eda2eagle

Status: Experimental · Ambitious · Work-in-progress

Universal netlist-to-Eagle-CAD-XML converter, and the first stepping stone toward a much larger vision: turning a vector PDF of a hand-drawn or printed schematic into a fully editable, netlist-aware EDA project.

The vision

This project is not just a file-format translator. It is the foundation layer of a longer roadmap:

1. Phase 1 — current: parse structured netlists (KiCad, SPICE, PADS ASCII, CadStar) and emit valid Eagle 9.6 XML, with auto-generated symbols, auto-placement, and net sanitization.
2. Phase 2 — next: ingest vector PDFs of schematics (exported from unknown EDAs, datasheets, old EDA tools, or photographed drawings vectorized via Inkscape/Potrace), extract wires, junctions, component symbols, and pin labels using geometric + topological heuristics.
3. Phase 3 — goal: reconstruct a full, editable, netlist-driven .sch from what was originally "paper" — so that legacy schematics, whiteboard designs, and scanned archives become first-class EDA projects again.

The output format is deliberately Eagle XML because it is the de-facto interchange lingua franca: the same .sch file is natively importable by:

EDA	Import path
KiCad	File → Import → Non-KiCad Project → Eagle
Altium Designer	File → Import → Eagle
OrCAD / Allegro	via Eagle importer or ULPP bridge
EasyEDA / LCEDA	Import Eagle .sch / .brd
Autodesk Fusion 360 Electronics	native Eagle engine
CircuitStudio	Eagle import
LibrePCB	via Eagle compatibility layer

In other words: one converter, every major EDA — free or commercial.

What it does today

• Format autodetection from file content (not extension): KiCad .net, SPICE .cir, PADS ASCII .asc, CadStar .txt.
• S-expression parser for KiCad, dedicated parsers for the other three formats.
• Automatic symbol generation: rectangular bodies with pins evenly distributed on two sides, sized to fit net labels without overlap.
• Auto-placement using a graph-based spiral algorithm (requires networkx); falls back to a 15-column grid otherwise.
• Net sanitization: names uppercased, illegal characters replaced, collisions disambiguated.
• Preview rendering with matplotlib so you can tune clearance before committing to the final file.
• Dual interface: CLI for scripting, GUI (tkinter) for double-click usage on Windows.

What it does NOT do (yet)

• It does not read PDFs, images, or raster scans.
• It does not perform OCR or symbol recognition.
• It does not reconstruct placement from visual cues — placement is purely topological today.

Those are exactly the problems Phase 2 and 3 are meant to solve.

Requirements

• Python 3.7+
• networkx (optional, enables smart placement)
• matplotlib (optional, enables visual preview)