1

I am trying to model a transformer using coupled electrical and magnetic circuits. SPICE has many advantages in assembling these circuits. I was able to model properly for an transformer with no load, although the solution is a little slow.

When I try to add a secondary to the transformer, the simulation is prohibitively slow.

Does anyone have any clue as to what the problem might be?

TR with no load

TR with secondary

Luiz Oliveira
  • 1,024
  • 5
  • 15
  • 2
    Last image, right two circuits - no nodes have any reference to ground. That typically causes excessive GMIN stepping and slow results. Try adding a 1GIG resistor to ground. – rdtsc Jan 27 '21 at 18:31
  • Wow!! @rdtsc That was the problem... in this case, i could ground directy, as in the image 1. Thank you! – Luiz Oliveira Jan 28 '21 at 13:48
  • 1
    @rdtsc Hey hey hey! You stole my best tip there! – winny Jan 28 '21 at 18:35
  • That looks like an unnecessary convoluted way to represent a [transformer](http://ltwiki.org/index.php?title=Transformers). Have you conidered two coupled inductors? If nonlinearity is needed, use the [Chan core](https://electronics.stackexchange.com/q/260510/95619). – a concerned citizen Jan 30 '21 at 00:26
  • @aconcernedcitizen, I was trying to model a 'topological-correct' transformer. In this models, some additional features can be modeled/checked, as flux densities in core limbs, yokes, tanks... also, some exotic connections can be done... zig-zag, phase-shift transformers... there is a lot of this in ATP... – Luiz Oliveira Jan 31 '21 at 13:48
  • @LuizOliveira Well, you know what you're doing, but I'd recommend using the simple voltage source instead of the behavioural one (as you did in your 2nd picture), and for simple `constant*I(device)` choose H or F sources, and, similarly, E or G for `constant*V(node)`; it's likely you'll have better chances for convergence and get faster results. But, unless I'm wrong, your `MUR(x)` seems to be `s/l/table()`, and then `REL(x,s,l)=l/s/MUR(x)`, which simplifies to `table()`. Also, you can accept your own answer, it helps future searches of the same topic, showing an acepted answer. – a concerned citizen Jan 31 '21 at 14:45
  • I tried recreating your first schematic, and it looks like you're trying to model a saturable transformer. Maybe it helps, maybe not, but there already is a nonlinear Chan core, and also a 3rd party library with the Jiles-Atherton model. If not, the behavioural inductor (`Flux=<...>`) can also be used to model simple saturable inductors. See the link I gave and the manual under `LTspice > Circuit Elements > L.` for more (if needed). – a concerned citizen Jan 31 '21 at 14:51
  • @aconcernedcitizen, thanks for the tips... about the MUR table, I was using it this way because I had the data for a specific configuration, now I have fixed it according to your suggestion. – Luiz Oliveira Feb 01 '21 at 13:22

1 Answers1

1

@rdtsc was right. If I ground the 2nd and 3rd circuits separetely, LTSpice finds a solution in an acceptable time, it is still not that fast, but this is probably due to non-linearity.

Luiz Oliveira
  • 1,024
  • 5
  • 15