Hello.

I am designing a network of audio transformers and I created a LTSpice model. Each transformer has three windings, same polarity, same number of turns
I used the Chan model and the LTWiki trick of putting the Chan modeled inductor in parallel with one of the three coupled inductors.
http://ltwiki.org/index.php?title=Transformers (section : "How about transformer saturation effects? )
The LTWiki states that in order not to load the Chan inductor, the inductance of the ideal coupled coils should be at least 10 times higher than the Chan modeled inductor.

I estimated the Chan inductor to have a 0,3mH inductance. So I gave my ideal inductors a dummy inductance of 4mH.
Now I am trying to understand the limitations of this trick. Even if thus I can model the saturation / hysteresis, I thought that, for instance, my modeled transformer impedance would be 4/0.3 larger than it really is.

Should I take that into account for my impedance matching ? (the audio transformers are subjected to an audio line-in signal, so it's probably in the 100 Ohms impedance range)

Do you see any other obvious limitations or things to take into account ?

 

I think I figured it out.

Suppose an ideal transformer (K1 L1 L2 1) with the L2 winding paralleled with the chan inductor L3
L3 non saturated inductance being in my case 0.3mH.
In order not to load L3, L2 is set to more than ten times higher than L3.
In this case, L2 is set to 4mH.
the resulting inductance of L2 || L3 being (L2*L3)/(L2+L3) is then close to 0.279mH
Since in my case the turn ratio between L1 and L2 is 1:1, I just setup L1 to be 0.279mH
Obviously I could have chosen L2 to an inductance way larger to make the resulting inductance closer to 0.3mH

It was really simple and I feel a bit stupid, unless i forgot something once again.