Menu
▾
▴
#54 Coupled inductors error
Hi, I am working on a netlist with a bunch of coupled inductors. I however get the following error on the following netlist, and I do not understand why. I have coupled them pair wise as per page 80 on this manual http://ngspice.sourceforge.net/docs/ngspice-manual.pdf
The Inductive System consisting of l0 lm1 lm2 lm3 lm4 k1 k2 k3 k4 has an incomplete set of K couplings, (missing ones are implicitly 0)
*circuit I1 0 1 dc 0 ac 1 PULSE (0 5 1u 1u 1u 1 1) L0 1 0 2.5208581772134705e-09 CL1 0 9 7.883376547497128e-14 LM1 9 10 6.302145443033676e-10 LS1 8 9 6.302145443033676e-10 CRL1 7 0 1.1085733333333333e-13 R1 7 8 0.05410091055648551 B1 0 7 V=0 CL2 10 13 7.883376547497128e-14 R2 11 12 0.05410091055648551 B2 10 11 V=0 LS2 12 13 6.302145443033676e-10 LM2 13 14 6.302145443033676e-10 CRL2 11 0 1.1085733333333333e-13 CL3 14 17 7.883376547497128e-14 R3 15 16 0.05410091055648551 B3 14 15 V=0 LS3 16 17 6.302145443033676e-10 LM3 17 18 6.302145443033676e-10 CRL3 15 0 1.1085733333333333e-13 CL4 18 21 7.883376547497128e-14 R4 19 20 0.05410091055648551 B4 18 19 V=0 LS4 20 21 6.302145443033676e-10 LM4 21 22 6.302145443033676e-10 CRL4 19 0 1.1085733333333333e-13 RLOAD 0 22 1.5839019060551192 K1 L0 LM1 1 K2 LM1 LM2 1 K3 LM2 LM3 1 K4 LM3 LM4 1 .control ac dec 10 1 100k write data.raw .endc .end
On 2020-12-31 14:56, Ian Jannasch wrote:
The inductive matrix is incomplete. I supposed you wanted tight coupling
between all 4 windings and filled in the rest.
These coefficients of all 1 will be the next problem: it results in a
non-definite positive matrix. Luckily it worked when I specified all
couplings to be 0.999.
Could you tell me what this circuit is supposed to do?
-marcel
Last edit: marcel hendrix 2020-12-31
Thank you for the answer, this does run, but I believe is not 100% what I need.
So this is the model of a 4 winding rogowski Coil, with LMx the mutual inductance of each winding and L0 the total, which is seen at the 'primary' or conductor being measured. (so compare current through I1 and L0 to voltage at 22)
In LTSpice I had thie coupling modeled as L0 LM1 LM2.....LMn and this gave what I think is reasonable results, but here something does not feel 100% right.
On 2020-12-31 17:37, Ian Jannasch wrote:
If you have a two winding transformer with turnsratio n and
mutual inductance L0, K (the coupling factor) is a function of
both n and L0. Given that all your K factors are 1, there is by
definition no leakage inductance and thus n must also be 1.
This is probably not how your Rogowski coil was constructed.
For a 4 winding transformer the coupling factor between two of the
windings has a quite complicated relationship with the other 9 free
parameters -- actually there is no unique model and more than one
solution exists. In that case it is better to use the
(canonical) Cantilever approach which uses ideal transformers
and explicit inductors that translate 1:1 to measured port
inductances.
If you show the explicit circuit diagram of your transformer
(10 independent parameters) I could be more concrete.
-marcel
Hi, it makes sense, as per this analytical model the assumption at this point is made that there are no leakage(ofcourse the goal is to adapt that later on). I am actually just trying to model the coupling between the wire in the middle and each separate loop.
Attached is a picture of the model, the R is the internal resistance and the capacitance values can be ignored for this discussion, The self inductance is the Ls values and the one we are currently interested in is the L_mutual values, every winding is moduled separately, so for n=10 there will be 10 repititions of the shown loop. (I hope the image is clear).
What I had in LT spice gave good results at higher turns values when compared to a numerical analysis.
On 2020-12-31 20:32, Ian Jannasch wrote:
It is ok when the turnsratios are all 1. If they are not, you have to
realize that this circuit has specific characteristics in the real
world: it is linear, passive and symmetrical. A bunch of arbitrary
coupling factors and turnsratios does not automatically guarantee
that, and NGSPICE specifically checks for the above mentioned
properties by testing if the inductance matrix is positive definite.
If you are doing actual measurements, it might be an idea to choose
a model that by design is symmetrical and possible, and then use
System Identification to find its parameters.
-marcel
Last edit: marcel hendrix 2021-01-01
makes sense, thank you for the time and help.
-Ian