def convert(self, dest, Tout=False):
N = dest(self.N)
_, T_s2d, _ = convert(self.N.P, N.P, self.N)
C = matrix_multiply(matrix_multiply(T_s2d, self.C), Hconj(T_s2d))
C[..., 0, 0] = C[..., 0, 0].real
C[..., 1, 1] = C[..., 1, 1].real
if Tout:
return NoisyTwoport(N, C), T_s2d
else:
return NoisyTwoport(N, C)
def convert(fromP, toP, a):
"""Converts multiports using permutation matrices as described in [#]
[#] J. Stenarson and K. Yhland "" IEEE Transactions on Instrumentation and
Measurements 2009 vol. x no. 4 pp. xx-yy.
"""
PB = inv(toP)
PA = fromP
P = matrix_multiply(PA, PB)
tau1 = P[..., :2, :2]
sigma1 = P[..., :2, 2:]
tau2 = P[..., 2:, :2]
sigma2 = P[..., 2:, 2:]
A = inv(tau1 - matrix_multiply(a, tau2))
B = -sigma1 + matrix_multiply(a, sigma2)
res = matrix_multiply(A, B)
return res, A, B
def convert(fromP, toP, a):
"""Converts multiports using permutation matrices as described in [#]
[#] J. Stenarson and K. Yhland "" IEEE Transactions on Instrumentation and
Measurements 2009 vol. x no. 4 pp. xx-yy.
"""
PB = inv(toP)
PA = fromP
P = matrix_multiply(PA, PB)
tau1 = P[..., :2, :2]
sigma1 = P[..., :2, 2:]
tau2 = P[..., 2:, :2]
sigma2 = P[..., 2:, 2:]
A = inv(tau1 - matrix_multiply(a, tau2))
B = -sigma1 + matrix_multiply(a, sigma2)
res = matrix_multiply(A, B)
return res, A, B
def passive_noise(twoport, Tamb=290):
if isinstance(twoport, YArray):
C = 2 * k * Tamb * (twoport + Hconj(twoport))
elif isinstance(twoport, ZArray):
C = 2 * k * Tamb * (twoport + Hconj(twoport))
elif isinstance(twoport, SArray):
eye = hftools.dataset.make_matrix(np.array([[1, 0], [0j, 1]]),
dims=tuple())
C = k * Tamb * (eye - matrix_multiply(twoport, Hconj(twoport)))
else:
raise Exception("Can only convert passive Y, Z, and S to NoisyTwoport")
return NoisyTwoport(twoport, C)
def make_passive_svd(S, delta=1e-6):
"""Force S-parameters in S to be passive.
The S-parameters of S are scaled such that the highest eigenvalue of S'
becomes |lambda|max < 1 - delta. The scaling is done using SVD.
Doshi, et.al, DesignCon 2012, "Fast and Optimal Algorithms for Enforcing
Reciprocity, Passivity and Causality in S-parameters"
"""
svd = np.linalg.svd
out = []
for ss in S:
ss = np.array(ss, copy=False)
u, s, v = svd(ss)
s = abs(s) - 1
s[s < 0] = 0
s = np.diag(s)
delta = matrix_multiply(matrix_multiply(u, s), v)
out.append(ss - delta)
return S.__class__(out, dims=S.dims)
def make_passive_svd(S, delta=1e-6):
"""Force S-parameters in S to be passive.
The S-parameters of S are scaled such that the highest eigenvalue of S'
becomes |lambda|max < 1 - delta. The scaling is done using SVD.
Doshi, et.al, DesignCon 2012, "Fast and Optimal Algorithms for Enforcing
Reciprocity, Passivity and Causality in S-parameters"
"""
svd = np.linalg.svd
out = []
for ss in S:
ss = np.array(ss, copy=False)
u, s, v = svd(ss)
s = abs(s) - 1
s[s < 0] = 0
s = np.diag(s)
delta = matrix_multiply(matrix_multiply(u, s), v)
out.append(ss - delta)
return S.__class__(out, dims=S.dims)
def switch_correct(b, a):
Sm = matrix_multiply(b, inv(a))
return Sm
def switch_correct(b, a):
Sm = matrix_multiply(b, inv(a))
return Sm
def test_3(self):
res = hfmath.matrix_multiply(self.m2, self.m)
self.assertAllclose(res, np.dot([[1., 2], [3, 4]], [[1., 2], [3, 4]]))
self.assertEqual(res.shape, (3, 2, 2))
self.assertEqual(res.dims, (self.J, self.mi, self.mj))
