gds2palace enables an RFIC FEM simulation workflow where GDSII layout files are simulated using the Palace FEM solver by AWS. setupEM provides a Python-based graphical user interface to configure and run gds2palace, instead of creating the simulation model code manually, and also start simulation in Palace.
When you install setupEM, the gds2palace workflow is automatically installed in the background. This enables creating a simulation model for AWS Palace. To actually run the simulation, you need to have AWS Palace installed, as described below. Palace installation is not done automatically!
An overview of the SetupEM user interface is given below in chapter "Using setupEM"
setupEM creates and runs simulation models for the AWS Palace FEM solver engine. The underlying solver AWS Palace can be installed in multiple ways. For a smooth interaction with the gds2palace workflow, it is recommended to create some scripts that help running the model and convert the Palace results to SnP Touchstone files.
For development of this workflow, Palace was installed using the Singularity/Apptainer installation method. This was rather simple and straightforward, even with no knowledge about container usage. The resulting apptainer file palace.sif can be integrated very easily in a Linux system like the Ubuntu 24.04 system used here, and can then be moved to other Linux machines using simple copy of the container file. The script to start Palace from the apptainer is included in the scripts directory in this repository.
Notes on installing the Palace solver using apptainer container manager: Installing Palace using Apptainer
Using the spack package manager, Palace can also be created from source with a few simple commands. All tools required by the build process will be downloaded and installed automatically by spack, so you can sit and watch while your system builds the software.
Notes in compiling Palace using the spack package manager for Linux: Installing Palace using spack
Thread on compiling Palace using the spack package manager for MacOS: Spack install for MacOS outdated?
You can use any of the installation methods described on the AWS Palace web site. The gds2palace workflow does not change, it only creates the input files for Palace and does not care how you installed Palace, or on what platform you run the actual Palace simulation from these model files. To start Palace from setupEM, a wrapper script run_palace is used, and this is where you point to your actual installation (even remote copy & remote simulation is possible).
As a Python program that uses the Qt library, setupEM works on Linux, Windows, MacOS and other platforms. The Palace solver itself is designed for Linux systems, but can you install it using the Windows Subsystem for Linux (WSL). Palace also works well on MacOS, installed using spack as described above.
To install setupEM, activate the Python venv where you want to install.
Documentation for the gds2palace workflow assumes that you have created a Python venv named "palace" in ~/venv/palace and installed the gds2palace module there.
If you follow these instructions, you now need to activate that venv and then install setupEM and dependencies via PyPI:
source ~/venv/palace/bin/activate
pip install setupEM
Later, if you want to upgrade to the latest version, you can do
pip install setupEM --upgrade
If you see this error message when trying to run setupEM:
qt.qpa.plugin: From 6.5.0, xcb-cursor0 or libxcb-cursor0 is needed to load the Qt xcb platform plugin. qt.qpa.plugin: Could not load the Qt platform plugin "xcb" in "" even though it was found.
you need to install additional Qt libraries:
sudo apt update
sudo apt install libxcb-cursor0 libxcb-xinerama0 libxcb-xkb1 libxcb-icccm4 libxcb-image0 libxcb-keysyms1 libxcb-randr0 libxcb-render-util0 libxcb-render0 libxcb-shape0 libxcb-shm0 libxcb-sync1 libxcb-xfixes0 libxcb-xinput0 libxcb-xv0 libxcb-util1 libxkbcommon-x11-0
The setupEM module also installs these dependencies:
- gds2palace
- PySide6
- scipy
- requests
To start setupEM, open a terminal window and activate the venv where you installed the setupEM module. Then with the venv activated, you can simply type setupEM to start the module main program.
The user interface of setupEM is organized in multiple tabs, which guide you through the model setup and simulation process. Behind the scenes, the setupEM user interface creates Python model code for gds2palace, and you can check the resulting code on the "Code" tab.
Colors in the user interface: In setupEM, yellow input fields always require your attention, whereas white fields can often be left to default values.
On this tab, you configure input files:
- GDSII layout file that provides geometry information and
- XML file that provides stackup information.
The fields for GDSII and XML file support drag & drop or you can use the Browse... buttons.
Some pre-processing of the layout is also defined here: You can specify a distance (in micron) which is used for via array merging, to speed up simulation by replacing many individual vias with one large via box. If your layout includes polygons with holes, you need to set the "Preprocess GDSII file" checkbox, otherwise you will get error messages during meshing.
Using the "Show stackup" button, you can visualize the stackup and see material information. Note that XML files for FEM simulation in Palace are different in some details (e.g. MIM) from XML used for the openEMS flow. This is because each stackup is optimized for the specific simulation method. Using the openEMS stackup for gds2palace might result in errors durining meshing, using the Palace stackup for openEMS might result in slower simulation.
Dielectric materials are color coded, to easily identify different permittivities. For metal layers, sheet resistance and thickness and distance to other layers is displayed.
On this tab, you configure the frequency range for simulation. Palace is configured to use an adaptive frequency sweep, so that dense sweeps with many points are created from a limited number of EM simulations. However, in general, more frequency points will take more simulation time.
If you want/need to simulate specific fixed frequencies, you can enter them in the "fpoint" field.
If you want/need to add specific fixed frequencies and store the resulting fields to disk for visualization in Paraview, you can enter them in the "fdump" field instead. All these frequency lists will be combined before simulation.
FEM simulation can't simulate at 0 Hz DC, but in setupEM you can specify start frequency 0 and the workflow will handle this behind the scenes: instead of 0 Hz, two low frequency points of 10 MHz and 20 MHz will be simulated, and the data will be extrapolated to 0 Hz in postprocessing.
As explained in the gds2palace documentation, ports are created by drawing rectangles on special layers to the GDSII file. For in-plane ports these must be rectangles, for via ports it must be lines (box with zero size in x or y direction). In setupEM, you then map the source layer for each of these ports, and define direction and target layer(s). Direction matters for polarity if multiple ports are connected to the same ground.
In the upper part of the tab, you make your settings and then need to apply to the port list below.
Palace does not evaluate the port voltage parameter, but we use it internally to specify if ports are "active" during S-parameter simulation. To get the full S-parameters, all ports must be simulated with non-zero voltage, which means that all ports are excited one after another. If you want to simulate only selected excitations, for faster total simulation time, you can set ports to zero voltage. In that case, you will not get full S-parameters data and the missing paths are set to 0 in the output file.
On this tab, you control the mesh used for FEM simulation, which has an effect on simulation time and accuracy. Finer mesh is more accurate, because it can model the actual fields in more detail, but it takes more time to solve.
Parameter "Mesh refinement at the edges" does what the name says, this is parameter "refined_cellsize" in gds2palace code. This is the mesh size used along the edges of polygons. If the actual geometry is smaller than this value, local mesh size will result from geometry dimensions, so this value does not specify a lower limit for global mesh size (as done in the IHP gds2openEMS flow). In many cases, a value of 2 or 5 microns will give great results for IHP SG13G2 simulation models. For physically small layouts, you might go down to 1µm, but small details will be included anyway, no matter what your setting is.
In the FEM workflow, we model conductors using surface impedance on the side walls, and don't need to mesh into skin effect. This is different from the IHP openEMS workflow (gds2openEMS) where solid conductors are modelled and refined_cellsize is used to (partially) mesh into skin effect! For this gds2palace FEM flow, that is not the case, and we can use much larger mesh cell size.
Parameter "Mesh cell maximum size absolute" works in combination with the cells/wavelength value, the mesh will use the lower of these two dimensions.
Parameter "Mesh basis function" is an expert setting that controls the order of FEM basis function. Use setting "most accurate", which means order=2 for basis functions. Only for a quick & dirty simulation, use "faster/less accurate", if you know what you are doing.
Parameter "Adaptive mesh iterations" does what the name says: Palace offers adaptive mesh refinement (AMR) but if we use mesh basis function order 2 ("most accurate") with mesh refinement of 2 micron or so, the initial mesh is usually fine enough and we don't need AMR. Starting from a coarse mesh plus AMR usually takes more simulation time than going for a finer initial mesh without AMR. If you experience something different, your feedback and example is much appreciated!
For the boundary conditions, absorbing boundary and pefect electric conductor are supported at the present time. You can specify the oversize of the dielectric layers from the metal drawing, and the additional layer of air that srrounds everything. Both these distances must NOT be zero, otherwise you will get mesh errors!
Here, you sepcify the target directory where your Python model code and simulation results are stored. You also need to specify a model name, the default is the name of the GDSII file but you can change this value.
The buttons are used top down: You can first preview the resulting model geometry, then create the mesh (and inspect in gmsh viewer if you wish). Close the gmsh window after each step.
To start simulation, use the "Run palace" button. If you are on Linux, this will start Palace using script "run_sim". If you are on Windows, this will start the Linux Subsystem for Windows (WSL) and open a command prompt in the simulation directory.
To start simulation on Linux, it is required that you have configured a script "run_sim" as described in the gds2palace documentation. You can find a template here in the gds2palace repository.
To start simulation on Windows from the WSL terminal, type
./run_sim
To create Touchstone SnP output from simulation results, please have a look at the scripts directory. Script "combine_snp" runs Python code "combine_extend_snp.py", which scans your directories (working directory and below) and converts simulation results to Touchstone file format. Supported input file format: Palace and Elmer S-parameter data.
Behind the scenes, the setupEM user interface created Python model code for gds2palace, and you can check the resulting code on the "Code" tab.
In the setupEM File menu, you can save and load simulation configurations, and you can also save and load a user defined "Default Settings" configuration. This includes the choice of simulation target directory and all other settings. Settings are stored in a JSON file with file extension ".simcfg". The "Default Settings" will be stored to the user home diretory.
Using "File > Import from *.py model", you can load settings from existing simulation model code, e.g. the examples included in the gds2palace repository. This import is based on detecting known keywords, with or without the settings[] syntax, and also works for openEMS Python models. Note that openEMS substrates model the MIM differently, and parameter "refined_cellsize" will usually be smaller in openEMS simulation, so you need to adjust these settings.
If you are on the "Code" tab, you can also export the Python model code using "File > Export to *.py model". This option is only required if you want to save the model code without running it. Buttons "Create mesh and model file" and "Run Palace" on the "Create Model" tab will also save the model code to the target directory, and run it from there.
In the Help menu, you can find links to relevant documentation pages. Also, there is an item Help > Version Information that gives information on your installed versions of gds2palace and setupEM and the latest version that are available online.
To upgrade gds2palace and its user interface setupEM to the latest version, do
pip install gds2palace --upgrade
pip install setupEM --upgrade
setupEM can be launched directly from KLayout through a helper script that you can download here:
https://github.com/VolkerMuehlhaus/setupEM/blob/main/src/scripts/klayout_setupEM.py
Download and save klayout_setupEM.py somewhere, e.g. ~/scripts
If you then start KLayout using this script, a new menu item is available: Tools > setupEM
#!/bin/bash
/usr/bin/klayout -e -rm ~/scripts/klayout_setupEM.py $1 $2 $3- Right-click → New → Shortcut
- Set target:
"<Path to KLayout>\klayout_app.exe" -e -rm "<Path to script>\klayout_setupEM.py"
- Name it e.g. setupEM via KLayout