I use this model:
* Saturable Core Model, copied from:
* _SPICE Models For Power Electronics_, Meares and Hymowitz.
*
.SUBCKT INDSAT 1 2 PARAMS: VSEC=1e-4 LMAG=1e-5 LSAT=1e-7 FEDDY=1e6
F1 1 2 VM1 1.0
G2 2 3 1 2 1.0
E1 4 2 3 2 1.0
VM1 4 5 0.0
RX 3 2 1E9
CB 3 2 {VSEC/500} IC=0
RB 5 2 {LMAG*500/VSEC}
RS 5 6 {LSAT*500/VSEC}
VP 7 2 250
VN 2 8 250
D1 6 7 DCLAMP
D2 8 6 DCLAMP
.MODEL DCLAMP D ( CJO={3*VSEC/(6.28*FEDDY*500*LMAG)} VJ=25 RS={LSAT/VSEC} )
.ENDS
You'll need to paste this into a subcircuit file, and load that file as a model or library. RTFM
It works by transforming terminal voltage into current in a capacitor; as the capacitor charges, flux increases. A diode clamps the flux, providing an exponential cutoff: flux cannot practically increase beyond this point. A second capacitor (the diode capacitance) even provides a lowpass filter characteristic, which has the effect of reducing flux at high frequencies: the permeability of the core drops. (The core also becomes lossy, i.e., the permeability has an imaginary component. Inductive reactance is already imaginary, so imaginary * imaginary = resistive loss!
) A voltage source measures the current flow into the clamp diode section, relaying this back to the terminals; thus terminal current is controlled by flux, and attempting to push excessive flux into the core will draw a huge current (just as should be the case!).
Which explains what the variables control:
VSEC: saturation flux in volt-seconds
LMAG: magnetizing inductance in henry (default, un-saturated inductance)
LSAT: saturated inductance in henry (this should be mu_r times smaller than LMAG)
FEDDY: cutoff frequency where permeability drops (as if due to eddy currents in a conductive core).
This is equivalent to a ferrite core with cross-sectional area A_e, turns N, permeability mu_r, inductivity A_L, and:
VSEC = A_e * N
LMAG = A_L * N^2
LSAT = A_L * N^2 / mu_r
A typical #77 ferrite might be mu_r = 2000 and FEDDY = 1e6 or so. High mu (up to 10-15k) ferrites have lower FEDDY (down to about 0.3e6), and lower loss, lower permeability types have higher FEDDY. Laminated iron will have FEDDY inversely proportional with lamination thickness, with several kHz being typical. (More complicated core loss curves might simply be modeled with an RLC network in parallel with the core, rather than at the flux level. For instance, the diffusion characteristic that nanocrystalline steel has.)
Tim