Non-Linear Inductor and Resistor Models for Transformer Simulation

[FredB]

Hi RG,

I was wondering if you had any more details about how to tweak non-linear inductor and resistor models for transformer simulation. I see the Gossner book lacks any pdf version that I can find. A trip to my library is on my list of things to do.

Non-linear inductor models I have some experience with, it's mostly about setting up the non-linear resistor to match core loss data that I am wondering about.

Best regards,

fred

[jazbo8]

Not sure if this is what you're looking for, but I think it is worth checking out, it's a part of my slow-growing SPICE library.

Jaz

[FredB]

Well, the manufacturers of electrical steels usually give one or two figures for core loss, such as 0.5 watt/lbs at 5 kiloGauss, and 2.1 watt/lbs at 10 kiloGause.

Those figures are two data points on a non-linear, probably experiential, function that describe losses in the core as a function of flux density. I would like to know that function so I can put it in my spice model for particular output transformers.

The complication here is that the steel manufacturers only make the raw materiel, which then has to be fashioned into transformer laminations by a laminations seller. The process the lamination sellers use to finish the raw electrical steel rolls into their products has a significant effect on the permeability and loss figures of the final product.

So without the lamination producers testing their own <censored> products and providing a full data sheet, like all electric component manufacturers do, it's all guess work on the part of transformer builders. For power transformers the published data points are enough to design a power transformer that operates withing the minimum induction and maximum loss figure given for the laminations used.

Although for audio transformer, those details are way too little to do a proper job, at least to my satisfaction. This is why a lot of audio transformer work done on the basis of experience and some traditional charts that were produced based on that experience.

I've found out that testing these things is easy enough to do, one just need a good data acquisition card. I can already tell, I'm going to doing a lot of my own testing.

[jazbo8]

I see, if the producers refuse to release basic specs, then you need to do a lot more work... Intusoft provide some basic equations here, if you don't already have them. Good luck with your project.

Jaz

[FredB]

I see, if the producers refuse to release basic specs, then you need to do a lot more work... Intusoft provide some basic equations here, if you don't already have them. Good luck with your project.

Jaz

That paper is similar to RG's article on transformer modeling. RG's model adds a non-linear resistor for the core hysteresis loss and a linear resistor for the core eddy current loss in parallel with the saturatable/non-linear inductor. Also the paper is missing a capacitors for the inter winding capacitance and self-capacitance of the windings.

My question is how to structure the flux dependent resistance gradient of the non-linear resistor simulating the core hysteresis loss.

Here is an interesting paper http://users.encs.concordia.ca/~pillay/c28.pdf that shows how that hysteresis loss is also frequency dependent. Although, it also lacks any clue about how to construct such a resistor.

[FredB]

I believe I done figured it out.

It requires a presumption without a curve to base everything on. Then it's just a matter of some curve fitting and some calculus.

Without seeing what the core loss curve does above 1.5 kilo Gauss, or what ever the highest given figure is, will limit the reliability of the results to that level.

I've got the curve fitted from the two data points given plus zero. Lookie:



Now I'll going to have to remember how to use my symbolic math app for this...