I am designing a transformer for a traction inverter using the AP method.

1) I have chosen a Metglas core with a saturation flux of 1.56 Tesla. I have chosen a operating flux density of 0.6 Tesla. What is the practical limitation preventing me from running the core at 1 Tesla or more, is it simply core loss?

2) At full output the core loss would be 1.3kW at 0.6 Tesla and 3.2kW at 1 Tesla. Copper loss is another 300W. I can attach to core to the cooling plate shared with the switching devices. Is cooling 1.6kW to 3.5kW reasonable?

3) At full output the MMF will cause a decrease in AL value and hence decrease in inductance. Are there guidelines for acceptable decrease or is the evaluation done using circuit simulation

4) Using 400A/cm^2 current density (good value?) and 30kHz switching frequency I have determined the primary will require three parallel layers of 80mm by 0.508mm copper foil with 7 turns. Is it acceptable to have the parallel layers uninsulated with each turn insulated or must all conductors be insulated from each other.

5) How are the foil ends terminated? Soldered to like thickness copper at a right angle to form a tab?

6) What materials should be used to gap the core, is hard plastic acceptable

Thanks for the help![enter image description here]1

  • \$\begingroup\$ The practical limitation is Remanence and high saturation currents when switch Off and On are out of phase random events. But ok if customer employs controllers with sensors memory and IGBT drivers to avoid this. \$\endgroup\$ Commented Feb 28, 2017 at 3:58
  • \$\begingroup\$ I suggest you use ganged multi-tap converter with AC-DC-AC(HF)-DC(HF)>AC 3phase with single phase shift, Soft switching and ZVS then perhaps you can reduce margin on core if all DC voltages are balanced \$\endgroup\$ Commented Feb 28, 2017 at 4:14
  • \$\begingroup\$ Have you considered nanocrystalline steel laminates and the oil direct air forced (ODAF) ? with multi tap windings. \$\endgroup\$ Commented Feb 28, 2017 at 4:20
  • 1
    \$\begingroup\$ Core losses >> copper losses sounds like an unbalanced design, generally best efficiency is when they are more equal. Is this a transformer? If so, why do you want to 'gap the core'(6). Core gaps are only normally used for flyback type converters. All foils insulated should improve current distribution and hence losses. However, that's to be weighed against thermal insulation and physical size of the insulation. While 400A/cm2 sounds high to me, your copper losses are << core losses, so it's obviously not a limiting factor. \$\endgroup\$
    – Neil_UK
    Commented Feb 28, 2017 at 5:45
  • 1
    \$\begingroup\$ What is the "AP method"? Also, what do you mean by this statement: "At full output the MMF will cause a decrease in AL value and hence decrease in inductance"? In other words, what has full-output got to do with this? \$\endgroup\$
    – Andy aka
    Commented Feb 28, 2017 at 9:43

1 Answer 1


Too long for a comment but not really an answer...

I would observe that the core loss vs flux density curves plotted on P.2 of the Metglas document cover the range 0.05 to 0.3Tesla despite the saturation flux density being quoted as 1.56T on page 1.

This is a pretty good hint that the expected range of operation is within that flux density range.

I don't know how you've calculated core loss at 0.6T and 1T but if you used the formula on that graph I'd caution against extrapolating it to flux densities outside that range - or even using the core outside that range at all - without explicitly talking to the supplier's application engineers and gaining their approval for what you're doing.

That formula is clearly an approximation (anything with a power of 1.74 is a "best fit" from plotted results not a physically realistic formula!).

So, given that it's an approximation of a non-linear phenomenon, you can expect unpleasant surprises if you apply it outside its applicable range.

As far as cooling is concerned ... if you are dissipating 1.3kw in this core, you can't expect to remove that heat by simply bolting one side of it to a heatsink, without knowing and modeling its thermal conductivity - or thermal resistance - neither of which I can find in the data so far - nor do we know the physical dimensions.

We do know the saturation flux density decreases with temperature (to 1.1T at 300C) and the Curie temperature is 399C. I can see thermal runaway leading to hard saturation. So operating at a flux density of 1T looks like a bad idea in any case.

Given a model of its thermal resistance you can model the resulting temperature increase resulting from core loss - and decide on an acceptable level of core loss to meet a realistic temperature.

I will freely admit to being out of my depth at these power levels and only scraping the surface of these issues. But they are real issues. And I'll repeat my suggestion to talk to the FAE.

But my guess is that you'll have to redesign for the manufacturer's recommended range of flux densities, which will probably reduce the core loss to something closer to the copper loss for a more efficient design. That will in turn reduce the cooling requirements and the problems resulting from temperature increase.


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.