reASITIC

A clean-room Python re-implementation of [ASITIC][asitic] — the planar RF inductor analysis tool originally developed at UC Berkeley by Ali M. Niknejad in 1999. The library exposes geometry builders, partial- inductance / Q / resistance / S-parameter / Pi-model analyses, and CIF / GDS / Sonnet / SPICE / FastHenry / Touchstone exports.

A clean-room Python re-implementation of the 1999 UC Berkeley ASITIC RF-spiral inductor analysis tool, faithful to the original but powered by NumPy / SciPy and a strict, fully type-annotated codebase.

🚀 Try reASITIC in your browser — no install   full Python + NumPy + SciPy + the reasitic wheel via Pyodide

User guide

• Installation
  • From PyPI
  • Development install
  • Runtime requirements
  • Verifying the install
  • Optional: legacy binary
• reASITIC tutorial
  • 1. Install
  • 2. Load a technology stack
  • 3. Build a spiral
  • 4. Analyse: L, R, Q, Pi-model
  • 5. Frequency sweep + Touchstone export
  • 6. Optimisation: maximise Q for a target L
  • 7. Multi-frequency design report
  • 8. Cross-validate against the legacy ASITIC binary
  • 9. The REPL
  • 10. Where to next?
• reASITIC cookbook
  • 1. Pick a 1-nH inductor for a 5 GHz LO
  • 2. 5 nH for 1 GHz with substrate-loss-limited Q
  • 3. Differential balun with 1:1 turns ratio
  • 4. Frequency-swept S-parameter export for SPICE
  • 5. Cross-validate against the legacy ASITIC binary
  • 6. Optimisation sweep across multiple operating points
  • 7. Differential transformer with optimal coupling
  • 8. MIM capacitor with proper dielectric stack
  • 9. Multi-metal series inductor (boosting L for fixed footprint)
  • 10. Parametric L-Q sweep visualisation
  • See also
• reASITIC FAQ
  • Q: The legacy run/asitic.linux.2.2 binary segfaults when I run Ind headlessly.
  • Q: Why do I need xvfb-run to drive the binary?
  • Q: I wrote a script that drives the binary via subprocess.run and my Python interpreter crashes.
  • Q: My spiral has a tiny inductance (< 0.1 nH) — is the result trustworthy?
  • Q: How accurate is the substrate shunt-cap model?
  • Q: My OPTSQ run never converges to the target L.
  • Q: I want the SPICE Pi-model at multiple frequencies. Is there a sweep?
  • Q: How do I cross-validate against measured data?
  • Q: Performance is slow on large spirals (>50 turns). What can I tune?
  • Q: Can I save my session in the binary’s BSAVE format?
  • Q: How do I add a new geometry builder?
  • Q: I get ImportError for matplotlib when I import reasitic.plot.
  • Q: How can I contribute?
• reASITIC CLI command reference
  • Command details
• Graphical interface
  • Layout
  • Mouse and keyboard
  • Console pane
  • Programmatic API
  • Testing without a display
• In-browser REPL
  • What you can do
  • CIF parity
  • How it works
  • Running the REPL locally

Reference

• API reference
  • Top-level package
  • Subpackages
• reASITIC ↔ ASITIC C function mapping
  • reasitic.units
  • reasitic.tech
  • reasitic.geometry
  • reasitic.inductance.grover
  • reasitic.inductance.partial
  • reasitic.inductance.filament
  • reasitic.resistance.dc
  • reasitic.resistance.skin
  • reasitic.quality
  • reasitic.network.twoport
  • reasitic.network.threeport
  • reasitic.network.touchstone
  • reasitic.network.sweep
  • reasitic.network.analysis
  • reasitic.persistence
  • reasitic.exports
  • reasitic.optimise
  • reasitic.inductance.eddy
  • reasitic.substrate.green
  • reasitic.substrate.fft_grid
  • reasitic.info
  • reasitic.substrate
  • reasitic.cli (REPL commands)
  • reasitic.validation.binary_runner
  • Subsumed by NumPy / SciPy / Python stdlib
  • Frontend (asitic_repl.c) bulk coverage
  • Vendored libraries (asitic_lapack.c / _linpack.c / _libf2c.c / _cxxrt.c)
  • GUI (X11 / Mesa → Tk)
• ASITIC Geometry Routine Notes
  • Per-Case Status Summary
  • Common Record Model
  • cmd_square_build_geometry (0x08056670)
  • cmd_spiral_build_geometry (0x08057248)
  • cmd_capacitor_build_geometry (0x0805bc3c)
  • cmd_ring_build_geometry (0x0805b450)
  • cmd_mmsquare_build_geometry (0x0805af5c)
  • cmd_trans_build_geometry (0x080576d4)
  • cmd_3dtrans_build_geometry (0x08057d40)
  • cmd_symsq_build_geometry (0x08059854)
  • cmd_sympoly_build_geometry (0x0805a45c)
  • Resume here (2026-05-09 round 2)
  • Remaining Work
• Benchmarks and cross-validation
  • Performance benchmarks
  • Cross-validation against the legacy binary
  • Continuous integration
  • Reproducing the benchmarks
• 100 % port milestone
  • Coverage by source file
  • What “100 % covered” means here
  • Feature surface
  • Quality gates
  • Cross-validation
  • Next steps
• Changelog
  • Unreleased

Quick install

pip install reASITIC          # base library
pip install reASITIC[plot]    # + matplotlib for plotting helpers

At-a-glance

import reasitic

tech = reasitic.parse_tech_file("BiCMOS.tek")
sp = reasitic.square_spiral(
    "L1", length=170, width=10, spacing=3, turns=2,
    tech=tech, metal="m3",
)
print(f"L = {reasitic.compute_self_inductance(sp):.3f} nH")
print(f"R = {reasitic.compute_ac_resistance(sp, tech, 2.4):.3f} Ω")
print(f"Q = {reasitic.metal_only_q(sp, tech, 2.4):.1f}")

Feature surface

• 117 / 117 original REPL commands

• 669 unit / integration / regression tests, 90 % line coverage

• 643 / 643 identified C functions covered (100 %) — see milestone

• mypy --strict clean across the entire public surface

• Greenhouse + Grover partial-inductance summation with filament-level current-crowding, Wheeler skin effect, substrate eddy correction

• Sommerfeld Green’s-function substrate model, FFT-accelerated coupling grid

• 2-port Y/Z/S algebra, π / π3 / π4 / πX extraction, 3-port reduction, Touchstone v1 reader/writer

• Geometry builders (10): square / polygon / wire / ring / via / transformer / 3D-transformer / symmetric-square / balun / capacitor

• Round-trip CIF / Sonnet / SPICE / FastHenry / Tek / Tek 4014 exports

• SLSQP-driven OptSq / OptPoly / OptArea / OptSymSq / BatchOpt

• JSON-based session persistence, full-binary cross-validation harness

Indices

• Index

• Module Index

• Search Page