Bitter solenoid simulator — interactive demo

Move the sliders to change the coil. The current is set automatically to reach the target field B0, and the page recomputes current, voltage, power, water flow, temperatures and the field map. Back to the overview.

Starting points

Results
Plate stack, to scale

Copper plates, insulator gaps and rows of cooling holes, drawn to scale. The moving marker shows water flowing from inlet to outlet.

Bz in the r–z half-plane

Field strength in a slice through the coil axis (coarse map for the picture). The box marks the central sphere (DSV) where uniformity is measured.

Advanced (work in progress)

Experimental panels kept for development: material catalog, mesoscopic emulation (not molecular dynamics), constraint chips, Swiss-roll RF, the fMRI layer (shims, B0 stability, RF SNR) and the reduced-N particle panel. Derivations and the full report live in the repository: PROJECT_REPORT.md, LaTeX report, and the README on GitHub. Model grades: continuum identities are established; contact scatter, ONB and the Langevin cloud are approximate; the legacy Swiss-roll SNR heuristic is speculative; the fMRI panel uses a quasi-static RF model.

Configuration

Catalog allow-list only. Typed values clamp to the catalog bounds and say so.

Emulation

Constraint chips
Emulation quantiles

Bars run from p05 to p95. The tick is the median. The dot is the continuum value. If the median voltage sits above the continuum voltage, that offset is the sampled contact resistance.

Langevin cloud in one RVE

A few dozen seeded carriers, not 10^28 particles. At room temperature the thermal speed hides the drift.

Pump-failure sketch

Side by side
Swiss roll μ_eff near Larmor (RF only; μ_eff(0) = 1)

μ_eff curve. SNR now comes from the quasi-static RF model in the fMRI panel; the E28 heuristic is kept only for reference.

fMRI layer: shims, B0 stability, RF SNR

Same equations as harmonics.py, stability.py and rfsnr.py (parity-tested). Every input default is an assumption; replace with vendor data.

Particle emulation (reduced N): P1 current elements, P2 conduction carriers

A browser port of bittersim.particles. P1 places Biot–Savart point current elements with the same deterministic Halton estimator as Python, so at equal N it matches Python to rounding (parity-tested). P2 runs Ornstein–Uhlenbeck carriers with a Bloch–Grüneisen τ(T) and uses Green–Kubo to get ρ. Its random stream differs from Python's, so P2 parity is statistical. This is not molecular dynamics. The full-N comparison tables are in the README.