Preliminary characterisation of the terminal behaviour of persistent-current switches in superconducting MRI magnets from operational service logs

Technical note v0.1 — preliminary, not peer reviewed. A full journal article is in preparation. When modelling the ramping of an MRI magnet, it is usually convenient to treat the heated persistent current switch (PCS) as an ideal open circuit, so that the power-supply current is equated with the winding current and the magnet dynamics is governed mainly by the winding inductance. In an installed magnet the switch sits inside the cryostat and is inaccessible to direct measurement. Its electrical behaviour, however, is reflected in the current and voltage recorded by the standard service power supply. While building a closed-loop emulator from MPS2 service logs we found that the ideal-open-circuit model fails to reproduce fast changes of the supply current that are synchronous with voltage-setpoint steps: the current of an inductive winding cannot exhibit such step changes. The operational data show that the current difference (I_psu − I_coil) flows through a finite-resistance parallel branch of the switch node. In some regimes, therefore, the supply current is not equal to the winding current, I_psu = I_coil + I_sw, and the PCS branch becomes a significant part of the system being controlled. We show that the effective electrical characteristic of this branch can be identified from routine service logs, incorporated into a closed-loop ramping model, and tested by transfer between magnets. To this end we propose a grey-box procedure that reconstructs the node voltage, separates the inductive component of the current, and determines a regime-dependent law R_sw(V_node; q). On a Siemens Verio 3 T magnet, four separate calibration windows identify a branch with a resistance of about 2 Ω below a +0.5 V threshold and a weakly resolved higher-resistance state. Cross-magnet evaluation comprised one successful frozen-parameter transfer and one active counterexample. Transfer of the frozen parameters to a magnet of a different model, a Siemens Avanto 1.5 T, reduced the RMSE of the closed-loop current reproduction by 45% (0.970 → 0.532 A), while a second Avanto unit does not reproduce the two-level form: forcing the same law degrades the agreement with the data. Thus the effective behaviour of the PCS branch can be quantitatively recovered from operational logs, but the identified law is magnet- and regime-dependent and is not universal even within a single vendor. Current sharing between a resistive PCS and the winding, as well as switch self-heating by the ramp voltage, are known from prior work. The contribution of this work is the quantitative identification of this branch from operational logs of installed MRI magnets and the demonstration of its dynamic significance in a closed-loop model of the system being controlled. A practical consequence is that models intended for developing and validating ramping algorithms must, in the relevant regimes, distinguish the supply current from the winding current. At the same time, the available logs determine only the effective external characteristic of the branch: the internal temperature of the switch zone, the physical nature of the observed states, and the complete electrothermal mechanism cannot be uniquely recovered from these data.