all the wire starts and ends and connect them accordingly.

You don't mention he primary/secondary isolation requirements. You may get away with very thin or no tape insulation.

If the leakage inductance is still outside your requirements, you may need further segmentation and interleaving.

At 1 Mhz, and depending on the calculated wire thickness and the required losses, each individual segment in turn may require to be bifilar. I suspect that the secondary being the higher current one, may require this technique.

Wurth Elektronik manufactures a transformer for that IC, look at datasheet and calculate your own transformer.
https://katalog.we-online.de/ctm/datasheet/750314781.pdf

The datasheet answered my isolation question.
The transformer is only rated for functional insulation.

Indeed, I only require functional isolation for my design.

There was nothing in the WE documentation about the transformer layer stack, hence my questions.

I notice that their leakage inductance is about 0.5% for their inductance, which I believe is quite good/low.

For info, I have inherited a couple of pre-existing designs both of which have four completely separate windings and they do not perform very well.

You're better off using off the shelf transformers if possible. Coilcraft also makes push pull transformers for this chip: **broken link removed**. I would try the PA6383 with the primary and secondary reversed.

schmitt trigger said:

Low leakage transformers are a challenge.
The decision to make it a pushpull increases the difficulty level. But it can be done.

With pushpull transformers it is very important that each of the two halves in both the primary and secondary are balanced, to minimize staircase saturation.
This means bifilar winding, in addition o interleaving.
In an effort to explain myself better, I've divided the winding segments in colors as shown below.

View attachment 139476

I would wind:
first layer; red and green
second layer: blue and light blue
third layer: pink and lime green.

Of course, you would have to keep tabs on all the wire starts and ends and connect them accordingly.

You don't mention he primary/secondary isolation requirements. You may get away with very thin or no tape insulation.

If the leakage inductance is still outside your requirements, you may need further segmentation and interleaving.

At 1 Mhz, and depending on the calculated wire thickness and the required losses, each individual segment in turn may require to be bifilar. I suspect that the secondary being the higher current one, may require this technique.

Click to expand...

Thanks again schmitt trigger.

Out of interest I went for exactly as you suggest regarding the layers. Also used two strands per winding due to the high frequency. It is now working well.

Glad that I could help.

Back in the late 1980s or early 1990s we got a request for quote from IBM. (When they still were in the mainframe business).

Anyways, the transformer was also a pushpull but rated for something like 50 amps. The designed was a mix of interleaved bifilar windings and copper foils. Several layers. The most complex transformer you can imagine.
I ran and characterized the prototypes. It was designed by our best engineer, which would patiently explain the intricate design details to rookie engineers like me.

It was a real pain to manufacture the contraption, and get good yields.
After the first year of production, we attempted to renegotiate the sale price. IBM decided to move the production to Taiwan instead.