Vector fitting: Passivity assessment not always working correctly

[Vinc0110]

The passivity violations of vector fitted models returned by VectorFitting.passivity_test() are sometimes incomplete. Consequently, passivity enforcement fails for these models, as it's missing some frequency bands of passivity violation.

Example with the Coax 2-port from this post: #524 (comment)

import skrf
import numpy as np
import matplotlib.pyplot as mplt

nw = skrf.Network('/home/vinc/Downloads/Coax_test_v6_iter-0.s2p')
i_re = 1
i_cc = 5
freqs = np.linspace(0, 10e9, 1001)

vf = skrf.VectorFitting(nw)
vf.vector_fit(i_re, i_cc)

print('VF passive? {}'.format(vf.is_passive()))
violations = vf.passivity_test()
print('violations:\n{}'.format(violations))

fig, ax = mplt.subplots(1, 1)
vf.plot_s_singular(freqs=freqs, ax=ax)
y_min, y_max = ax.get_ylim()

for violation in violations:
    ax.plot([violation[0], violation[0]], [y_min, y_max], color='r')
    ax.plot([violation[1], violation[1]], [y_min, y_max], color='r')

ax.grid(True)
fig.tight_layout()
mplt.show()

Output:

VF passive? False
violations:
[[0.00000000e+00 8.65201086e+07]
 [3.30821054e+09 4.01429811e+09]]

It clearly misses the violation of $\sigma_1$ starting from about 5 GHz.

[Vinc0110]

The passivity assessment was filtering unbounded violation bands. This is now also fixed in PR #678. The enforcement routine had to be adapted accordingly, so I did not create a new PR for this issue (they are related).

New output for unbounded violations:

VF passive? False
violations:
[[0.00000000e+00 8.65201086e+07]
 [3.30821054e+09 4.01429811e+09]
 [4.93075857e+09            inf]]