Newbie in need for hints

[wolfgang6444]

I am absolutely new to OpenEMS and EM-simulations in general. Attached is my first simulation-script. I am trying to simulate the inductance of a planar PCB-loop. The ultimate goal is to simulate coupled inductors in the presence of a metal target as used in inductive posi
trial3.txt


tion-sensors.
While the results around 100MHz and the resonance at 1.2GHz are quite plausible, I have some questions:
1)How can I avoid the unphysical negative real part of the impedance at 1.2GHz?
2)Is there a more effective way to achieve my goal:
I observe that the simulation takes drastically more time as I lower the center-frequency. This is somehow plausible, as the
transient simulation needs to last longer and the required minimum timestep is probably somehow imposed by the smalles features
in the mesh. On the other hand these features probably do not make a difference at 100MHz. So is there a way to automatically
'smoothen' the mesh based on the simulated frequency range?
3)Is my stimulating port the best choice? 70nH at 1.2GHz are about 500Ohm, so I am essentially voltage driving my inductor?

Many thanks,

Wolfgang

[VolkerMuehlhaus]

Hi Wolfgang!

I believe the negative real part at resonance is just a numerical artefact if the total impedance is really large, i.e. |Z| >> re{Z}
Does this really matter to your use case?

The lumped port with 50 Ohm is fine, you don't need to tweak the port impedance. I have done many inductor simulations this way and confirmed against measurements and other simulators. That was for RFIC inductors, but similar impedance mismatch.

Regarding automatic meshing: If you are willing to create your openEMS models in Python instead of Matlab, you could use the wonderful automatic meshing add-on that Mustafa presented here:
#421

Good luck!
Volker

[wolfgang6444]

Hi Volker,
many thanks for your reply. Yes the negative real part is not really of any importance. |S11| < 1.004. I just had to clip it to 1 when using vectfit3.
I am struggeling a bit with the DC-resistance: the related zero is around 100kHz so it is virtually impossible to get data from openEMS for frequencies that low. I am using vectfit3 to extrapolate. But I still need to run open EMS with an accuracy limit of 10^-6 to get a reasonable match to the 100mOhm @dc I calculated. That takes > 12h.

Anyhow maybe I should just simulate a PEC-track and then add the DC-resistance as ESR manually.
I am still struggeling to get the SSM from vectfit3 into the control package - some formats must be different.
I will take a look at the automatic mesh generation. I prefer octave for the math, but I can define geometrie and mesh offline in python.
So there is progress ....

Wolfgang

trial5.txt

[VolkerMuehlhaus]

Hi Wolfgang,

agreed, the DC response can be off a bit because we stop the time domain simulation when energy is low enough. We would to simulate "forever" to get the exact DC value, so there's some compromise.

I am not familiar with vectfit3. If you are looking for SPICE model extraction, this is a new code that works well for my RFIC stuff. It offers "black box" lumped model fit based on scikit-rf library, in addition to some fixed topology model that are NOT wideband.

https://github.com/iic-jku/snp2le

It#s easy to use if you are familar with python, just install as a Python module (as described in readme) and type snp2le on the command line to start the GUI.

[wolfgang6444]

That's interesting, but can't spectre simulate touchstone ootb?

Wolfgang

[VolkerMuehlhaus]

@wolfgang6444 Yes, Spectre can, so that's a non-issue in Cadence flows. This model extraction is designed for all-open-source RFIC simulation workflows, which don't have that capability built into the circuit simulators yet.