The task was to develop a design method that would make sure that small ferrite filters would prevent the transfer of dangerous amounts of energy into pyrotechnic devices. In principle the idea is simple, a test on a single ferrite bead gives an indication of its inductance and with the resistance of the circuit, an estimate of attenuation or filtering effect is made.
However, the configuration allowed only a small amount of space for the ferrite material, the form was a coaxial line with non-standard conductor materials and the terminations were themselves other coaxial arrangements. To make it more interesting, the manufacturer's data for the properties of the material provided permeability values over only a small part of the frequency range specified. We were working to MILSTD 1576 or variants of it for commercial applications and needed to go well above the gyromagnetic resonance frequency of the material.
I developed a model of the device treating it as a small short, transmission line with the structure into which it led being similarly treated. As a result we had a model of a weird communications line. However for testing we could not actually look at the receiving end as that was the location of the pyrotechnic material and covered. So we used a vector signal analyser and measured the input impedance of the device over the range of frequencies we were interested in (only dummy pyro materials in place!) and with this plot, after a lot of jiggling about we were able to get a small signal model of the properies of the ferrite that were consistent with its initial permeability (known), the drop off above the gyromagnetic resonance frequency and some estimates of the loss component (complex permeability) beyond the frequency covered by the manufacturer's data.
From the resulting model we could calculate small signal attenuation to be sure the amount of signal passing to the fuse wire (bridgewire) was low enough to be safe. However, that was all small signal stuff. The MILSTD 1576 specification (or variant thereof) called for significant input power to be put in by conjugate matching. As the power was applied, the ferrite bead absorbed it, heated up and changed its characterisitics. That never improved the matching so the power input was never thereby increased. Tests on live devices later showed the designs were safe. The live testing was another interesting experience especially as we did some deliberate tests (and two unintentional ones!) to find out what safety margin we really had.