gdsfactory/gplugins

gdsfactory plugins

gplugins 0.13.0

gdsfactory plugins:

• devsim TCAD device simulator.
• meow Eigen Mode Expansion (EME).
• femwell Finite Element Method Solver (heaters, modes, TCAD, RF waveguides).
• gmsh mesh structures.
• tidy3d Finite Difference Time Domain (FDTD) simulations on the cloud using GPU.
• lumerical For Ansys FDTD and Circuit interconnect.
• klayout for fill, dataprep and testing.
• ray for distributed computing and optimization.
• sax S-parameter circuit solver.
• schematic: for bokeh schematic editor and path_length_analysis.
• meep for FDTD.
• mpb for MPB mode solver.
• elmer for electrostatic (capacitive) simulations.
• palace for full-wave driven (S parameter) and electrostatic (capacitive) simulations.
• vlsir for parsing GDS-extracted circuit netlists into Spice, Spectre and Xyce Schematic File formats.

Installation

You can install most plugins with:

pip install "gdsfactory[full]" --upgrade

or

pip install "gplugins[devsim,femwell,gmsh,schematic,meow,meshwell,ray,sax,tidy3d]" --upgrade

Or install only the plugins you need. For example:

pip install "gplugins[schematic,femwell,meow,sax,tidy3d]" --upgrade

Non-pip plugins

The following plugins require special installation as they can't be installed with pip:

• For Meep and MPB you need to use conda or mamba on MacOS, Linux or Windows WSL (Windows Subsystem for Linux) with conda install pymeep=*=mpi_mpich_* -c conda-forge -y
• For Elmer, refer to Elmer FEM – Installation for installation or compilation instructions each platform. Gplugins assumes ElmerSolver, ElmerSolver_mpi, and ElmerGrid are available in your PATH environment variable.
• For Palace, refer to Palace – Installation for compilation instructions using Spack or Singularity. Gplugins assumes palace is available in your PATH environment variable.

Getting started

• Read docs
• Read gdsfactory docs
• 

gplugins's Issues

path_length_analysis notebook issue

the notebook has some issue

https://gdsfactory.github.io/gplugins/notebooks/path_length_analysis.html

@tvt173

paralell build of notebooks

• How to speed up? Run in parallel. Combine with jupyterbook
• Detect failed notebooks

Expose them as an artifact for

papermill notebook.ipynb

@flaport

Bug in specifying multiple volumes for materials in Palace

Problem

There is a bug in the Palace plugin that prevents using multiple volumes for materials. That is, for example multiple substrates cannot currently be specified.
The problem happens with this dict comprehension, which will contain multiple keys of the same name without combining the values.

gplugins/gplugins/palace/get_capacitance.py

Lines 54 to 56 in cef1c38

	material_to_attributes_map = {
	v["material"]: physical_name_to_dimtag_map[k][1] for k, v in bodies.items()
	}

Solution

Refactor dict comprehension to combine values

Using kweb for gds visualization in jupyter book

I am trying to use the jupyter based kweb gds viewer to display chips on our local lab server. Currently we are compiling markdown and ipynb files using jupyter book. I have managed to get the kweb to display gds files within the browser. I first launch a kweb server using jupyter run and a python script that sets the KWEB port and then launches the server. However, this server is only able to display the most recently executed chip.plot() command when the notebooks are built using jupyter book. The built site is able to communicate with a python kernel using thebe but jupyter book does not currently support using a local kernel. As of right now I am not sure if it is possible to switch between different gds files within the browser.

A dropdown menu might be enough to make this work. A way for compiled jupyter books to interact with the kweb server would also help.

The notebook being compiled with jupyter book:

Purple Chip A

import gdsfactory as gf
import amf
from kweb.server_jupyter import start

chipA = gf.import_gds("/home/st4eve/purple/mask-design/purple-chip-A.gds")
chipB = gf.import_gds("/home/st4eve/purple/mask-design/purple-chip-B.gds")

start()

chipA.plot()

Script that starts kweb server:

from kweb.server_jupyter import start
import os
os.environ['KWEB_PORT'] = '8000'
start()

Command to start kweb server:
jupyter run jupyter_kweb.py

fix dataprep notebook

There is sill an issue with the fill script

#123

@sebastian-goeldi

AttributeError                            Traceback (most recent call last)
File ~/gplugins/gplugins/klayout/dataprep/regions.py:243
    236 fill_cell = d.get_fill(
    237     d[LAYER.FLOORPLAN] - d[LAYER.WG],
    238     size=(0.1, 0.1),
    239     spacing=(0.1, 0.1),
    240     fill_layers=(d[LAYER.WG],),
    241 )
    242 c = d.get_kcell()
--> 243 c << fill_cell
    244 fill_cell.write("fill.gds")

File ~/kfactory/src/kfactory/kcell.py:943, in KCell.__lshift__(self, cell)
    937 def __lshift__(self, cell: KCell) -> Instance:
    938     """Convenience function for [create_inst][kfactory.kcell.KCell.create_inst].
    939 
    940     Args:
    941         cell: The cell to be added as an instance
    942     """
--> 943     return self.create_inst(cell)

File ~/kfactory/src/kfactory/kcell.py:884, in KCell.create_inst(self, cell, trans, a, b, na, nb, libcell_as_static, static_name_separator)
    881 else:
    882     assert cell.layout().library() is not None
    883     lib_ci = self.kcl.layout.add_lib_cell(
--> 884         cell.kcl.library, cell.cell_index()
    885     )
    886     kcell = self.kcl[lib_ci]
    887     for port in cell.ports:

AttributeError: 'Cell' object has no attribute 'kcl'

Show gmsh-generated mesh in Elmer and Palace examples

The electrostatic simulation example in the documentation could show the generated mesh as well. I assume this can be done with pyvista or meshio.

Notebooks to update:

• https://gdsfactory.github.io/gplugins/notebooks/elmer_01_electrostatic.html
• https://gdsfactory.github.io/gplugins/notebooks/palace_01_electrostatic.html
• https://gdsfactory.github.io/gplugins/notebooks/palace_02_fullwave.html

HFSS plugin

We already have a couple of Ansys plugins. How could we add RF design specific plugin using HFSS?

some examples:

• https://github.com/PhilReinhold/pyHFSS
• https://github.com/qiskit-community/qiskit-metal/tree/main/qiskit_metal/renderers/renderer_ansys

@nikosavola
@zlatko-minev
@jagandecapri

Issue on page /notebooks/workflow_3_cascaded_mzi.html

virtual lab + sequencer

How do we define what we want to measure from gdsfactory?

it would be great to define the testing and data analysis sequence

Ideally we can also store it on the GDS itself using labels so that they remain as part of the GDS file

Issue on page /notebooks/tidy3d_01_tidy3d_modes.html

coupler issue in the docs due to serializer

`asyncio.run` not working in Elmer example notebook

Changes from #67 broke running the Elmer documentation. Apparently asyncio calls don't work in Jupyter Notebooks without special handling. I ran the docs as a .py file so this was missed.

RuntimeError: asyncio.run() cannot be called from a running event loop

Support dielectric loss analysis for Palace scattering

Is your feature request related to a problem? Please describe.
Support dielectric loss analysis for Palace if materials with loss tangents are specified.
This requires changes in

gplugins/gplugins/palace/get_scattering.py

Lines 84 to 86 in d342c82

	# TODO list here attributes that contained LossTAN
	# palace_json_data["Domains"]["Postprocessing"]["Dielectric"] = [
	# ]

and

gplugins/gplugins/palace/get_scattering.py

Line 109 in d342c82

# TODO palace_json_data["Boundaries"]["Postprocessing"]["Dielectric"]

Describe the solution you'd like
Handle materials with loss tangents by adding dielectric analysis to corresponding dimtags and add some result fetching for DrivenFullWaveResults.

Employ pyvista trame backend instead of panel

Using the panel backend for pyvista the Elmer electrostatic notebook shows the following

/tmp/ipykernel_8249/763369231.py:3: PyVistaDeprecationWarning: `panel` backend is deprecated and is planned for future removal.
  pv.set_jupyter_backend("panel")

trame is preferred instead.

Necessary changes include modifying the backend in notebooks and updating dependencies for trame instead of panel.

Provide default mesh_parameters.portnames in Elmer and Palace

Currently, specifying ports for Elmer and Palace always requires specifying portnames. By default, the implementations of run_capacitive_simulation_elmer and run_capacitive_simulation_palace should set portnames=c.ports by default.

results = run_capacitive_simulation_elmer(
    c,
    ...,
    mesh_parameters=dict(
        portnames=c.ports,
        ...
    )

add pdk compiler functions

When building PDKs, there are many functions that are shared:

• plot waveguide modes
• save cells into separate GDS files
• Create new PDK
• ....

How about making a pdk compiler module?

Add electrostatic simulations with Palace

Add support for electrostatic simulations with Palace, similar to Elmer done in #42

analog

How could we add analog design worlkflow?

https://github.com/Patarimi/pyWAS

@Patarimi

Issue on page /notebooks/path_length_analysis.html

At the moment we rely on the info for path length analysis

how could we extract path length from a GDS instead of a gdsfactory component?

@nikosavola
@tvt173

Issue on page /notebooks/elmer_01_electrostatic.html

notebook not rendering correctly

@nikosavola

Color the Gmsh entities according to LayerViews

Is your feature request related to a problem? Please describe.
Right now, inspecting meshes in GMSH requires manually toggling visibility of the layers, because there is no obvious coloring scheme.

Describe the solution you'd like
Layers are already associated with colors in gdsfactory, so we could simply call gmsh/model/setColor on the entities when we assign them a physical label associated with a Layer. This would require a change upstream in meshwell to support colors, which could just be another attribute of the meshwell entity.

Issue on page /notebooks/tidy3d_00_tidy3d.html

issues on tidy3d notebook

Bibliography not working in Elmer plugin docs

In https://gdsfactory.github.io/gplugins/notebooks/elmer_01_electrostatic.html#id1. The references smolic_capacitance_2021 and marxer_long-distance_2023 don't seem to work. Either something is wrong with the .bib file or the config selecting it

Add Elmer to plugins listing

Elmer is currently missing from the README and https://gdsfactory.github.io/gplugins/

Add linting for Markdown

Add linting for Markdown files as done in gdsfactory/gdsfactory#2001 for gdsfactory.

femwell/solve_thermal.py

$ python solve_thermal.py 
Traceback (most recent call last):
  File "/Users/lukasc/Downloads/solve_thermal.py", line 219, in <module>
    from gplugins.gmsh.mesh2D import mesh2D
ModuleNotFoundError: No module named 'gplugins.gmsh.mesh2D'

• Mac OSX 11.6.7, Intel processor
• pip install: gdsfactory, gplugins, femwell

I have the latest version of gplugins (0.5.0)

$ pip install gplugins[gmsh,femwell]
Collecting gplugins
  Obtaining dependency information for gplugins from https://files.pythonhosted.org/packages/b9/dd/c790baea29ff148805030d562cbc63f6c0eaa4b428f7d672b5f8ddb7f847/gplugins-0.5.0-py3-none-any.whl.metadata
  Using cached gplugins-0.5.0-py3-none-any.whl.metadata (6.0 kB)
Requirement already satisfied: gdsfactory[cad]>=7.4.0 in /usr/local/lib/python3.10/site-packages (from gplugins) (7.4.6)

Seems like mesh2D is missing:

$ python 
Python 3.10.9 (main, Dec 15 2022, 18:20:40) [Clang 13.0.0 (clang-1300.0.29.30)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import gplugins.gmsh
>>> dir(gplugins.gmsh)
['MeshTracker', '__all__', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '__version__', 'annotations', 'break_geometry', 'cleanup_component', 'create_physical_mesh', 'fuse_polygons', 'get_layer_overlaps_z', 'get_layers_at_z', 'get_u_bounds_layers', 'get_u_bounds_polygons', 'get_uz_bounds_layers', 'list_unique_layerstack_z', 'map_unique_layerstack_z', 'mesh', 'mesh_from_polygons', 'meshtracker', 'order_layerstack', 'parse_component', 'parse_gds', 'parse_layerstack', 'refine', 'round_coordinates', 'tile_shapes', 'to_polygons', 'uz_xsection_mesh', 'xy_xsection_mesh']

Hoping that mesh2D was simply replaced by mesh, I made the change:

from gplugins.gmsh import mesh as mesh2D

And it runs, but then there is another error:

$ python solve_thermal.py 
dict_keys(['substrate', 'box', 'core', 'shallow_etch', 'deep_etch', 'clad', 'slab150', 'slab90', 'nitride', 'ge', 'undercut', 'via_contact', 'metal1', 'heater', 'via1', 'metal2', 'via2', 'metal3'])
2023-09-09 14:10:55.633 | WARNING  | gdsfactory.pdk:get_active_pdk:721 - No active PDK. Activating generic PDK.

2023-09-09 14:10:55.765 | INFO     | gdsfactory.technology.layer_views:__init__:790 - Importing LayerViews from YAML file: '/usr/local/lib/python3.10/site-packages/gdsfactory/generic_tech/layer_views.yaml'.
2023-09-09 14:10:55.766 | INFO     | gdsfactory.pdk:activate:334 - 'generic' PDK is now active
Traceback (most recent call last):
  File "/Users/lukasc/Downloads/solve_thermal.py", line 227, in <module>
    heater2 = gf.components.straight_heater_metal(length=50, heater_width=2).move(
  File "/usr/local/lib/python3.10/site-packages/gdsfactory/component.py", line 2153, in move
    raise ValueError(move_error_message)
ValueError: 
You cannot move a Component. You can create a new Component, add a reference to the other Component and then move the reference.

For example:

# BAD
c = gf.components.straight()
c.xmin = 10

# GOOD
c = gf.Component()
ref = c.add_ref(gf.components.straight()) # or ref = c << gf.components.straight()
ref.xmin = 10

port to pydantic>=2

The core functionality is supported for pydantic2 now since release 0.4.0

However some other plugins are not available when using pydantic2

• gdsfactory
• gplugins #71
• ray ray-project/ray#39162
• sqmodel tiangolo/sqlmodel#632
• dagster dagster-io/dagster#15162

Add documentation on how to write your plugin

Automate lumped ports on either side of waveguides for Palace scattering

Is your feature request related to a problem? Please describe.
Support automatically adding lumped ports on at least waveguides with ports for Palace scattering simulations

Describe the solution you'd like
Helper function or a step abstracted away in run_scattering_simulation_palace.

femwell plugin uses get_material_index which is deprecated

UserWarning: get_material_index is deprecated
warnings.warn("get_material_index is deprecated")

gplugins/femwell/mode_solver.py uses get_material_index which is deprecated

@simbilod
@HelgeGehring

Adaptive meshing for capacitive Elmer

Electrostatic simulations for Elmer support adaptive meshing and there is part of an API for it implemented in gplugins. However, I could not get it working yet. This issue would be to get it working and edit the .sif.j2 file accordingly.

Infer portname delimiter in Elmer electrostatics

The portname delimiter is now assumed to be port_delimiter = "__". The default value of # does not work in Elmer, as it tries to evaluate a statement with the symbol. Supporting the default value should be done by converting layers using it automatically to Elmer supported ones.

Issue on page /notebooks/klayout_dataprep.html

the fill does not look right

@SkandanC
@sebastian-goeldi

Support waveports in Palace scattering

Is your feature request related to a problem? Please describe.

Handle adding waveports to ports at the edge of the simulation area automatically.

Blocked by simbilod/meshwell#20

Describe the solution you'd like
Automatically add a waveport to ports at the edges of the simulation area when running run_scattering_simulation_palace.

improve schematic

How can we improve the schematic capture?

Options:

• gf web: convert the jupyternotebook into a web app
• https://github.com/TeledyneLeCroy/SignalIntegrity
• Use interconnect plugin
• Use a node editor See

@proppy
@tvt173
@sebastian-goeldi
@s910324

Elmer notebook not rendering in docs

Describe the bug
Elmer notebook fails to run in docs

This is caused by Pyvista recently deprecating the panel backend from pip install pyvista[jupyter] install dependency. pyvista/pyvista@0cd21cc

Suggested fix
Change the backend to trame, which is recommended by pyvista.

Tee output from Elmer and Palace to a log file and stdout and stderr

Currently, both of the simulators only write to a log file (which also is not separate for stdout and stderr). Ideally the output would also be teed to stdout and stderr such that it would be easier to debug and catch in Ray for example.

Support distributing Palace scattering simulations excitations with Ray

Is your feature request related to a problem? Please describe.
For solving the full S matrix, Palace requires a different simulation for each excitation. These are currently run in parallel with asyncio but could additionally be distributable with Ray similarly to Automated model extraction

gplugins/gplugins/palace/get_scattering.py

Lines 206 to 222 in d342c82

	tasks = []
	for json_file, n_processes_json in zip(json_files, n_processes_per_json):
	tasks.append(
	execute_and_stream_output(
	(
	[palace, str(json_file)]
	if n_processes == 1
	else [palace, "-np", str(n_processes_json), str(json_file)]
	),
	shell=False,
	log_file_dir=json_file.parent,
	log_file_str=json_file.stem + "_palace",
	cwd=simulation_folder,
	)
	)
	await asyncio.gather(*tasks)

Describe the solution you'd like
Support distribution with Ray

enable CDL schematic conversion

At the moment gplugins support YAML schematic representation but recently some users asked for CDL support to use gdsfactory to generate cadence compatible schematics

How about suppor converting from YAML to CDL?

Add Bug report and Feature request templates from main repo

I think essentially the same Bug report and Feature request templates should be added from the main repo. Maybe the fixme folder can be left out though.

Issue on page /notebooks/sax_01_sax.html

https://gdsfactory.github.io/gplugins/notebooks/sax_01_sax.html#phase-shifter-model

@flaport
@simbilod

Run Palace tests in CI

Run Palace tests in CI. This requires installing Palace in GitHub Actions and adding palace to GITHUB_PATH. This currently requires compilation, hence taking a long time. Pre-compiling in let's say a Docker image or Singularity container seems like a good approach.

Path length analysis notebook has a missing figure

Describe the bug
Path length analysis notebook has a missing figure defined in

gplugins/docs/notebooks/path_length_analysis.py

Line 81 in 3fa1a1e

# 

To Reproduce

Expected behavior
No missing figures

Suggested fix
Add figure to repo or imgur

Expose built HTML from `test_docs` as artifact

Expose built HTML from test_docs GH action as artifact. I find this helpful for debugging formatting etc. when avoiding trying to install all the deps. like meep locally

Refactor capacitance path to use `gdsfactory.config`

Currently, Pdk.get_capacitance_path is used but using a gdsfactory-specific config as opposed to a PDK-attached one makes more sense. See gdsfactory/gdsfactory#1995 (comment) for details.

Unified geometry interface for simulation components

Is your feature request related to a problem? Please describe.
Currently, most device simulators derive their simulation properties (in particular geometries) from a patchwork of logic, e.g. from functionality contained in gdsfactory proper, from other gplugins, and self-implemented ones. This leads to a duplication of efforts and makes things harder to maintain.

Describe the solution you'd like
Ideally, there should be a single "source of truth" when it comes to simulated geometries. This could be implemented as a component base class that combines a layerstack and a component (and whatever else might be necessary that I'm not aware of currently) and provides a single interface from which all relevant attributes can be extracted easily. Developers could then implement whatever additional logic is necessary for a particular simulator on top of this interface.

I have implemented a draft of this base component in the tidy3d plugin rework (WIP) in #138 (see here). It's currently very incomplete and I've submitted it mostly because I am seeing a lot of issues crop up recently along similar lines (e.g. #119, #121, #122, #131) that could all be addressed with a common interface and I think it would be good to have a discussion.

Some considerations:

• Immutability: A simulation component should be immutable. It is derived from a particular layerstack and device, and if any of that information changes, it should be considered a new component.
• Owing to immutability, a component class should be hashable (implemented). This makes it easy to identify components for e.g. caching of simulation results for particular components or comparisons.
• The base component should have a fixed interface (allow_extra=False) to provide a strong guarantee of what is and isn't provided by this class. Children should modify their own behavior as needed.
• Ports: Since components have ports (usually), the component base should be able to provide information on port placement and sizing. I have started to implement this, and especially port sizing can (and probably should) be automated, i.e. I think one should be able to just ask the simulation component for the port sizes and it should provide them while handling spacing between ports and simulation boundaries automatically.
• Material properties: I am undecided whether this should be done in the base component or in the derived components that different simulators use. Probably the latter because material definitions differ greatly between different simulators.
• Visualization: Since all the geometry information is already contained in the component, there should be a way to visualize the component independent of different simulators. I have plotting implemented for the derived tidy3d component, but ideally this should live in the component base.

Long term this would enable us to add new simulators with relative ease and add new functionalities to existing ones, e.g. fabrication corner analysis is something that can simply extend the functionality of a base component (either via direct implementation or mix-ins) with little to no changes required to existing simulators (that already use the SimulationComponentBase)

If you agree that this is something that we should try to pursue, then it would be good to have a discussion on the actual interface that this component should provide. The one I've implemented is relatively bare-bones. It implements everything that's needed for tidy3d simulations, so there are very likely to be things that I've missed.

• tidy3d #239
• meep
• fdtdz
• lumerical @seanlam97

@joamatab @nikosavola @simbilod @HelgeGehring @mdecea

Export raw NumPy array capacitance matrix for ElectrostaticResults

Add helper function to get the raw capacitance matrix instead of the CDict format.

A hopefully working implementation is already done in

gplugins/gplugins/typings.py

Line 18 in d5aecd0

# TODO uncomment after move to pydantic v2

but requires Pydantic 2, which gdsfactory does not yet support until gdsfactory/gdsfactory#1879 is merged.

Allow arbitrary lumped port directions

Is your feature request related to a problem? Please describe.
awslabs/palace#75 is now merged and arbitrary rotation ports are supported. The implementation for Palace scattering in gplugins however supports only an orientation from ["+X", "+Y", "-X", "-Y"].

Describe the solution you'd like
Generalise inferring the lumped port direction and use new Palace format for arbitrary orientation

gplugins/gplugins/palace/get_scattering.py

Lines 409 to 428 in d342c82

	def _xy_plusminus_direction(point_1, point_2):
	# TODO update after https://github.com/awslabs/palace/pull/75 is merged
	delta_x = point_2[0] - point_1[0]
	delta_y = point_2[1] - point_1[1]
	angle = atan2(delta_y, delta_x)
	angle_deg = degrees(angle) + 360
	directions = ["+X", "+Y", "-X", "-Y"]
	index = round(angle_deg / 90) % 4 # TODO check if shift is correct
	return directions[index]
	lumped_two_port_directions = {
	ports[0]: _xy_plusminus_direction(
	*[component.get_ports_dict()[port].center for port in ports]
	)
	for ports in itertools.chain(
	lumped_two_port_pairs, [tuple(reversed(e)) for e in lumped_two_port_pairs]
	)
	}

Refactor Elmer examples to use `interdigital_capacitor_enclosed` from gdsfactory

Refactor Elmer examples to use interdigital_capacitor_enclosed merged to main in gdsfactory/gdsfactory#1994.
This includes

• docs
• tests

scalable way for testing and documenation

as we increase the number of plugins, the tests and the documentation build time is going to increase

how can we run in paralel?

improve dataprep

• make layers lazy loaded
• Use tiling processor for klayout, so big layouts are fast to process
• Add options (flatten=False). By default at the moment it flattens, which inceases size of the GDS
• Write more exmples

https://gdsfactory.github.io/gplugins/notebooks/klayout_dataprep.html

@sunjin-choi
@SkandanC
@tvt173
@sebastian-goeldi