CoilGen2

Interactive design and field evaluation for MR gradient and shim coils.

Define a support surface and target field, optimize the surface current, and turn its contours into a winding. CoilGen2 brings these steps into a Python library, a desktop workbench and a browser workbench with the same numerical engine.

Installation · Desktop guide · Browser guide · Examples · Documentation

GUI gallery: winding layouts, field comparisons and fabrication checks.

The table below reports earlier ISMRM 2027 abstract designs, with fields evaluated on their mounted conductor centerlines. Their configurations and numbers are separate from the current GUI presets.

Example	ROI radius (mm)	Mode efficiency	Relative RMS Bz error (%)
A · Shielded y-gradient	20	1.759 mT/m/A	4.432
B · Biplanar x-gradient	20	1.636 mT/m/A	11.766
C · Cylindrical X²−Y² shim	20	12.635 mT/(m²·A)	4.572
D · Conical Z2 shim	50	2.442 mT/(m²·A)	10.577

Efficiencies are fitted Bz-mode coefficients of the mounted winding per ampere, using the same current in each part. The linear examples target 20 mT/m. Shim coefficients multiply x²−y² (C) and z²−(x²+y²)/2 (D), with coordinates in meters; their efficiencies therefore use mT/(m²·A). Figure sources and numerical results.

Installation

On Windows, run Setup CoilGen2.cmd from the checkout, then open Start CoilGen2 GUI.cmd. Choose Browse examples… → Gradients → Cylindrical Y gradient (Gy), then Compute full design. When the job finishes, use Jobs → View results → Run design choices to inspect the model that was submitted. Engineering checks shows the separate stream function, sweep and fabrication verdicts. For the browser, use Start CoilGen2 Browser.cmd and follow the first-run guide.

Requires Python 3.10+ and Git. In a terminal:

git clone https://github.com/Philipp-MR/Coilgen2.git
cd Coilgen2
python -m venv .venv

Activate the environment:

Platform	Command
Windows PowerShell	.\.venv\Scripts\Activate.ps1
Linux / macOS	source .venv/bin/activate

Install the workbench and scientific dependencies, then launch:

python -m pip install --upgrade pip
python -m pip install -e ".[gui,plot]"
python -m coilgen.gui

Choose File → Load template to start with a gradient or shim design. New gallery designs fit contours on the support with a 0.5 mm displacement cap and a turn-spacing guard. If a whole-loop fit fails, eligible corners can still be rounded within those limits. Compare Contours before routing with Review conductor centerlines after a run; Fit shows the whole coil. The connectors and mounted conductor have their own field, clearance and bend checks. On Windows, subsequent launches can also use Start CoilGen2 GUI.cmd. For scripts without the GUI, install python -m pip install -e ".[research,plot]". Full installation guide.

Design and inspect

• Optimize: generated or imported support meshes, analytic or imported target fields, suppression regions, regularization and constrained objectives.
• Build the winding: extract and smooth contours, connect loops, route returns, and set conductor cross-section and placement.
• Evaluate the fields from the stream function, contours and conductor centerlines. Inspect spacing, bend and clearance diagnostics, resistance and inductance estimates.

The workbench and Python scripts share the numerical engine. Saved projects retain editable geometry, targets and processing settings.

Conductor cross-sections and sweeping

Edit the cross-section in millimeters, sweep it along the winding, and choose centered, inner or outer mounting on the support. Profiles can be rectangular or custom convex polygons, including elliptical and trapezoidal profiles.

Illustrative circular-profile sweep meshes and bend details · Profile editor · Sweep and mounting controls.

Check the field after each geometry change

In the abstract's three controlled smoothing examples, 3D RMS curvature fell 39.8–61.3% at 0.5 mm arclength sampling. The smoothing-induced RMS Bz change stayed below 0.010% of target RMS, with the original signed currents retained.

Later operations mattered more: the biplanar example's relative RMS Bz error rose from 4.193% on the connected route to 11.766% after path realization and mounting. These comparisons use 2,048 independent volume samples per design, without gain or offset fitting. Methods, stage comparisons and provenance.

These are numerical examples. Discrete winding fields use thin-filament Biot–Savart evaluation; the abstract's profile sweeps failed configured geometric checks and are not ready for fabrication. Physical assumptions · Interpreting results.

Python workflow

from coilgen import CylinderCoilDesign

design = CylinderCoilDesign(
    target_bz_expression="y",
    target_scale=0.01,             # 10 mT/m
    radius_m=0.12,
    length_m=0.30,
    target_region_radius_m=0.045,
    contour_levels=16,
)
result = design.run()

Geometry uses meters and fields use tesla. This is a starter design; the abstract figures use separately saved configurations. See Getting started for plots, reports and exports.

Citation

Created and maintained by Philipp Amrein. CoilGen2 continues the original CoilGen work by Philipp Amrein, Feng Jia, Maxim Zaitsev and Sebastian Littin. Please cite the original paper and record the CoilGen2 version or commit:

Amrein P, Jia F, Zaitsev M, Littin S. CoilGen: Open-source MR coil layout generator. Magnetic Resonance in Medicine. 2022;88(3):1465–1479. doi:10.1002/mrm.29294.

Citation metadata · Contributing · Contact · Report an issue