FEA modelling of Litz wire

[Spehro Pefhany]

Anyone done FEA modelling of power magnetics incorporating litz wire?

Creating a physical model of the individual strands in a 3D winding
looks a little challenging..

 

[legg]

Anyone done FEA modelling of power magnetics incorporating litz wire?

Creating a physical model of the individual strands in a 3D winding
looks a little challenging..

You might look here first:

http://engineering.dartmouth.edu/inductor/papers.shtml

http://engineering.dartmouth.edu/inductor/programs.shtml

They use piece-wise linear analysis and develop physical models for
litz wire. There's a matlab app as well, that might be useful.

FEA modeling of mixed materials has always looked like a head-ache to
me. Takes a while to figure out that the CAD package can't do it, too.

RL

 

[Spehro Pefhany]

It's probably (definitely!) easier to breadboard.

John

The cores are custom ferrite shapes- iterations are a bit slow and a
bit expensive.

 

[legg]

The cores are custom ferrite shapes- iterations are a bit slow and a
bit expensive.

They already exist, or are you forced to specify a new shape?

Never had much trouble finding core suitable core shapes, but never
had much luck convincing mfrs to produce them in 'outside of the
application' material. Complex shapes can also be prototyped from
simpler structures.

FEA modeling with sheet/strands should tell you a lot about non-litz
window fills - at least enough to know whether litz is potentially
justified. Is this just a layering issue or a predictably troublesome
core shape?

RL

 

[Spehro Pefhany]

They already exist, or are you forced to specify a new shape?

It's all special/custom, but at least standard materials.

Never had much trouble finding core suitable core shapes, but never
had much luck convincing mfrs to produce them in 'outside of the
application' material. Complex shapes can also be prototyped from
simpler structures.

It's amazing (to me) what can be done when unit cost is not very
important.

FEA modeling with sheet/strands should tell you a lot about non-litz
window fills - at least enough to know whether litz is potentially
justified. Is this just a layering issue or a predictably troublesome
core shape?

RL

I'm hoping we'll be close enough, then just do an a/b comparison with
the litz in a prototype.

 

[Spehro Pefhany]

Dremel?

John

CNC

 

[legg]

It's all special/custom, but at least standard materials.

But they know what they want........

It's amazing (to me) what can be done when unit cost is not very
important.

Meaning the sky's the limit, or that they can make anything needed,
for peanuts, in volume?

 

[Spehro Pefhany]

But they know what they want........

Sure, it should work reliably. ;-)

Meaning the sky's the limit, or that they can make anything needed,
for peanuts, in volume?

Meaning, if I don't care too much about unit cost (unfamiliar
concept), they can make whatever I can imagine (almost).

 

[legg]

Sure, it should work reliably. ;-)


Meaning, if I don't care too much about unit cost (unfamiliar
concept), they can make whatever I can imagine (almost).

Well, then. I would advise you to do a simple optimum volts per turn
at frequency sweep, based on the power loss budget, before fiddling
with winding structures. For any isolated topology with leakage
limitations to power transfer duty cycle, there will be an optimum
frequency range for the core volume, or an optimum volume range for a
spot frequency.

Ferrite core designers seem to have forgotten that most parts,
nowadays, are fixed with a close-to-unity turns, resulting in a
situation that does not permit physical interpolation of core loss
data in increments of less than 1/2 order of magnitude in a regulated
system.

RL

 

[Spehro Pefhany]

Well, then. I would advise you to do a simple optimum volts per turn
at frequency sweep, based on the power loss budget, before fiddling
with winding structures. For any isolated topology with leakage
limitations to power transfer duty cycle, there will be an optimum
frequency range for the core volume, or an optimum volume range for a
spot frequency.

Ferrite core designers seem to have forgotten that most parts,
nowadays, are fixed with a close-to-unity turns, resulting in a
situation that does not permit physical interpolation of core loss
data in increments of less than 1/2 order of magnitude in a regulated
system.

RL

What's the best way to deal with a non-negotiable large amount of
leakage inductance?

 

[legg]

What's the best way to deal with a non-negotiable large amount of
leakage inductance?

Either resonant, or clamped recovery, preferably with recovery
directly into the intended load.

It's not non-negotiable if you have control over the magnetic
structure.

RL

 

[TheGlimmerMan]

Anyone done FEA modelling of power magnetics incorporating litz wire?

Creating a physical model of the individual strands in a 3D winding
looks a little challenging..

What FEA are you trying to resolve? Thermals? Mechanical stresses?

Or are you talking about electrical characterization of such items?
THAT is NOT FEA.

 

[TheGlimmerMan]

It's probably (definitely!) easier to breadboard.

John

Truly, the only way to properly "select" an inductor or transformer is
to experiment and characterize in the lab. I spent two years going over
a power supply for the med industry where a single turn more OR a single
turn less than the final number made a huge difference.. THEN simply
using Litz improved it further.

Rather than trying to model Litz, one should use standard models for
ordinary transformation. THEN use different Litz configurations and
FIND the best one for that form factor xformer and operating frequency.

There will ALWAYS be an improvement. THAT is all one needs to count
on. Put numbers to the standard, and KNOW that you will get better
performance from the change to Litz (even simple bi or tri filar yields
results). The inductor was the same way. The result is/was not merely
better performance, but the unit ran down to a lower drop out point on
the line side as a result as well. So it would work all the way down to
3.5 Volts when it used to drop out at 6 Volts.

 

[TheGlimmerMan]

It's all special/custom, but at least standard materials.

Once you have had one batch made, subsequent batches cost far less.

 

[legg]

On Feb 3, 5:22 pm, Spehro Pefhany <[email protected]>
wrote:

Magnetic shield further away?

George H.

You mean a BIGGER box ?

Pure heresy.

RL