The prototype setup is used to test new passive and active shielding methods. The experimental part of the project focuses on identifying, modeling, and mitigating magnetic noise sources:

• Environmental noise (power-line harmonics, building fields),
• Mechanical coupling (vibration-induced flux changes),
• Electronic and Johnson noise (shield material and sensor electronics).

A hybrid shielding system is under development, consisting of:

• a compact inner shielding pot made of μ-metal which contains the sensitive magnetometers, an outer shield as well as additional shielding geometries from μ-metal or Metglas forms the passive shielding
• a movable outer coil frame (Helmholtz and gradient coils) for active field compensation,
• and a low-noise feedback control system (here are analog PID and digital control loops under investigation).

The effectiveness is assessed through Low-Frequency Spectral Density (LSD) and coherence analysis between internal and external field sensors.

To further improve shielding efficiency, new compact geometries are being developed using:

• Finite Element Modelling (COMSOL) for field simulation,
• and Physics-Informed Neural Networks (PINNs) for fast, data-driven optimization of geometry and current distributions.

PINNs are trained on COMSOL-generated datasets to predict optimal combinations of shield thickness, opening geometry, and coil configuration, reducing the computation time by orders of magnitude.
This allows rapid exploration of design trade-offs between accessibility, weight, and magnetic performance — a critical step toward clinically deployable fMCG systems.