10 kVA, 400 V / 230 V, 50 Hz single phase. Laminated E-I-style core model, 160 primary turns.
Flux. Faraday's law for a sinusoidal supply gives the peak flux density B = V1 / (4.44 f N1 A). Raising voltage or lowering frequency raises B; above the knee the steel saturates and the magnetising current spikes.
Core loss. Hysteresis (Steinmetz): Ph = kh f Bn Vcore. Classical eddy loss of a sheet of thickness t: Pe = (π2 σ f2 B2 t2 / 6) Vcore, so halving the sheet thickness cuts eddy loss by 4. For thick sheets the field cannot penetrate: the model multiplies by the exact slab factor 3/x (sinh x - sin x)/(cosh x - cos x), x = t/δ, with skin depth δ = 1/√(π f μ σ).
Equivalent circuit. Series R1+R2' = 0.32 Ω, X1+X2' = 0.80 Ω (2 % and 5 % of the 16 Ω base) with the core branch (loss current in phase, magnetising current from the B-H curve) shunted at the primary terminals. Voltage regulation ≈ Rpu cosφ ± Xpu sinφ. Efficiency peaks when copper loss equals core loss, at x = √(P0/Psc) of rated load.
Tests. Open circuit at rated voltage gives core loss P0, Rc and Xm; short circuit at rated current gives Psc, Req and Xeq. Efficiency predicted from test data alone is compared with the full circuit solution above.
Simplifications, stated honestly. Single phase, linear leakage, no excess (anomalous) loss, no hysteresis loop in the magnetising current (anhysteretic H(B)), fixed relative permeability for the skin depth, magnetising-current fundamental used in the phasor solution (the distortion is shown as THD), lumped thermal model with illustrative constants. Material constants are typical order-of-magnitude values for each class, not a datasheet; the 10 kVA unit is a made-up round-number design. The 3D picture shows 36 plates for clarity; a real core has hundreds.
Independent educational model. Flux sparks are a visual aid at 1/100 speed.