Category Archives: Allgemein

Open Source EM for IHP: September 2026 Update

Posted on by

Many new features and capabilities have been implemented for openEMS and Palace EM workflows, thanks to coding support by Anthropic Claude.

Major new features shared by both workflows

  • XML stackup files now support derived layers (e.g. for IHP resistors), variables (including variable override from the EM model code) and relative position of metals and dielectrics. Documentation can be found in the docs folder of the repositories.
  • A new graphical stackup editor is now available for creating and modifying the XML stackup files, including import from Keysight ADS *.subst and *.ltd files.
    For gds2openEMS, this is built into the gds2openEMS package. For gds2palace, the stackup editor is built into the setupEM tool. SetupEM is the GUI tool to create and run models for the gds2palace workflow.
  • The gds_prepare_for_EM package is designed to clean up GDSII layout files prior to simulation, to remove all those details and “decoration” that would slow down EM simulation a lot. This package was completely redesigned and now does an amazing job, much more robust than the initial version.
  • Documentation is now available in *.md format also.

Changes specific to the gds2openEMS flow

  • One major change is that documentation and examples now install gds2openEMS as a module from PyPi (pip install gds2openEMS), instead of the local copy of the modules folder used before. The old method with local copy still works, but the new method using Python module installation is more convenient.
  • The second major change is that examples and documentation now use the settings[] dictionary syntax for simulation model settings, similar to the gds2palace workflow. The previous model syntax with loose variables ist still supported, so you can choose whatever suits your needs. Long term, the settings[] dictionary syntax is preferred because it allows to switch between openEMS and Palace workflows more easily.
  • The repository for the openEMS flow was renamed to gds2openEMS. The previous repository name and URL will be redirected, so it doesn’t break.

Changes specific to the gds2palace flow

  • Completely redesigned the core mesh algorithm, which previously had thrown Palace MFEM error message for certain stacked chip configurations. Now, metals are properly cut from dielectrics no matter if they cross dielectric boundaries.
  • There is a new setting to override refined_cellsize for specific layers, for example if you have a complex ground plane but don’t want to mesh that with the same level of detail as the signal trace. See settings[‘refined_cellsize_override’] in the User’s Guide. This setting is not yet available in the setupEM GUI.
  • Surrounding air layer thickness may now be 0 on one or more sides, for example to use PMC boundary symmetry or to use a PEC boundary as ground. Sheet metal layers from GDSII layers may now be located on the outer simulation boundary, overriding the boundary material.
  • gds2palace can now create thermal simulation models for Elmer, a multi-physics FEM solver. At this moment, thermal models can be excited by user defined thermal sources, see documentation in doc folder.

Changes to setupEM (graphical user interface for gds2palace)

  • The setupEM GUI can now start Palace simulation in WSL (Windows Subsystem for Linux) and report simulation progress to the setupEM log. This is now similar to running setupEM on Linux, it is no longer required to start Palace in the WSL terminal window manually.
  • setupEM now provides a built-in S-Parameter viewer to inspect results.
  • The setupEM package was extended to support thermal simulation models for Elmer. There is a separate ‘setupThermal’ program included in that package for building thermal models, see documentation in the repository. Elmer thermal simulation can be started directly from setupThermal.

GDSII geometry cleanup prior to simulation

Layouts are usually simple and clean in the initial design phase, but simulating a „final“ layout that was already prepared for tape-out with density rules etc. can be a challenge.

To solve these issues and create a more simulation-friendly layout, a collection of tools is provided at https://github.com/VolkerMuehlhaus/gds_prepare_for_EM

Open Source FEM: Create AWS Palace models from GDSII layout files

Posted on by

Palace is an open source FEM solver created by AWS. It can be scaled from single computer to large high performance simulation clusters and cloud-based computing.

To simplify the use of Palace for RFIC EM simulation, I created the gds2palace workflow. Input is GDSII layout file + technology stackup file, output are model files for the Palace FEM solver.

github repository of gds2palace:
https://github.com/VolkerMuehlhaus/gds2palace_ihp_sg13g2

Quick overview of gds2palace with examples:
https://github.com/VolkerMuehlhaus/gds2palace_ihp_sg13g2/blob/main/doc/IHP_Palace_Workflow_evaluation_October%202025.pdf

Documentation including User’s Guide:
https://github.com/VolkerMuehlhaus/gds2palace_ihp_sg13g2/tree/main/doc

AWS Palace FEM solver home page:
https://awslabs.github.io/palace/stable/

Open Source EM: New user friendly openEMS workflow for SG13G2

Posted on by

At FSiC 2025, I presented a new user friendly openEMS workflow, with many improvements over using native openEMS, e.g. automatic meshing.

Conference slides
Github download

The files provided here enable openEMS EM simulation with layouts created for the IHP SG13G2 RFIC technology.

The difference to the workflow available at https://github.com/IHP-GmbH/IHP-Open-PDK/tree/main/ihp-sg13g2/libs.tech/openems is that we now have an additional layer of abstraction, to simplify the re-use of model code and enable this new functionality:

  • Automatic mesh generation
  • The model code can load files directly from GDSII
  • Ports (in-plane or via port) are defined in GDSII file on special layers
  • The technology stackup is read from an XML file
  • Merging of via arrays is supported
  • Touchstone SnP file output is supported

Overall, models in this new workflow looks much cleaner because most “routine” stuff is moved into external libraries.