When an Interconnect Becomes a Transmission Line

01 / 10

When a 30 mm wire changes the circuit

The schematic calls it a connection. Why do simulation and the bench say it moved a 2.45 GHz filter response?

Illustrative engineering case: a battery-powered condition-monitoring node contains a 5 mm IC escape, a 30 mm PCB antenna feed, and a 1 m test cable. The first schematic treats all three as ideal wires. The RF simulation includes the feed as a guided structure, and the lab calibration stops at a connector. The three descriptions therefore do not place voltage and phase at the same locations.

The physical length alone is not the decision. A useful question is: how much can the signal vary while information travels from one endpoint plane to the other?That requires propagation velocity, the signal variation that matters, and an accuracy target.

Think about itFor a 10° allowance, which of the 5 mm escape, 30 mm feed, and 1 m cable can be lumped at a 2.45 GHz carrier?

Answer

Under the lesson’s ε_eff = 3.4 uniform-line model, none. They accumulate about 27.12°, 162.75°, and 5,424.85° respectively. Even the 5 mm escape exceeds 10°—well before the familiar quarter-wavelength mark of 90°.

5 mmIC escapePhysically small; still 27.12° at the carrier in the declared model.
30 mmAntenna feed184.52 ps delay and 162.75° of unwrapped carrier phase.
1 mTest cableMore than six nanoseconds and many complete carrier cycles.

Common misconceptionShorter than a quarter wavelength is always lumped.

A quarter wavelength is 90°. A design allowing only 10° cannot wait until 90° to change models. Wavelength fractions are engineering shortcuts only after the allowable error, propagation model, signal content, and endpoints have been declared.

02 / 10

Choose the model from delay and variation

Which frequency should decide the model: carrier, occupied bandwidth, or a fast edge?

They answer different questions. Carrier phase tells how far a sinusoidal component rotates between the endpoint planes. Occupied-band phase spread tells how much relative phase can change across a declared band. Edge content tests whether a fast transition can stay nearly uniform along the connection. The engineering purpose decides which quantity binds.

1 · PlaneName both endpointsLength is separation between specific planes, not a drawing dimension.
2 · VariationName carrier, band, and edgeDo not replace a real spectrum with one universal “highest frequency.”
3 · AccuracyDeclare allowed phaseHere the local teaching criterion is 10°, not a universal rule.
4 · DecisionExpose the inequalityIf relevant phase exceeds the allowance, use the distributed model.

$\begin{aligned}v_p&=\frac{c}{\sqrt{\epsilon_{\mathrm{eff}}}}\\t_d&=\frac\ell{v_p},\quad\theta=360^\circ f t_d\end{aligned}$ v_p is phase velocity [m/s]; c is exactly 299,792,458 m/s; ℓ is plane separation [m]; t_d is one-way delay [s]; f is ordinary frequency [Hz]; and θ is unwrapped phase magnitude [degrees].

For the edge comparison, this module uses $f_{\mathrm{knee}}=\frac{0.35}{t_r}$ . The 0.35 constant corresponds to a particular rise-time/frequency-response relationship; it is a selectable engineering proxy, not a hard edge of the waveform’s spectrum.

Class 1 · embedded decision model

Lumped-or-Distributed Ruler

Declare an error allowance, then compare carrier phase, occupied-band phase spread, and a 0.35/t_r edge proxy. The result is a model choice, not fabrication sign-off.

tl-ruler/1.0.0

Status · calculated

Distributed model required

Carrier phase binds: 162.75° is 16.27× the declared 10.0° allowance. Increasing length, frequency, or ε_eff increases electrical length in this model.

Propagation velocity: 1.62585 × 10⁸ m/s
One-way delay: 184.52 ps
Carrier wavelength: 66.361 mm
Carrier electrical length: 162.75°
Occupied-band phase spread: 0.06643°
Edge proxy: 175.000 MHz
Delay / edge time: 9.226%
Carrier length at allowance: 1.843 mm

Transparent decision checks · values at named endpoint planes
Check	Calculated phase	Inequality	Result
Carrier phase	162.75°	162.75° ≤ 10.0°	Exceeds allowance
Phase spread across occupied bandwidth	0.06643°	0.06643° ≤ 10.0°	Pass
Edge-proxy phase	11.62°	11.62° ≤ 10.0°	Exceeds allowance

Endpoint convention: +z runs from source-side plane P_S to load-side plane P_L; phasors use e^+jωt. Text above is the complete diagram equivalent.

Illustrative node comparison · current ε_eff, carrier, edge proxy, and phase allowance
Connection	Delay	Carrier case	Edge-proxy case
5 mm IC escape	30.75 ps	27.12° · distributed model	1.937° · lumped candidate
30 mm antenna feed	184.52 ps	162.75° · distributed model	11.62° · distributed model
1 m test cable	6.151 ns	5,424.8° · distributed model	387.49° · distributed model

Model: calculated, uniform, nondispersive, low-loss line with v_p = c/√ε_eff. It omits frequency-dependent material behavior, conductor/dielectric loss, mode conversion, discontinuities, launches, and termination response. A future Tools candidate may generalize it; this embedded version remains a teaching decision record. The 0.5 relative-bandwidth warning is this ruler's conservative inspect trigger, not a universal physics boundary.

Server-rendered default · printable / no-script record

30 mm between P_S and P_L requires the distributed model.

Model tl-ruler/1.0.0: ε_eff = 3.4, carrier 2.45 GHz, occupied bandwidth 1 MHz, 2 ns edge, and 10° allowance. Uniform, nondispersive, low-loss approximation; +z is P_S → P_L.

v_p: 1.62585 × 10⁸ m/s
One-way delay: 184.52 ps
λ at 2.45 GHz: 66.361 mm
Delay / rise time: 0.09226

Criterion	Inequality	Result
Carrier phase	162.75° > 10°	Exceeds
Band phase spread	0.06643° ≤ 10°	Passes
Edge-proxy phase	11.62° > 10°	Exceeds

The carrier is the binding input. The conclusion is analytic, not measured. Without JavaScript, these equations, inequalities, and the final six-case synthesis remain complete.

Common misconceptionCarrier frequency is the only frequency that matters.

A carrier may bind an absolute phase budget while a narrow occupied band produces little phase spread. A fast control edge on the same route can bind a different decision. State the signal mode and the error being controlled.

Common misconceptionDelay matters only on digital lines.

Delay rotates the phase of every sinusoidal RF component. It affects coherent combination, filter placement, calibration-plane movement, and group delay even when no logic edge exists.

03 / 10

A line stores energy everywhere

What physical model replaces the single ideal wire once position matters?

Take an electrically tiny slice of length Δz. Current along its conductors encounters series resistance and magnetic energy storage. Voltage between conductors produces electric energy storage and dielectric leakage. Repeat that slice continuously and voltage and current become functions of position as well as time.

Definition · uniform-line infinitesimal slice. R′ [Ω/m] and G′ [S/m] model loss; L′ [H/m] and C′ [F/m] store magnetic and electric energy. Values need not remain frequency independent in a real structure.

R′ [Ω/m] is series resistance per metre, including conductor-loss behavior in the chosen model.
L′ [H/m] relates current to magnetic energy stored per metre.
G′ [S/m] is shunt conductance per metre for dielectric leakage/loss.
C′ [F/m] relates voltage to electric energy stored per metre.

The prime means “per unit length,” not a derivative. Multiplying by Δz gives the lumped element value of one tiny slice; taking Δz toward zero produces the distributed model.

Go deeperEnergy storage sets the lossless line scales

When R′ = G′ = 0, the line’s wave impedance is √(L′/C′) and its velocity is 1/√(L′C′). The ratio of magnetic to electric storage sets the wave’s V/I ratio, while their product sets how quickly a disturbance advances.

04 / 10

From local laws to traveling waves

How does one RLGC slice force voltage and current to propagate rather than change everywhere at once?

Kirchhoff’s voltage and current laws applied to the slice say that a change with position is paid for by its series impedance and shunt admittance. With the portfolio phasor convention e^+jωt, ordinary frequency is f [Hz] and angular frequency is ω = 2πf [rad/s].

$\frac{d V}{d z} = -(R' + j\omega L') I$ $\frac{d I}{d z} = -(G' + j\omega C') V$ V(z) is phasor voltage [V]; I(z) is current [A] defined positive in +z; and z is distance [m] from the source-side plane.

Differentiate either equation once more and substitute the other. The coupled first-order laws become two second-order wave equations with the same propagation constant.

$\frac{d^{2}V}{d z^{2}} = \gamma ^{2}V \qquad \frac{d^{2}I}{d z^{2}} = \gamma ^{2}I$ $V(z) = V^{+}e^{-\gamma z} + V^{-}e^{+\gamma z}$ $I(z) = (\frac{V^{+}}{Z_{0}})e^{-\gamma z} - (\frac{V^{-}}{Z_{0}})e^{+\gamma z}$ V⁺ and V⁻ are voltage-wave amplitudes referenced at z = 0. The minus sign in the reverse-wave current follows the declared positive current direction.

Think about itWhich term travels toward the load when +z points from source to load?

Answer

V⁺e^−γz. Since γ = α + jβ, its time-and-space phase is ωt − βz under e^+jωt, which advances in +z as time increases. The V⁻ term travels in −z.

Both terms are kept here so the notation is complete. What creates V⁻ at a termination is a boundary-condition problem owned by Module 03.2; no reflection formula is needed yet.

05 / 10

Characteristic impedance is a wave ratio

How can a lossless line have a real 50 Ω value without containing a row of 50 Ω resistors?

Take only the +z traveling term and divide its voltage by its current. The common spatial exponential cancels. The result is the line’s characteristic impedance, a property of the uniform propagation mode at the stated frequency.

$Z_{0} = \frac{V^{+}}{I^{+}} = \sqrt{\frac{R' + j\omega L'}{G' + j\omega C'}}$ Z₀ is characteristic impedance [Ω]. It may be complex for a lossy line. The square-root branch is chosen for a passive forward wave.

In the ideal lossless limit, Z₀ = √(L′/C′), which is real. If the line continues without end, energy keeps moving into new distributed electric and magnetic fields. The source can therefore supply real power even though the line does not dissipate that power locally.

Common misconceptionZ₀ is a 50 Ω resistor distributed down the cable.

Z₀ is a traveling-wave V/I ratio. R′ is the distributed series-loss term. Confusing them hides the distinction between power transported by an ideal line and power dissipated by conductor or dielectric loss.

Common misconceptionA 50 Ω label guarantees the same impedance at every frequency and plane.

The label describes a nominal interface or mode under stated conditions. Real line impedance can vary with frequency and geometry, while total V/I at a plane can also depend on forward and reverse waves. Module 03.2 develops that boundary response.

06 / 10

Propagation constant separates loss and phase

What does the exponent e^−γz do to a forward wave as it moves along the line?

Its real part reduces amplitude; its imaginary part accumulates phase. Keeping those jobs separate prevents attenuation, electrical length, and delay from collapsing into one vague idea of “line length.”

$\gamma = \sqrt{(R' + j\omega L')(G' + j\omega C')} = \alpha + j\beta$ $\lambda = \frac{2\pi }{\beta } \qquad v_{p} = \frac{\omega }{\beta }$ γ is propagation constant [1/m]; α is attenuation constant [Np/m]; β is phase constant [rad/m]; λ is wavelength [m].

Calculated normalized sketch · one forward sinusoidal component between the endpoint planes. Conditions: uniform line, constant α and β at one frequency. No measured data.

Definition · consistent attenuation units

One amplitude ratio and one power ratio produce the same dB loss.

Over length ℓ, forward voltage-wave amplitude is multiplied by e^−αℓ and power by e^−2αℓ. Therefore 20 log₁₀(e^−αℓ) and 10 log₁₀(e^−2αℓ) both equal −8.68589 αℓ dB. The numerical dB value agrees because the quantity and logarithm factor are paired correctly.

07 / 10

Lossless and low-loss are conditional models

What evidence earns the right to drop R′ and G′ from the exact equations?

Physical size is not that evidence. Compare each loss term with the reactive term beside it at the frequency of interest. A low-loss approximation requires both dimensionless ratios to be small over the declared band.

$\frac{R'}{\omega L'}\ll1\qquad\text{and}\qquad\frac{G'}{\omega C'}\ll1$ These checks compare series resistance with inductive reactance per metre and shunt conductance with capacitive susceptance per metre. “Small” must fit the required accuracy.

Lossless · R′ = G′ = 0: Z₀ = √(L′/C′), α = 0, β = ω√(L′C′), and v_p = 1/√(L′C′).
Low loss · first order: Z₀ ≈ √(L′/C′), β ≈ ω√(L′C′), and α ≈ R′/(2Z₀) + G′Z₀/2.

Derived · checked 2.45 GHz RLGC example

A line can be low loss while remaining electrically long.

Choose L′ = 307.5309 nH/m and C′ = 123.0124 pF/m so the lossless scales are Z₀ = 50.000 Ω and v_p = 1.62585 × 10⁸ m/s. Add illustrative R′ = 0.800 Ω/m and G′ = 3.00 µS/m at 2.45 GHz.

$\frac{R'}{\omega L'}$ 1.6899 × 10⁻⁴
$\frac{G'}{\omega C'}$ 1.5843 × 10⁻⁶
Exact Z₀50.0000 − j0.00419 Ω
Exact γ0.008075 + j94.6814 m⁻¹
Attenuation0.07014 dB/m of wave transmission

Both ratios are small, and the first-order α is 0.008075 Np/m, agreeing with the exact result to the shown precision. Yet the 30 mm carrier phase is still 162.75°: low loss says little about electrical shortness.

Think about itWhat happens as R′ and G′ approach zero while L′ and C′ remain positive?

Answer

α approaches zero, Z₀ approaches the real value √(L′/C′), and β approaches ω√(L′C′). The line can still have any electrical length; lossless does not mean lumped.

Common misconceptionA low-loss approximation is valid because the line looks physically small.

Loss validity follows R′/(ωL′), G′/(ωC′), frequency, and required accuracy. Lumped validity follows delay and signal variation. These are different decisions and can produce different answers.

08 / 10

Phase velocity is not group delay

If one sine wave has a clear phase velocity, will a modulated waveform keep its shape?

Phase velocity follows a constant-phase point of one sinusoidal component. A modulation envelope depends on the relative phase of nearby components, so its delay is governed by the slope of β with angular frequency.

$v_{p} = \frac{\omega }{\beta }$ $\tau _{g} = \frac{\ell d\beta }{d\omega } \qquad v_{g} = \frac{d\omega }{d\beta }$ τ_g is group delay [s] across length ℓ; v_g is group velocity [m/s]. For forward phase φ = −βℓ, τ_g = −dφ/dω = ℓ dβ/dω.

If β(ω) is a straight line through the relevant band, phase and group velocities are the same constant in this model. Curvature makes different spectral components accumulate different delay, so the waveform can distort. One value of ε_eff cannot capture that behavior across an arbitrary real line and band.

Illustrative, not material data · β [rad/m] versus ω [rad/s]. Full geometry and material dispersion extraction is deferred to Module 03.3.

Common misconceptionPhase velocity and group velocity are interchangeable.

They agree only under a suitable nondispersive approximation. For a wideband waveform, inspect β(ω) or measured/simulated group delay over the band rather than dividing one center-frequency phase by frequency and assuming the result applies everywhere.

09 / 10

Every length lives between two planes

When two engineers quote different phase for the same line, are they necessarily disagreeing?

Not until their reference planes match. This lesson places P_S at the source-side endpoint, defines z = 0 there, and places P_L at the load-side endpoint z = ℓ. Positive current and +z point from source toward load.

Definition · source-to-load direction and named endpoint planes. Moving P_Sto P_S′ changes the included delay and phase even before a discontinuity or termination response is analyzed.

Plane P_S: Source-side endpoint · z = 0 · wave amplitudes referenced here.
Direction: +z and positive current point toward P_L.
Plane P_L: Load-side endpoint · z = ℓ · termination response deferred.

Derived · moving one plane

Ten millimetres is 54.25° at 2.45 GHz in the default model.

Moving the source-side plane 10 mm toward the load removes 61.51 ps and 54.25° of unwrapped carrier phase from the reported path. The physical line did not change; the quantity assigned to the remaining two-plane network did.

IEEE 370 uses carefully controlled fixture and reference-plane concepts for high-frequency interconnect characterization. This lesson uses those ideas only for orientation; VNA/TDR calibration, de-embedding, and uncertainty remain in Path 08.

10 / 10

Classify the sensor-node interconnects

Which simplest model survives the same 10° requirement for the carrier and the 2 ns edge?

Use the default uniform ε_eff = 3.4 approximation between named endpoints. For the narrowband carrier case, compare carrier phase with 10°. For the edge case, compare the explicitly chosen 175 MHz proxy phase with 10°. This produces six decisions—not one label permanently attached to each piece of copper.

Illustrative model-choice record · tl-ruler/1.0.0
Connection / signal	Binding value	10° criterion	Chosen model	Limitation / revisit evidence
5 mm IC escape 2.45 GHz carrier	27.12°	27.12° > 10°	Distributed model	Constant ε_eff; revisit with extracted β(f), geometry, or a changed phase budget.
5 mm IC escape 2 ns edge proxy	1.937°	1.937° ≤ 10°	Lumped candidate	0.35/t_r proxy; revisit with the actual driver spectrum and allowed waveform error.
30 mm antenna feed 2.45 GHz carrier	162.75°	162.75° > 10°	Distributed model	Constant ε_eff; revisit with extracted β(f), geometry, or a changed phase budget.
30 mm antenna feed 2 ns edge proxy	11.62°	11.62° > 10°	Distributed model	0.35/t_r proxy; revisit with the actual driver spectrum and allowed waveform error.
1 m test cable 2.45 GHz carrier	5,424.8°	5,424.8° > 10°	Distributed model	Constant ε_eff; revisit with extracted β(f), geometry, or a changed phase budget.
1 m test cable 2 ns edge proxy	387.49°	387.49° > 10°	Distributed model	0.35/t_r proxy; revisit with the actual driver spectrum and allowed waveform error.

The 5 mm connection changes class: distributed for the 2.45 GHz carrier, lumped candidate for this 2 ns edge proxy. The 30 mm feed is distributed for both, although its edge check is only modestly over the chosen limit. The 1 m cable is distributed by a wide margin.

Path 03 capstone artifact · part 1

Write the reference-plane map and model-choice record.

Your record should let another engineer reproduce the decision without guessing what “short,” “fast,” or “50 Ω” meant.

Draw P_S and P_L; mark +z from source to load and record ℓ.
State e^+jωt, the +z term e^−γz, units, and the propagation approximation.
Name the operating mode, carrier, occupied-band definition, edge proxy, and allowed error.
Show every inequality, identify the binding one, and choose the simplest adequate model.
Record the limitation and the simulation, measurement, stackup, or requirement change that triggers review.

Think about itWhat question naturally follows once the connection must carry forward and reverse waves?

Answer

What boundary condition exists at its termination, and what wave returns? Module 03.2 will develop reflection, standing waves, and input-impedance transformation from that need.

Ungraded review

Check your understanding

Answer each question in your own words, then reveal the model answer.

01Why is ‘shorter than one quarter wavelength’ not a complete lumped-model criterion?
Model answer
Because adequacy depends on allowed error and the relevant signal variation. Compare propagation delay with carrier phase, occupied-band phase spread, and edge content under a stated criterion; a quarter wavelength is 90°, which may be far beyond the permitted phase error.
02For e^(+jωt) and +z from source to load, which exponential represents a forward wave?
Model answer
V⁺e^(−γz). Its phase term is e^(−jβz), so increasing z produces the spatial phase lag of a wave traveling toward the load. The reverse voltage term is V⁻e^(+γz).
03What physical roles do R′, L′, G′, and C′ play?
Model answer
R′ [Ω/m] models series conductor loss, L′ [H/m] magnetic energy storage, G′ [S/m] shunt dielectric leakage/loss, and C′ [F/m] electric energy storage. Real values can vary with frequency, material, temperature, and geometry.
04A lossless line has a real 50 Ω characteristic impedance. Where is the 50 Ω resistor?
Model answer
There need not be one. The real Z₀ = V⁺/I⁺ = √(L′/C′) is the voltage-to-current ratio of one traveling wave. An infinitely long ideal line accepts power because energy continues to move into fresh distributed L′ and C′, not because each slice dissipates it.
05When do phase velocity and group velocity agree in this lesson’s simple model?
Model answer
When β is proportional to ω over the signal band. Then ω/β and dω/dβ are the same constant. Curvature in β(ω) means different frequency components have different group delay and the single-velocity model is insufficient.
06What must accompany a final lumped-or-distributed decision?
Model answer
Named endpoint planes and +z direction; signal mode and relevant frequency variation; phase or delay allowance; selected model and approximation; units; binding inequality; limitation; and the evidence that would cause the decision to be revisited.

Sources and further study

Accessed 5 September 2026. Equations and diagrams are redrawn from the stated distributed model; no source illustration or paywalled table is reproduced.

Theory and constants

David M. Pozar, Microwave Engineering, 4th ed., Chapter 2, Transmission Line Theory. Stable theory source; geometry-specific extraction is outside this module.
NIST/CODATA, speed of light in vacuum, 2022 CODATA value: 299,792,458 m/s exactly.
University of Kansas EECS 723, transmission-line handouts: distributed model, telegrapher equations, wave equation, Z₀, γ, and lossless limits.

Interconnect and measurement orientation

IEEE, IEEE 370-2020, active; published 8 January 2021, with errata dated 21 January 2022. Normative interconnect-characterization context; detailed procedure is deferred.
Keysight, Port Extensions: practical orientation on moving a VNA measurement reference plane to account for line delay and loss.
Tektronix, ABCs of Probes: source for the response-dependent rise-time/bandwidth relationship and why 0.35 is conditional.

Review a prerequisiteFields & Waves

Next in Transmission Lines & MatchingTerminations, Reflections & Standing Waves

Back to Transmission Lines & Matching

When an Interconnect Becomes a Transmission Line

When a 30 mm wire changes the circuit

Choose the model from delay and variation

Lumped-or-Distributed Ruler

Distributed model required

30 mm between PS and PL requires the distributed model.

A line stores energy everywhere

From local laws to traveling waves

Characteristic impedance is a wave ratio

Propagation constant separates loss and phase

One amplitude ratio and one power ratio produce the same dB loss.

Lossless and low-loss are conditional models

A line can be low loss while remaining electrically long.

Phase velocity is not group delay

Every length lives between two planes

Ten millimetres is 54.25° at 2.45 GHz in the default model.

Classify the sensor-node interconnects

Write the reference-plane map and model-choice record.

Check your understanding

Sources and further study

Theory and constants

Interconnect and measurement orientation

30 mm between P_S and P_L requires the distributed model.