When the circuit inductance is very low (1.4mH in this example). If identification was made with PowerSpy to calculate the differential inductance Ld = (Vmag- rmag x Imag)/(di_dt), with a very low inductance the accuracy of the calculation for Ld is a function of how accurate the measurement of rmag was made in DC (at current higher than 10-50A) and the di/dt used during your ramp: di/dt shall be high in order to have a significant inductive voltage drop across the magnet. Please recheck both points.

• There’s a coupling which is significant between the load and the output filter: SIRIUS_S is [600uH + 2x70uF + 2x80mOhm] in // with [0.3125 Ohm – 1.4mH]

In blue, the bode plot of the filter with a circuit composed of 0.3125 Ohm and 10mH: decoupling is perfect, Fco = 500Hz, and Vloop tuning is independent of the load,

In red, the bode plot of the filter with a circuit composed of 0.3125 Ohm and 1.4mH: there’s a coupling, Fco has moved to almost 600Hz and the curve in the low frequency domain is below the 0dB axis è *tuning of the Vloop has to be made with the load.*

1. Generate a small signal step response in Vmode with a table from the FGC3 (narrow pulse, short ON time, long OFF time): make sure that the minimum current in the load is always greater or equal than 15A during the OFF time,
2. Retune Ki, Ku and then kTs with the objective to get a voltage response on the step with kxi = 0.7 (overshoot of 5%) and a rise time 10-90% corresponding to a small signal BW of 1 kHz (probably circa 350us). If you use Powerspy for Vloop tunig, your choice but don’t forget that the control is made in DSP board @13.5kHz è what you will see in powerspy does not corresponds to real life, a scope is always better. For first iteration why not Powerspy but take care.
3. Once the tuning is done for Vloop, you can replay with tuning of Iloop and the internal script of FGC3.