B.3 Application to Radial Waveguides
B.3.3 Adaptations for Modes with Azimuthal Variations
For n > 0, the procedure is the same as for Eρ and Hφ. For the other components one
It can be observed that, except for a different sign, the ratio ofEφtoHρ is the same as the ratio Eρ and Hφ. Therefore, the transfer matrix for these field components can be derived in an analogous way and can be written as
−Eφ(ze,i) concluded that the matrices express the same conditions. The difference in sign of the electric field in (B.47) leads (by construction of the wave solution) to an orientation of the transverse component of the wave vector which is the same as in (B.42).
For mode order index n = 0 and with PEC boundary conditions at the outermost layer interfaces it is thereby shown that the same transfer matrix can be used for the determi-nation of the wave numbers of the radial waveguides. The following cases can be shown in an analogous way:
• Cases with n >0 and with the complementary sinusoidal azimuthal dependence.
• Non-PEC boundary conditions with decaying fields away from the waveguide analog to the method in Section 4.3.3.
B.4 Analytical Derivatives for the TRM B.3.4 Approximate Solution for a PEC-bounded Single Layer
For the single layer it follows that
kρ=ki =ω√
εiµi (B.48)
If a loss tangent tanδ of the dielectric medium is defined by εi =ε0εr,i(1−jtanδ), then for small loss tangents, using the series expansion of the square root, the wave number can be given as
This is the same result for the attenuation due to dielectric loss (in an unlayered medium) as the one derived through considerations of power in [91, Eq. A2.8].
B.4 Analytical Derivatives for the TRM
The numerical efficiency and robustness can be improved by making use of analytical derivatives of the function of which the roots need to be determined when using one of the root-search techniques that involve derivatives of the characteristic function. Consider the following form of the transfer matrix for layer i
Ti = cos(θi) −jεr,isin(θi)/ki θi =kiti. Note that the sign chosen for the ki of the intermediate layers is irrelevant. This is because cos(kiti), sin(kiti)/ki, and kisin(kiti) are all even functions for their complex argument ki (see also Appendix F.2) and therefore invariant to its sign.
The transfer function for two consecutive layer i and j is given by the matrix product
T{i,j} =TiTj (B.51)
and for all N layers analogously by T =T{1,2,···,N} =
where the product is defined recursively by
The characteristic functionf(kz)for the general case (arbitrary non-magnetic materials at outermost layers) is then given by
f(kz) =−jksT1,1/εr,s−jkcT2,2/εr,c−T2,1 +kskcT1,2/(εr,sεr,c), (B.54) where for the space below the first layer (“substrate”) ks = ±q
k2z−k0,s2 and εr,s = εs/ε0. For the space above the uppermost layer (“cladding”) kc = ±q
k2z−k20,c and εr,c = εc/ε0. In both cases the sign for the wavenumber has to be chosen such that the real part is positive which corresponds to a wave that decays in the transverse direction(s) away from the waveguide.
B.4.1 First Derivative of the Transfer Matrix
The first derivative of the transfer matrix of layeri with respect to the longitudinal wave number kx can be given as
The first derivative of the transfer matrix for two consecutive layer iand j is given by the sum of matrix products involving the transfer matrices and their derivatives:
∂T{i,j}
∂kx = ∂(TiTj)
∂kx = ∂Ti
∂kxTj+Ti∂Tj
∂kx (B.56)
The first derivative of the overall transfer matrix for all N layers can be formulated in a recursive way as
B.5 Characteristic Impedance and Wave Impedances of Parallel Plate Waveguides B.4.2 First Derivative of the Characteristic Function
The derivative of the characteristic function is given by differentiation of (4.26) as
∂F(kz)
B.5 Characteristic Impedance and Wave Impedances of Parallel Plate Waveguides
B.5.1 Impedances of the Homogeneously Filled Waveguide
Consider a parallel plate waveguide with perfectly conducting plates, width w and plate separation d. The homogeneous filling has a permittivity ε and a permeability µ0. The characteristic impedance can be expressed as a function of the per-unit-length capacitance and inductance which are easily derived as
L0 =εw/d, (B.60)
C0 =µ0d/w. (B.61)
The characteristic impedance of the (fundamental) TEM mode is then given as Z0 =
Equation (B.62) can applied, e.g., when using the characteristic impedance Z0 from port solutions of FEM full-wave simulations or the network parameter of a transmission line segment to determine the wave impedance η of the material.
B.5.2 Impedances of the Layered Waveguide
For the ith layer of the multilayer waveguide, a TM-wave exists for which the longitudinal impedance can be related to the longitudinal wave number as
ηTMx = kx
ωεi. (B.63)
I.e., the wave impedance is related to the permittivity of the respective layer, but the longitudinal wave number is the same for all layers and is determined with one of the methods discussed before.
Next, the characteristic impedance of the layered waveguide is considered for the case where it can be related to capacitances and inductances. For the silicon structures that are of interest in this thesis this has been found to be applicable for the slow-wave and quasi-TEM regimes where the (effective) wave number has been with good accuracy approximated as
kx =ω√
Using L0 which is identical to the empty guide and C0 for the characteristic impedances, one obtains
In the last step of (B.66) an effective wave impedance ηeff is defined that is analog to the one of the homogeneously filled guide and is related to the characteristic impedance by only geometrical parameters. It is different in that it cannot in general relate the transverse electric and magnetic field components of the mode. For the effective wave impedance, also (B.63) can be used together with(B.64) as
ηeff= kx
From the last result in (B.67) it can be seen that with these assumptions, the TM- and TE-mode impedances coincide.
Note that these results are equivalent to those used in other publications. For the modeling
B.5 Characteristic Impedance and Wave Impedances of Parallel Plate Waveguides of the multi-conductor lines in [25], a characteristic impedance
Zc= 1
√εeffc0C0 (B.68)
is used where c0 is the speed of light, C0 is the capacitance of the structure with all dielectrics replaced by vacuum, and
εeff = kzc0
ω 2
. (B.69)
This formulation can be traced back to [131–133]. It should be noted that, while [132]
relates the characteristic impedance to the (real) phase velocity, it is related to the com-plex propagation constant by [131]. Applicability of the latter is assumed with the above formulas.