InvenTree DC-bias curve plugin

Shows a capacitor's DC-bias (derating) curve on its InvenTree part page, in a DC bias panel below Notes. If the part has no curve, the panel says "No DC-bias curve available".

The curve comes from the part's Cbias Curve text parameter. It's read and interpolated exactly as GroundTruth reads it (components/bias.py), so the plot shows the same capacitance GroundTruth uses in a simulation.

The panel

The screenshots show a 220 nF 100 V X7R 0805 whose Cbias Curve is -2%@10V, -12.2%@25V, -48.8%@50V, -65.9%@75V, -74.1%@100V.

Tiles across the top, from left to right:

• Nominal: the capacitance at 0 V (the Capacitance parameter), with the dielectric and the rated voltage.
• At 50 V: half the rated voltage. GroundTruth uses this voltage for a single Cbias value given without one.
• At 100 V: the rated voltage.
• Curve data: how many points the curve has and the voltage range they cover.

The bar under a capacitance shows how much of the nominal is left: 112.6 nF at 50 V is about half the bar.

The chart

• The line is the capacitance against the DC voltage across the capacitor. Between the given points (the dots) it is a monotone cubic (PCHIP) through (0 V, nominal) and the points, so it never overshoots between them. Above the last point it continues dashed as C0 / (1 + (V/V1)²) through that point (extrapolated). It's dashed only when the chart goes past the last point, which isn't the case here.
• The shaded area between the line and the dashed nominal line is the capacitance lost to DC bias.
• "rated 100 V" marks the rated voltage. The chart runs up to it, or to 1.25 × the last point when the part has no rating.
• The ring and the dotted line (here at 50 V) mark the voltage chosen in "Capacitance at". Clicking the chart moves it there.
• Hovering shows the voltage, the capacitance and the change at the pointer.
• ΔC % / Capacitance at the top right switches the y-axis between the change from the nominal and the capacitance itself. The choice is remembered in the browser.

Reading off values

• Capacitance at: type a DC voltage to get the capacitance and its change from the nominal. It starts at half the rated voltage, and is marked "extrapolated" above the last point.
• Points: the curve's given points as capacitance and change, with the nominal at 0 V on top. Percent points are turned into capacitance against the nominal.
• Cbias Curve (below them, in the first screenshot): the parameter's text as written, with a copy button.

On a wide screen these cards sit to the right of the chart; in a narrow panel they move below it. The panel follows InvenTree's light and dark themes.

Which parameter is drawn

A bare number in the curve is read in µF, as in GroundTruth. Up to 12 points are allowed. A curve that can't be read shows what's wrong with each point instead of a chart. A point more than 1.2 × the nominal (a typo such as 750nF for 75nF) is drawn, with a warning above the tiles.

Install

1. In Admin Center, Plugins, Install Plugin:
   • Package Name: inventree-cbias-plugin, with the Source URL left empty: the latest release from PyPI. (Or pip install inventree-cbias-plugin in InvenTree's environment, or add it to plugins.txt.)
   • To install the newest code on main instead, also set the Source URL to git+https://github.com/afkiwers/inventree_cbias.git. Without git+, InvenTree passes the URL to pip as a package index (pip install -i ...), and pip finds nothing there. The Package Name must still be inventree-cbias-plugin; inventree_cbias fails.
2. Activate CbiasPlugin: the ⋯ menu on its row, then Activate.
3. Turn on Enable interface integration (Settings, Plugin settings). Panels need it.
4. If the plugin's settings say "Error loading settings", or the panel says "Failed to load module .../static/plugins/cbias/cbias_panel.js" (a 404), the server's processes haven't all loaded the new plugin yet. Activation asks the background worker to copy cbias_panel.js into the static files, and a worker that hasn't loaded the plugin skips the copy without an error. Deactivate and activate the plugin again, or run invoke static in the server container. Do the same after every upgrade of the plugin.

The panel appears on a part when its category path contains capacitor (case doesn't matter), or when the part has a Capacitance, Cbias or Cbias Curve parameter. The parameter names and the category keyword can be changed in the plugin's settings.

In use on an InvenTree 1.1+ server (Python 3.14); written against InvenTree's 0.18 plugin interface. On 1.1+ the panel reads /api/parameter/ (parameters for every model), and on older servers it falls back to /api/part/parameter/. It reads through the logged-in user's own API session, so InvenTree's permissions apply. Parameter text is inserted only as text, never as HTML.

Development

The interpolation is a port of GroundTruth's Python. tests/reference.json holds GroundTruth's own answers, and the JavaScript is checked against them, bit for bit at the time of writing:

node --test tests/test_curve.mjs
# after a change to GroundTruth's components/bias.py or components/values.py, regenerate the reference:
<GroundTruth>/.venv/Scripts/python tests/make_reference.py tests/reference.json

GitHub Actions run on every push and pull request (.github/workflows/checks.yaml): flake8, these curve tests, and a test build of the package that checks cbias_panel.js is in the wheel.

Releasing

Publishing a GitHub release uploads the package to PyPI (.github/workflows/pypi.yaml):

1. Set CBIAS_PLUGIN_VERSION in inventree_cbias/version.py (e.g. 0.3.0), merge it to main.
2. On GitHub, Releases, Draft a new release, with a new tag of the same version (0.3.0 or v0.3.0), then Generate release notes and Publish release. The workflow stops if the tag and version.py differ.

The upload uses PyPI's Trusted Publishing, so no API token is stored in the repository. Set it up once, before the first release: on pypi.org, Your account, Publishing, Add a new pending publisher (GitHub): PyPI project name inventree-cbias-plugin, owner afkiwers, repository inventree_cbias, workflow pypi.yaml, environment pypi. The first release then creates the project on PyPI.