---
title: Two-Port Impedance Model
url: https://www.emergentmind.com/topics/two-port-impedance-model
type: topic
---

# Two-Port Impedance Model

Searching arXiv for recent and relevant papers on two-port impedance modeling and closely related formulations.
A two-port impedance model represents a device, subsystem, or interaction by relating two port voltages and two port currents, conventionally as
\[
\begin{bmatrix}V_1\\V_2\end{bmatrix}
=
\begin{bmatrix}Z_{11}&Z_{12}\\Z_{21}&Z_{22}\end{bmatrix}
\begin{bmatrix}I_1\\I_2\end{bmatrix}.
\]
In this form, \(Z_{11}\) and \(Z_{22}\) are driving-point impedances, while \(Z_{12}\) and \(Z_{21}\) are transfer or mutual impedances. In recent literature, however, the topic spans more than fixed frequency-domain \(Z\)-matrices: it includes hybrid immittance descriptions, measured operating-point-specific models, modal equivalent circuits, and energetic port-Hamiltonian formulations when a stationary impedance matrix is not mathematically faithful [2010.00919][2509.24683][2210.14057].

## 1. Canonical meaning of a two-port impedance description

The standard reciprocal two-port is written either in impedance form,
\[
[Z]=\begin{bmatrix} z_{11} & z_{12} \\ z_{12} & z_{22} \end{bmatrix},
\]
or in admittance form,
\[
[Y]=\begin{bmatrix} y_{11} & y_{12} \\ y_{12} & y_{22} \end{bmatrix}.
\]
Reciprocity is expressed by \(Z_{12}=Z_{21}\) or \(Y_{12}=Y_{21}\), while non-symmetry means \(Z_{11}\neq Z_{22}\) or \(Y_{11}\neq Y_{22}\). This distinction is central in microwave and antenna structures, especially reciprocal but non-symmetric networks that require three complex degrees of freedom for complete characterization [2010.00919].

In wave-oriented formulations, the same two-port may be represented by scattering parameters. For measured microwave cavities and related systems, the impedance–scattering relation is
\[
Z = Z_{0}^{1/2}(1+S)(1-S)^{-1}Z_{0}^{1/2},
\]
with \(Z_0\) the diagonal matrix of channel characteristic impedances; for equal real reference impedance \(Z_0\), the inverse relation can be written as
\[
S=(Z-Z_0I)(Z+Z_0I)^{-1}.
\]
Accordingly, a two-port impedance model is often one coordinate choice inside a larger immittance/scattering framework rather than an isolated formalism [1309.3302][2412.17884].

## 2. Modal and equivalent-circuit parameterizations

A general equivalent circuit for lossy reciprocal non-symmetric two-ports can be derived by eigenstate decomposition. In that construction, the complete model is specified by three complex parameters,
\[
\lambda_1,\;\lambda_2,\;p,
\]
where \(\lambda_1,\lambda_2\) are modal immittances and \(p\) encodes eigenvector shape, asymmetry, and modal mixing. The corresponding admittance decomposition is
\[
[Y]=[Y^a]+[Y^b],
\]
with each term a rank-one modal contribution. When \(p=1\), the formulation reduces to the usual even/odd lattice interpretation; when \(p\) is real, the eigenstates are orthogonal; when \(p\) is complex, the modes are non-orthogonal. A key property is that for passive reciprocal networks the real parts of the modal immittances are non-negative, avoiding negative-real-part branch artifacts that can arise in simpler \(T\) or \(\Pi\) fits [2010.00919].

An allied synthesis viewpoint treats the passive linear subsystem as a black-box multiport impedance \(Z(s)\), fits it as a positive-real rational matrix, and then realizes it exactly by a finite lumped circuit. In the two-port case,
\[
\begin{pmatrix}V_1\\V_2\end{pmatrix}
=
\begin{pmatrix} Z_{11}(s) & Z_{12}(s)\\ Z_{21}(s) & Z_{22}(s)\end{pmatrix}
\begin{pmatrix}I_1\\I_2\end{pmatrix},
\]
and the off-diagonal terms are realized by multiport transformer structure, especially Belevitch transformers. The positive-real property is decisive: a finite passive lumped circuit exists exactly realizing \(Z(s)\) only if \(Z(s)\) is positive real [1505.04116].

## 3. Energetic completion beyond ordinary \(Z\)-parameters

A notable limitation of the conventional two-port impedance picture appears in explicitly time-varying devices. For a time-varying capacitor, the classical constitutive law
\[
Q(t)=C(t)V(t),\qquad I(t)=\frac{d}{dt}\big(C(t)V(t)\big)=C(t)\dot V(t)+\dot C(t)V(t)
\]
is mathematically valid, but incomplete as a physical one-port model. Using the storage function
\[
S(Q(t),t)=\frac{1}{2C(t)}Q^2(t),
\]
its derivative is
\[
\frac{d}{dt}S(Q(t),t)=V(t)I(t)-\frac{1}{2}\dot C(t)V^2(t).
\]
The extra term has no associated port in the one-port description, so the missing power must be exchanged with the mechanism that changes the capacitance [2210.14057].

The complete formulation promotes \(C\) to a state and introduces a second, mechanical-like modulation port:
\[
\dot Q(t)=I(t),\qquad \dot C(t)=U(t),
\]
\[
V(t)=\frac{Q(t)}{C(t)},\qquad F(t)=-\frac{1}{2C^2(t)}Q^2(t).
\]
The two ports are then \((I,V)\) and \((U,F)\), with power \(VI\) and \(FU\), and the balance law becomes
\[
\frac{d}{dt}S(Q(t),C(t))=V(t)I(t)+F(t)U(t).
\]
In this completed representation, the device is a nonlinear, time-invariant, lossless two-port state-space system.

This model is not a conventional LTI impedance matrix. The paper is explicit that no ordinary frequency-domain impedance matrix exists globally, because the relation depends multiplicatively on the time-varying state/input \(C\) or \(U\). The closest faithful description is port-Hamiltonian:
\[
\begin{cases}
\dot Q = I,\\
\dot C = U,\\
V = Q/C,\\
F = -Q^2/(2C^2).
\end{cases}
\]
Accordingly, a local incremental linearization around an operating trajectory may yield a linear time-varying small-signal two-port matrix, but not a fixed impedance matrix [2210.14057].

## 4. Measurement, extraction, and calibration of two-port models

A frequency-domain two-port impedance representation is used for a converse magnetoelectric magnetometer with two electrical interfaces: a resonator/excitation side and a pickup-coil side. Its measured model is
\[
\begin{bmatrix} u_{\mathrm{res}}(s)\\ u_{\mathrm{coil}}(s) \end{bmatrix}
=
\begin{bmatrix} Z_{11}(s) & Z_{12}(s)\\ Z_{21}(s) & Z_{22}(s) \end{bmatrix}
\begin{bmatrix} i_{\mathrm{res}}(s)\\ i_{\mathrm{coil}}(s) \end{bmatrix},
\]
where \(Z_{11}\) is the composite cantilever impedance, \(Z_{22}\) the coil impedance, \(Z_{21}\) the forward transfer impedance, and \(Z_{12}\) the reverse transfer impedance. The parameters are obtained by measuring S-parameters with a Bode 100 Vector Network Analyzer under fixed AC excitation amplitude of \(10\ \mathrm{mV}\), \(1\ \mathrm{Hz}\) resolution, averaging factor \(5\), and \(100\ \mathrm{Hz}\) measurement bandwidth, and then converting them to Z-parameters in MATLAB. Under open-circuit output conditions,
\[
H_{\mathrm{oc}}(s)=\frac{u_{\mathrm{coil}}(s)}{u_{\mathrm{res}}(s)}=\frac{Z_{21}(s)}{Z_{11}(s)},
\]
while under short-circuit conditions,
\[
H_{\mathrm{sc}}(s)=\frac{i_{\mathrm{coil}}(s)}{u_{\mathrm{res}}(s)}
=
-\frac{Z_{21}(s)}{Z_{11}(s)Z_{22}(s)-Z_{12}(s)Z_{21}(s)}.
\]
The model is explicitly an LTI approximation around a specific operating condition and does not account for dependence of impedance on excitation amplitude and external magnetic field [2509.24683].

A complementary extraction route starts directly from Maxwell’s equations in a finite-element setting. For an \(N\)-branch system, one excites one branch at a time with current \(I_0\), computes branch voltages from terminal-averaged compensated scalar-potential differences, and assembles the impedance matrix column by column via
\[
Z_{ij}=\frac{V_{ij}}{I_0}.
\]
For two ports, this yields \(Z_{11},Z_{21}\) from excitation of port 1 and \(Z_{12},Z_{22}\) from excitation of port 2. The method supports arbitrary conductor geometry and inhomogeneous permittivities and permeabilities, and includes a mandatory low-frequency stabilization scheme [2009.08232].

Metrological and bench-measurement variants often use terminal-pair or modal formulations rather than a direct \(2\times2\) \(Z\)-matrix. In a two terminal-pair digital impedance bridge, each standard is modeled as a two-port \(\Pi\) network with transadmittance \(Y_X\), high-to-shield stray admittance \(y_{HX}\), and low-to-shield stray admittance \(y_{LX}\), and the corrected ratio model is
\[
W = W^{(\mathrm r)}(1+\epsilon_W),
\]
\[
\epsilon_W = -\frac{1}{2}\Delta g_{\mathrm{FR}} + \frac{1}{2}(z_1+z_2)\left[(Y_B+y_{HB})-(Y_A+y_{HA})\right].
\]
In the twin-wires method with a two-port VNA, the measured transmission includes modal contamination,
\[
S_{dd}^{\mathrm{meas}} = S_{dd21} + 2S_{cc21}(\varepsilon_{21}^{c})^2,\qquad
S_{cc}^{\mathrm{meas}} = S_{cc21} + 2S_{dd21}(\varepsilon_{21}^{d})^2,
\]
so hybrid isolation and splitter magnitude balance enter directly into the extracted impedance error budget [1408.2638][1410.3201].

## 5. Statistical and interconnected two-port models

In complicated wave-scattering systems, the two-port impedance matrix is treated statistically. The random coupling model uses the normalized impedance
\[
Z_{n}=R_{avg}^{-1/2}\left(Z-iX_{avg}\right)R_{avg}^{-1/2},
\]
where \(Z_{avg}=R_{avg}+iX_{avg}\) contains nonuniversal average response due to coupling and short trajectories. For a reciprocal two-port, the impedance variance ratio
\[
\Xi_Z=\frac{\operatorname{Var}[Z_{12}]}
{\sqrt{\operatorname{Var}[Z_{11}]\operatorname{Var}[Z_{22}]}}
\]
is predicted to be a universal function of the loss parameter
\[
\alpha=\frac{f}{2Q\Delta f},
\]
whereas the corresponding scattering variance ratio is generally nonuniversal unless \(\alpha\gg 1\) [1309.3302].

A geometrically refined average impedance for a realistic two-port chaotic cavity is
\[
Z_{avg}=Z_{R}+R_{R}^{1/2}z R_{R}^{1/2},
\]
where \(Z_R\) is the radiation impedance and the matrix \(z\) sums short ray trajectories connecting port \(n\) to port \(m\):
\[
z_{n,m}=\sum_{b(n,m)}\{-p_{b(n,m)}\sqrt{D_{b(n,m)}}\exp[-(ik+\kappa)L_{b(n,m)}-ikL_{port(n,m)}-i\beta_{b(n,m)}\pi]\}.
\]
In the two-port case this includes not only \(z_{11}\) and \(z_{22}\), but also the off-diagonal inter-port terms \(z_{12}\) and \(z_{21}\), including the direct trajectory between the two ports without wall bounces [1006.3040].

For assemblies of interconnected subnetworks, a closed-form equivalent scattering matrix is
\[
\tilde{\mathbf S}
=
\mathbf S_{\mathcal{NN}}
+
\mathbf S_{\mathcal{NC}}
\left(
(\mathbf S_{\mathrm{con}})^{-1}-\mathbf S_{\mathcal{CC}}
\right)^{-1}
\mathbf S_{\mathcal{CN}},
\]
and the equivalent two-port impedance matrix can then be recovered, for equal real reference impedance \(Z_0\), by
\[
\tilde{\mathbf Z}_{2\text{-port}}=
Z_0(\mathbf I+\tilde{\mathbf S}_{2\text{-port}})
(\mathbf I-\tilde{\mathbf S}_{2\text{-port}})^{-1}.
\]
The same work argues that scattering parameters have a fundamental advantage for connection schemes that require some interconnections to be treated as delayless, lossless, reflectionless, and reciprocal two-port scattering systems [2412.17884].

## 6. Applications, limits, and recurrent misunderstandings

In grid-tied voltage-source converters, the fundamental small-signal model is inherently coupled. In the \(dq\) frame,
\[
\begin{bmatrix} U_d(s)\\ U_q(s) \end{bmatrix}
=
\begin{bmatrix} Z_{dd}(s) & Z_{dq}(s)\\ Z_{qd}(s) & Z_{qq}(s) \end{bmatrix}
\begin{bmatrix} I_d(s)\\ I_q(s) \end{bmatrix},
\]
and, after transformation, the sequence-domain model is likewise a \(2\times2\) impedance matrix. The paper shows that an accurate SISO sequence model derived by closed-loop equivalence gives identical stability conclusions to the full MIMO model, whereas a reduced SISO model based on the strong-grid assumption may lead to wrong results if the PLL bandwidth is large [1704.04157].

In physical human–robot interaction, two-port analysis is used in a hybrid form rather than a pure \(Z\)-matrix form. For series damped elastic actuation under velocity-sourced impedance control, the controlled actuator is modeled as a two-port hybrid immittance between the human side and the virtual-environment side. The paper provides necessary and sufficient conditions for two-port passivity and proves the necessity of a dissipative element parallel to the series elastic component and the necessity of a virtual coupler with dissipation for absolute stability and two-port passivity. In a different nonreciprocal setting, a three-electrode Hall device becomes a two-port when one electrode is chosen as common ground; the resulting two-port admittance and scattering description predicts anti-reciprocity at
\[
\Omega_n=\frac{\pi}{r}(2n+1),
\]
and self-matching in the low-\(\alpha\) regime for simple rational values of \(r\), especially \(r=2\) [2011.00664][1609.06543].

Several recurrent misunderstandings are directly addressed in the literature. One is that a constitutive equation may be mathematically correct but physically incomplete: for the time-varying capacitor, \(I=C\dot V+\dot C\,V\) is valid, but not a complete one-port model. Another is that any two-port can be assigned a fixed global \(Z\)-matrix: the cME magnetometer model is explicitly an operating-point-specific LTI approximation, and the time-varying capacitor admits at most local small-signal linearization around a trajectory. A third is that every port-based quantum or nanoscale device already has a derived two-port matrix model: the low-frequency double-quantum-dot paper provides a one-port admittance seen from the gate electrode,
\[
Y_{\mathrm{eq}}=i\omega(C_{\mathrm{geom}}+C_Q+C_t)+\frac{1}{R_{\mathrm{Sis}}},
\]
and states that a genuine two-port extension would require additional terminal definitions and electrostatic coupling relations beyond the derived model [2509.24683][2210.14057][1812.06056].

Source: https://www.emergentmind.com/topics/two-port-impedance-model