0

Using a technique to model transformers for a wide frequency range, I got at a circuit like this one: https://1drv.ms/u/s!AhR86nrv-ZA7ivoz9BZIoXfs4GwXIQ?e=L6mcKd

LTSpice takes a long time to solve this circuit, being prohibitive. If I remove the mutual inductances, the circuit is solved much more quickly. However, mutual inductances do not change the number of nodes in the circuit.

Does anyone get a clue on how to solve the circuit quickly on LTSpice?

*Some background about the circuit, it's very similar to this one below, but bigger and with a lot of mutual inductances.

enter image description here

Luiz Oliveira
  • 1,310
  • 5
  • 16
  • 3
    Add the schematic as an image, not a link to a tool that can't display it. –  Jan 28 '21 at 19:50
  • 1
    Add a schematic screenshot and/or a netlist if its short – Voltage Spike Jan 28 '21 at 19:56
  • Can you reduce the computational effort by increasing the steps?? i.e. don't do an analysis at every 0.000001Hz if you only care about 10Hz resolution. – Kyle B Jan 28 '21 at 21:41
  • If you won't show the schematic, try reading this. The number of nodes is not everything when trying to solve the circuit, a considerable part is the care taken to build the schematic. Just because it's a simulator that can run at a click of a button, it doesn't mean you can throw anything in there and bam!, you have a space-ship. – a concerned citizen Jan 28 '21 at 21:50
  • 6
    Does this answer your question? Slow solution of LTSPICE electric-magnetic model for transformers – Elliot Alderson Jan 28 '21 at 22:05
  • @BrianDrummond, sorry, but I can't show the schematic because I've solve only by netlist. Too big to draw. I will edit the question to help. – Luiz Oliveira Jan 29 '21 at 19:34
  • @VoltageSpike, the netlist isn't short. – Luiz Oliveira Jan 29 '21 at 19:34
  • @KyleB, unfortunately, increasing the time step does not help. – Luiz Oliveira Jan 29 '21 at 19:34
  • What about the schematic? – Voltage Spike Jan 29 '21 at 19:34
  • @aconcernedcitizen, the link contains great suggestions, but I didn't find anything to help me, unfortunately. I have already solved many circuits like this in LTSpice, but as the circuit grows with many mutual inductances, the solution takes too long. From what I read, I think it may be related to the fact that matrices are no longer sparse when there are many mutual inductances. – Luiz Oliveira Jan 29 '21 at 19:35
  • @ElliotAlderson, no... I created that topic and luckily there was an adequate response. – Luiz Oliveira Jan 29 '21 at 19:35
  • 3
    Have you tried introducing tiny leakages to ground from any problematic node? Or many nodes. – winny Jan 29 '21 at 19:51
  • What is the purpose of this simulation? – Bruce Abbott Jan 30 '21 at 00:04
  • @LuizOliveira If all the inductors are to be coupled with the same coupling, then a single statement like K L1 L2 L3 ... will do. Otherwise, that schematic should not pose any problems, particularly since it looks to be completely linear. If you could share the contents of your LTspice .asc file, I'll take a closer look (otherwise, recreating the schematic by myself won't ensure that I have the same build as you). Plus, the picture is very fuzzy for my eyes. – a concerned citizen Jan 30 '21 at 00:29
  • @BruceAbbott, this circuit represents a transformer in an adequate manner for any signal applied... since 50Hz to some tens of MHz. So, it can be used by transformers manufacturers to evaluate voltage distribution along the windings for any input signal: lightning impulse, switching impulse, momentary overvoltages, etc. – Luiz Oliveira Jan 31 '21 at 13:52
  • @aconcernedcitizen, unfortunately, i don't have an .act, just the .cir... I fill it automatically using a spreadsheet... after all, it uses some coupled circuits in each element (that represents a turn, or a disk winding) to represent the losses in this element for any frequency. – Luiz Oliveira Jan 31 '21 at 13:58
  • @LuizOliveira Then can you be sure that you have the exact same nodes as in the picture? What I mean is, while not impossible, verifying a netlist for connections is far more prone to error than a visual schematic. Plus, generating the netlist is as simple as clicking a menu entry (Tools > Export Netlist). – a concerned citizen Jan 31 '21 at 14:47
  • How long does the simulation take? How long does it take if you halve the number of elements? – Bruce Abbott Jan 31 '21 at 19:01
  • @BruceAbbott, I cannot remove half of my circuit, but I can remove the coupled branches that generate the frequency-dependent losses. Each set of these adds a node for each node in the circuit. In that case, I got:
    • Complete circuit (3 coupled branches) [624 nodes]: 1h10
    • Circuit with 2 coupled branches [468 nodes]: 29min
    • Circuit with 1 coupled branches [312 nodes]: 5min
    • Circuit without coupled branches [156 nodes]: 25s

    So I believe that the biggest chance is that the problem is really the number of nodes and not the mutual inductances.

     – Luiz Oliveira Feb 01 '21 at 15:49

0 Answers0