Resonace estimation/AR filter

[VolkerMuehlhaus]

Hi Thorsten @thliebig

I recently had a model where energy convergence was the key to getting accurate results, and my usual -40dB looked good (no visible ripple in results), but upon close examination the results was quite far off. The model really needed -50dB or better -60dB.

https://github.com/VolkerMuehlhaus/gds2openEMS/tree/main/more_examples/measured_vs_simulated/more_accurate_models_L6n2#energy-end-criterion-applied-to-inductor-model

Commercial FDTD usually have some resonace estimation/AR filter to compensate for the error from stopping early, and I now started to explore and implement such a solution in my gds2openEMS workflow. Question: Instead of implementing that in my openEMS-based workflow only, I wonder if this is something to implement into openEMS directly? Claude Opus 5.5 seems to be very capable to work on this topic.

Best regards
Volker

[thliebig]

Sure why not. I think openEMS would be the correct place for it I guess. There are very early attempts in this, but not very successful.
I am sure that an LLM won't have much issues with it these days.

[VolkerMuehlhaus]

Hi Thorsten @thliebig

here is a document for my implementation in gds2palace:
https://github.com/VolkerMuehlhaus/gds2openEMS/tree/main/more_examples/resonance_estimation

Claude also created a document for the openEMS maintainer how a possible native implementation (directly in openEMS) could look like:
https://github.com/VolkerMuehlhaus/gds2openEMS/blob/main/more_examples/resonance_estimation/openEMS_native_implementation.md

Important: The approach was tested with various examples from my RFIC work, but no other use cases.

[VolkerMuehlhaus]

BTW, I opened an issue on a misleading message when openEMS is stopped early:
thliebig/openEMS#236