Electronics Guide

Resistance (R)

Series resistance per unit length, measured in ohms per meter (Ω/m), represents the conductor's opposition to current flow due to the finite conductivity of the metal. At DC and low frequencies, resistance is determined primarily by the conductor's cross-sectional area and material resistivity. However, at higher frequencies, the skin effect causes current to concentrate near the conductor surface, effectively reducing the cross-sectional area and increasing resistance. Surface roughness further increases resistance by increasing the effective path length for current flow.

Inductance (L)

Series inductance per unit length, measured in henries per meter (H/m), arises from the magnetic field surrounding current-carrying conductors. For a transmission line, both the external inductance (due to magnetic flux between conductors) and internal inductance (due to magnetic flux within the conductors themselves) contribute to the total value. The inductance depends on the geometry of the conductors and their separation, and at high frequencies, skin effect reduces the internal inductance component as current becomes concentrated at the surface.

Conductance (G)

Shunt conductance per unit length, measured in siemens per meter (S/m), represents leakage current flowing through the dielectric material between conductors. This parameter accounts for losses in the insulating material and is related to the dielectric loss tangent. In good dielectrics, conductance is typically very small at DC but increases with frequency due to polarization losses in the dielectric material. For high-quality transmission lines, G is often negligible compared to other loss mechanisms.

Capacitance (C)

Shunt capacitance per unit length, measured in farads per meter (F/m), represents the ability to store electric charge between the conductors. This parameter depends on the geometry of the conductors, their separation, and the permittivity of the dielectric material between them. Unlike resistance and inductance, capacitance is largely independent of frequency for most practical transmission lines, though the effective dielectric constant may show some frequency dependence in certain materials.

The Heaviside Distortionless Condition

Oliver Heaviside, who cast the telegrapher's equations in their modern form in the 1880s, identified a special balance among the four parameters:

R/L = G/C

When this condition holds, the propagation constant factors cleanly into α = √(RG) and β = ω√(LC). The attenuation is then independent of frequency and the phase velocity is constant, so a pulse arrives attenuated but undistorted in shape. Real cables violate the condition badly—R/L is normally far larger than G/C—which is why long lines smear pulses rather than merely shrinking them.

Heaviside's insight led to the practice of loading telephone cables with series inductance at regular intervals to raise L and move the line toward the balanced condition, a technique that extended the usable range of voice circuits before electronic amplification became practical. Modern high-speed links pursue the same goal by the opposite route: rather than flattening the channel, they compensate for its slope with equalization.

Lossless Line (R = 0, G = 0)

For an ideal lossless transmission line:

Z₀ = √(L/C)

This fundamental result shows that the characteristic impedance of a lossless line is real, frequency-independent, and depends only on the ratio of inductance to capacitance per unit length. This is the impedance most commonly specified for transmission lines (50Ω for RF systems, 75Ω for video, etc.).

Low-Loss Line at High Frequency

For practical transmission lines with small losses (R << ωL and G << ωC), which is typical at high frequencies:

Z₀ ≈ √(L/C) × [1 - j(R/2ωL - G/2ωC)]

The characteristic impedance becomes slightly complex, but the magnitude remains close to √(L/C) with a small phase shift. Note the sign: because conductor loss normally dominates dielectric loss at moderate frequencies, R/ωL exceeds G/ωC and the reactive part is negative, so a real lossy line looks slightly capacitive rather than inductive.

The RC Regime at Low Frequency

Below the frequency where ωL overtakes R, a line enters a diffusive regime. Dielectric conductance is usually still negligible there (G << ωC), so:

Z₀ ≈ √[R/(jωC)]

The magnitude √(R/ωC) rises without bound as frequency falls, and the phase settles at -45 degrees. This is the regime that governs long, thin on-chip interconnects and low-frequency behavior on PCB traces; it is the reason a broadband model of a real line shows Z₀ climbing steeply toward DC instead of holding at 50 ohms. Only in the rare case where both R >> ωL and G >> ωC—a leaky line such as a long, wet buried cable at very low frequency—does the impedance return to a real value, Z₀ ≈ √(R/G).

Reflection and Input Impedance

Characteristic impedance acquires its practical importance through the reflection coefficient. When a wave traveling on a line of impedance Z₀ meets a load Z_L, the fraction of the incident voltage wave sent back toward the source is:

Γ = (Z_L - Z₀)/(Z_L + Z₀)

A matched load (Z_L = Z₀) gives Γ = 0, an open circuit gives Γ = +1, and a short circuit gives Γ = -1. In RF work the same mismatch is often quoted as a voltage standing wave ratio, VSWR = (1 + |Γ|)/(1 - |Γ|), or as return loss in decibels, -20 log₁₀|Γ|. Reflection mechanisms and termination strategies develop these consequences in detail.

Looking into a lossless line of length ℓ terminated in Z_L, the input impedance is:

Z_in = Z₀ (Z_L + jZ₀ tan βℓ)/(Z₀ + jZ_L tan βℓ)

This single expression contains most of the useful behavior of transmission lines. A quarter-wavelength line (βℓ = 90 degrees) reduces it to Z_in = Z₀²/Z_L, the impedance inverter behind quarter-wave matching transformers, and it turns a short into an open and an open into a short. A half-wavelength line reproduces Z_L exactly. A matched load makes Z_in = Z₀ for any length, which is precisely why matched systems are insensitive to trace length.

Phase Velocity

The phase velocity v_p describes how fast a point of constant phase (such as a wave peak) travels along the transmission line. For a lossless line, the phase velocity is:

v_p = ω/β = 1/√(LC)

Since the inductance and capacitance per unit length depend on the geometry and the permittivity and permeability of the dielectric material, the phase velocity is related to the speed of light c by:

v_p = c/√(ε_rμ_r)

where ε_r is the relative permittivity and μ_r is the relative permeability of the dielectric. For most practical dielectrics, μ_r ≈ 1, so:

v_p ≈ c/√ε_r

Common PCB materials such as FR-4 have ε_r ≈ 4.2 to 4.5 in the bulk, giving a phase velocity around 47 to 49 percent of the speed of light, or roughly 14 to 15 cm per nanosecond. A stripline, fully embedded in that dielectric, sees the bulk value and propagates at about 180 picoseconds per inch. On microstrip, where part of the field travels in the air above the trace, the effective permittivity falls to roughly 3.0 to 3.5 and the signal is correspondingly faster—about 16 to 17 cm per nanosecond, or 145 to 160 picoseconds per inch. The familiar rule of thumb of 6 inches per nanosecond corresponds to a homogeneous dielectric with ε_r ≈ 4, so it fits stripline well and understates microstrip speed by 10 to 15 percent.

Group Velocity

For signals containing multiple frequency components (such as digital pulses), the group velocity v_g describes how fast the signal envelope propagates. It is defined as:

v_g = ∂ω/∂β

For lossless, non-dispersive transmission lines, the group velocity equals the phase velocity. However, in dispersive media where the phase velocity depends on frequency, the group velocity differs from the phase velocity. This dispersion causes pulse spreading and is a critical consideration in high-speed digital systems and fiber optic communications.

A uniform TEM line in a single homogeneous dielectric is only weakly dispersive, and what dispersion it has comes mostly from frequency-dependent loss. Two structures depart from that ideal. Microstrip is inhomogeneous—part of its field is in air and part in the substrate—so as frequency rises the field concentrates in the higher-permittivity substrate, the effective permittivity climbs toward the bulk value, and the line slows down. Waveguides and any structure operating near a modal cutoff are strongly dispersive by construction, since phase velocity there is a steep function of frequency.

Time Delay

The propagation delay per unit length τ_d is the reciprocal of the propagation velocity:

τ_d = 1/v_p = √(LC)

For a line of length ℓ, the total time delay is:

T_d = ℓ × τ_d = ℓ/v_p

This delay drives timing budget analysis, clock distribution, and length matching of differential pairs and parallel buses. The numbers are unforgiving: at 180 picoseconds per inch, one millimeter of length mismatch is about 7 picoseconds of skew. Against a 25 Gbps unit interval of 40 picoseconds, that single millimeter consumes roughly 18 percent of the bit period, which is why intra-pair matching is specified in tenths of a millimeter on serial links.

Electrical Length

The electrical length of a transmission line is often more important than its physical length. It is expressed in terms of wavelength or degrees of phase shift:

θ = βℓ = 2πℓ/λ

where λ is the wavelength at the operating frequency. A quarter-wavelength line (θ = 90°) has special impedance-transforming properties, while a half-wavelength line (θ = 180°) appears electrically transparent, replicating the load impedance at the input.

Lossless Line Model

The lossless transmission line model assumes R = 0 and G = 0, leaving only the reactive elements L and C. While no real transmission line is truly lossless, this approximation is excellent for many practical situations, particularly:

• Short interconnects in digital systems where propagation time matters more than attenuation
• High-quality RF transmission lines over moderate distances at frequencies well below cutoff
• Initial design and analysis where loss effects can be added as perturbations later

For a lossless line, the propagation constant simplifies to:

γ = jβ = jω√(LC)

This means α = 0 (no attenuation) and the phase constant β = ω√(LC) is purely real. The characteristic impedance is real and frequency-independent:

Z₀ = √(L/C)

Waves propagate without amplitude decay, and reflections at impedance discontinuities are purely geometric, depending only on impedance ratios. The lossless model enables clean analytical solutions for reflection coefficients, standing wave patterns, and impedance transformations.

Low-Loss Line Model

For most practical transmission lines at high frequencies, losses are small but not negligible. The low-loss approximation assumes R << ωL and G << ωC, which allows perturbative treatment of losses while maintaining analytical tractability.

Under these conditions, the attenuation constant can be approximated as:

α ≈ (R/2Z₀) + (GZ₀/2)

where Z₀ ≈ √(L/C). This shows that attenuation comes from two sources: conductor losses (first term) and dielectric losses (second term). The phase constant remains approximately:

β ≈ ω√(LC)

This model is widely used for cable design, PCB trace analysis, and RF link budgets, where losses must be accounted for but do not fundamentally change the wave behavior. Channel loss mechanisms examines how these terms accumulate across a complete interconnect.

General Lossy Line Model

When losses are significant (such as long cables at high frequencies, poor conductors, or lossy dielectrics), the full lossy transmission line equations must be used without approximation. The propagation constant retains both substantial real and imaginary parts:

γ = √[(R + jωL)(G + jωC)]

In this regime, several important effects occur:

• Frequency-dependent attenuation: Higher frequencies experience greater loss, causing signal distortion and pulse spreading
• Complex characteristic impedance: The impedance becomes frequency-dependent and complex, affecting matching networks
• Dispersion: Different frequency components travel at different velocities, causing pulse broadening
• Reduced reflection coefficients: Losses damp reflections, making mismatches less severe but also making time-domain reflectometry less effective

Analyzing lossy lines often requires numerical methods or computer-aided tools, particularly for time-domain analysis of complex signals.

Resistance Variation

The series resistance R exhibits the strongest frequency dependence due to two primary mechanisms:

Skin Effect: At DC and low frequencies, current distributes uniformly across the conductor's cross-section. As frequency increases, current concentrates exponentially near the surface within a depth called the skin depth δ:

δ = √[2/(ωμσ)]

where μ is the permeability and σ is the conductivity of the conductor. For copper at room temperature, a convenient approximation is δ ≈ 66/√f μm, with f expressed in megahertz: the skin depth is about 66 μm at 1 MHz and only about 2.1 μm at 1 GHz.

As the effective conducting area decreases, resistance rises. For a round conductor of radius a in the regime where the skin depth is much smaller than the radius:

R_AC/R_DC ≈ a/(2δ) + 1/4

Because δ falls as 1/√f, the resistance climbs as √f once this regime is reached. The transition begins near the frequency where the skin depth equals the conductor radius, or roughly half the thickness for a flat trace. A 35 μm (one ounce) copper trace reaches that point at about 14 MHz, so essentially every digital edge and RF carrier of practical interest operates well inside the skin-effect regime. For PCB traces the estimate is optimistic: current crowds toward the edges of the trace and toward the face nearest the reference plane, so the effective conducting perimeter is smaller than geometry alone suggests.

Proximity Effect: In transmission lines with closely spaced conductors, the current distribution is further modified by magnetic fields from neighboring conductors. This proximity effect can increase resistance beyond the skin effect alone, particularly in closely coupled differential pairs or multi-conductor cables.

Inductance Variation

The total inductance consists of external inductance (from magnetic flux between conductors) and internal inductance (from flux within conductors). While external inductance remains essentially constant with frequency, internal inductance decreases as skin effect concentrates current near the surface.

At low frequencies where current is uniform, the internal inductance contributes significantly. As frequency increases and skin effect dominates, internal inductance approaches zero, and the total inductance approaches the external inductance value:

L(f) → L_external as f → ∞

For most high-frequency applications, this frequency dependence of inductance is much weaker than that of resistance and can often be neglected.

Capacitance Variation

Capacitance is generally the most frequency-independent of the four parameters for good quality dielectrics. However, some frequency dependence exists due to:

• Dielectric dispersion: The relative permittivity of some materials varies slightly with frequency due to molecular polarization mechanisms
• Apparent permittivity from roughness: Measurements on rough copper consistently extract a slightly higher effective permittivity, and therefore a slightly longer delay, than the same stackup with smooth foil—an artifact of the field following the contoured surface rather than a change in the resin itself

A larger source of capacitance variation on PCBs is not frequency at all but position. Glass-reinforced laminates are inhomogeneous at the scale of the weave: glass bundles have a higher permittivity than the surrounding resin, so a trace running above a bundle sees more capacitance and travels more slowly than one running above a resin-rich window. This fiber-weave effect is a leading cause of intra-pair skew and is managed by rotating the routing relative to the weave or by specifying spread-glass fabrics. Aside from these effects, capacitance can be treated as frequency-independent up to many gigahertz.

Conductance Variation

Dielectric conductance increases approximately linearly with frequency due to dielectric loss mechanisms. This is commonly expressed through the loss tangent tan(δ):

G = ωC × tan(δ)

The loss tangent represents the ratio of energy dissipated to energy stored in the dielectric per cycle. For low-loss materials like PTFE, tan(δ) ≈ 0.0002, while standard FR-4 has tan(δ) ≈ 0.02 at 1 GHz. Because G is proportional to frequency, dielectric losses become increasingly important at higher frequencies.

Classical Skin Effect Theory

Skin effect arises from the interaction between current flow and the magnetic field it creates. The time-varying magnetic field induces eddy currents that oppose the main current in the conductor's interior, forcing current toward the surface. This self-induction effect strengthens with frequency, progressively excluding current from the conductor core.

For a cylindrical conductor, the current density decreases exponentially from the surface:

J(x) = J_surface × e^-x/δ

where x is the depth below the surface and δ is the skin depth. The skin depth represents the depth at which current density has fallen to 1/e (about 37%) of its surface value.

For a round wire the exact AC resistance follows from a Bessel-function solution of the diffusion equation, conventionally written with Kelvin functions:

R_AC/R_DC = (q/2) × [ber(q)bei'(q) - bei(q)ber'(q)] / [ber'²(q) + bei'²(q)]

where q = √2 × a/δ, a is the conductor radius, and ber and bei are the Kelvin functions (the real and imaginary parts of a Bessel function of complex argument). For q much greater than unity this expression converges to the a/(2δ) + 1/4 approximation given earlier, and hence to the √f behavior; below q ≈ 1 the ratio approaches unity and the conductor behaves as it does at DC.

For rectangular conductors like PCB traces, the analysis is more complex, but the fundamental behavior remains: resistance increases as √f when skin depth is much smaller than conductor dimensions.

Surface Roughness Effects

Real conductors are not perfectly smooth. PCB copper foils, in particular, have intentional surface roughness to promote adhesion to the dielectric substrate. This roughness significantly affects high-frequency resistance beyond what classical skin effect predicts.

When skin depth becomes comparable to or smaller than the roughness feature size, current must travel a longer path following the surface contours, increasing the effective length and hence resistance. Several models describe this effect:

Hammerstad-Bekkadal Model: One of the earliest practical models relates the roughness-enhanced resistance to RMS surface roughness R_q:

K_rough = 1 + (2/π) × arctan[1.4(R_q/δ)²]

where K_rough is the resistance multiplication factor. Roughness effects become significant once δ approaches R_q. The model's weakness is visible in its own form: the arctangent saturates, so the correction factor cannot exceed 2 no matter how rough the surface or how high the frequency. Measurements above roughly 10 GHz routinely exceed that ceiling, which is what motivated the physics-based models that followed.

Huray "Snowball" Model: Scanning electron micrographs of electrodeposited copper show clusters of roughly spherical nodules stacked in pyramidal piles, and this model takes that geometry literally. It treats the surface as an array of conducting spheres and computes the additional power lost to the field scattering from them, parameterized by an effective sphere radius and the number of spheres per unit area. Because it is grounded in the actual loss mechanism rather than in a path-length correction, it does not saturate and tracks measurements far better at high frequency.

Cannonball-Huray Model: A practical simplification of the Huray model that fixes the sphere arrangement to a close-packed stack, allowing the sphere radius and tile area to be derived from a single supplier-published roughness number rather than from fitting to measured loss data. This matters in practice, because foil vendors publish R_z while the general Huray model requires parameters that most designers cannot measure.

Practical implications include:

• Foil grades are specified by R_z, the mean peak-to-valley height: standard electrodeposited (HTE) foil runs roughly 5-10 μm, reverse-treated foil roughly 3-5 μm, very low profile (VLP) roughly 2-4 μm, and hyper very low profile (HVLP) roughly 1-2 μm
• Relative to ideally smooth copper, standard foil can nearly double conductor loss in the 10 GHz region; low-profile grades recover much of that margin
• The penalty grows through the band where skin depth becomes comparable to the roughness features—roughly 1 to 20 GHz for common PCB copper—since below it the current does not yet hug the profile
• Links at 56 Gbps PAM4 and above generally require low-profile foil, because conductor loss at these rates competes directly with the equalization budget
• Smoother foil adheres less well, so the choice trades insertion loss against peel strength and process yield

Design Considerations

Managing skin effect and roughness losses requires several strategies:

• Conductor selection: Use low-roughness copper for high-frequency applications; accept higher roughness for better adhesion in lower-frequency designs
• Wider traces: At high frequencies where current is confined to the surface, increasing trace width proportionally decreases resistance
• Plating quality: Smooth plating reduces roughness penalty, but plating itself adds resistance if poorly controlled
• Material choice: Lower-loss dielectrics become more important at higher frequencies where conductor losses are already significant
• Accurate modeling: Modern signal integrity tools include roughness models; ensure your simulator accounts for these effects with parameters matched to your fabrication process

Polarization and Loss Tangent

When an electric field is applied to a dielectric, the material becomes polarized as charges and dipoles respond to the field. In AC conditions, the polarization must continuously reverse direction, and the energy required for this reorientation is partially dissipated as heat.

The complex permittivity captures both energy storage and dissipation:

ε = ε' - jε''

where ε' is the real part (energy storage) and ε'' is the imaginary part (energy loss). The loss tangent is the ratio:

tan(δ) = ε''/ε'

A low loss tangent (below 0.001) indicates a high-quality dielectric; values above 0.01 indicate significant losses. By long-standing convention the loss angle is written δ, which is unrelated to the skin depth δ of the preceding section; context distinguishes them, as the loss angle appears only inside tan δ. The loss tangent relates directly to the conductance parameter:

G = ωC tan(δ)

Dielectric loss therefore increases linearly with frequency, while conductor loss increases only as √f. The two curves cross, and the crossover is the single most useful fact in material selection: below it, smoother copper buys the most improvement; above it, a lower-loss resin does. For standard laminates and typical trace widths the crossover falls in the low gigahertz, which is why multi-gigabit designs are dominated by dielectric choice while radio-frequency and lower-speed designs are often limited by the conductor.

Physical Loss Mechanisms

Dipolar Relaxation: Polar molecules or groups within the dielectric have permanent dipole moments that attempt to align with an applied electric field. At lower frequencies, dipoles can follow field oscillations, but as frequency increases, the inertia of molecular reorientation causes a phase lag between the field and polarization. This lag means energy is absorbed from the field and converted to molecular kinetic energy (heat). Each type of dipole has a characteristic relaxation frequency where loss peaks.

Ionic Conduction: Mobile ions in the dielectric, even in very small concentrations, contribute to loss by drifting in response to the applied field. This mechanism is essentially a very small DC conductivity added to the loss. It matters most at lower frequencies and in materials with higher moisture content or residual processing chemistry.

Electronic Polarization: At very high frequencies (optical and above), electronic cloud distortion in atoms becomes the dominant polarization mechanism. The losses associated with electronic polarization are usually negligible at RF and microwave frequencies.

Interfacial Polarization: In composite or heterogeneous dielectrics (like fiber-reinforced materials), charge accumulation at interfaces between different phases causes additional loss. This Maxwell-Wagner polarization is significant in glass-reinforced PCB materials and can vary with fiber weave pattern.

Common Dielectric Materials

Understanding typical material properties helps in design choices. Material selection treats the sourcing and cost dimensions of this decision in more depth.

PTFE: Excellent high-frequency properties, with ε_r ≈ 2.1 and tan(δ) ≈ 0.0002 to 0.0004. The non-polar molecular structure leaves almost no dipole to relax, which is the physical reason for the low loss. Used in demanding RF applications, but expensive, mechanically soft, and awkward to plate and bond.

FR-4: The ubiquitous PCB laminate has ε_r ≈ 4.2 to 4.5 and tan(δ) ≈ 0.015 to 0.025 at 1 GHz. Note that FR-4 is a flammability and reinforcement grade rather than a single specified material, so electrical properties vary substantially between suppliers and resin systems—a fact that regularly surprises designers who treat published values as guaranteed. The relatively high loss limits long traces to moderate frequencies.

Rogers RO4000 series: Hydrocarbon ceramic laminates that process much like FR-4. RO4003C is specified at ε_r = 3.38 ± 0.05 and tan(δ) = 0.0027 at 10 GHz, with a separate design value of 3.55 published for impedance calculation over 8 to 40 GHz; RO4350B is a flame-retardant counterpart at ε_r = 3.48 and tan(δ) = 0.0037. They occupy the middle ground between FR-4 and PTFE in both performance and cost.

Polyimide: Used in flexible circuits and high-temperature applications, with ε_r ≈ 3.4 to 3.6 and tan(δ) ≈ 0.002 to 0.008 depending on formulation. Polar groups in the molecular structure lead to higher loss than PTFE, and the material absorbs moisture readily.

Low-loss laminates with FR-4-like processing: Modern resin systems reach tan(δ) ≈ 0.002 to 0.008 while remaining compatible with conventional multilayer fabrication. These materials, rather than exotic substrates, carry most production links at 25 Gbps and beyond, and they are normally paired with low-profile copper so that the conductor does not become the limiting loss term.

Frequency Dependence

Dielectric properties vary with frequency due to dispersion. The Debye model describes relaxation for a single dipole species:

ε(ω) = ε_∞ + (ε_s - ε_∞)/(1 + jωτ)

where ε_s is the static permittivity, ε_∞ is the high-frequency limit, and τ is the relaxation time. Real materials often show multiple relaxation times, requiring more complex models (Cole-Cole, Havriliak-Negami, etc.).

For PCB materials in the microwave range, the permittivity typically decreases slightly with frequency (1-5% variation from DC to 10 GHz), while loss tangent may increase gradually. Accurate modeling for broadband digital signals requires frequency-dependent material parameters, usually obtained from vendor data or measurements.

Environmental Effects

Dielectric properties are sensitive to environmental conditions:

• Moisture absorption: Water has high permittivity (ε_r ≈ 80) and loss tangent. Even small amounts of absorbed moisture significantly increase both ε_r and tan(δ). Conformal coating or hermetic packaging protects against this.
• Temperature: Molecular mobility increases with temperature, generally increasing loss tangent while decreasing permittivity. Temperature coefficient of permittivity (TCε) is an important specification for temperature-stable designs.
• Aging: Some materials degrade over time through oxidation or chemical changes, increasing loss. Glass transition temperature and thermal cycling accelerate aging.

Causality Principle

Causality states that the impulse response h(t) of any physical system must be zero for t < 0. In other words, there is no output before an input is applied. This requirement is so fundamental that violations would contradict basic physical law.

When we transform to the frequency domain, causality places strict constraints on the mathematical form of the frequency response. Specifically, if a frequency response is to represent a physically realizable system, its real and imaginary parts (or equivalently, magnitude and phase) cannot vary independently—they must be related through specific integral transforms.

Kramers-Kronig Relations

The Kramers-Kronig relations are mathematical expressions of causality in the frequency domain. They state that the real and imaginary parts of any causal, linear frequency response are Hilbert transforms of each other.

For a complex response function H(ω) = H'(ω) + jH''(ω):

H'(ω) = (1/π) × P∫[H''(ω')/(ω' - ω)]dω'
H''(ω) = -(1/π) × P∫[H'(ω')/(ω' - ω)]dω'

where P denotes the Cauchy principal value of the integral, and the integration is over all frequencies from -∞ to +∞.

These relations mean that if you know the real part of the response at all frequencies, you can calculate the imaginary part (and vice versa). This is not merely a mathematical curiosity—it has practical implications for measurement and modeling.

Application to Transmission Lines

Complex Permittivity: The complex permittivity ε(ω) = ε'(ω) - jε''(ω) must satisfy Kramers-Kronig relations:

ε'(ω) - ε_∞ = (2/π) × P∫[ω'ε''(ω')/(ω'² - ω²)]dω'
ε''(ω) = -(2ω/π) × P∫[(ε'(ω') - ε_∞)/(ω'² - ω²)]dω'

This means you cannot arbitrarily choose both the dielectric constant and loss tangent as functions of frequency—if one is specified, the other is determined by causality. This constraint is important when developing broadband dielectric models from limited measurement data.

Propagation Constant: The complex propagation constant γ(ω) = α(ω) + jβ(ω) must also obey Kramers-Kronig relations. This means attenuation and phase response are not independent. For example, a transmission line with frequency-dependent attenuation will necessarily exhibit dispersion (frequency-dependent phase velocity), even if the dielectric is non-dispersive.

Characteristic Impedance: The complex characteristic impedance Z₀(ω) similarly obeys these relations. A frequency-dependent resistance necessarily implies a reactive component, which is why lossy lines always have complex characteristic impedance.

Practical Implications

Model Validation: Any electrical model that violates Kramers-Kronig relations is non-causal and therefore unphysical. When developing or using transmission line models, verifying KK compliance ensures physical consistency. Models that fit measured data well at discrete frequencies but violate KK relations will produce non-physical results in time-domain simulation.

Measurement Extrapolation: If you measure one component of a complex parameter (say, dielectric loss) over a limited frequency range, Kramers-Kronig relations can help extrapolate the real part (permittivity) beyond the measured range, or validate the consistency of measurements.

Dispersion and Loss Connection: The relations prove that any frequency-dependent loss must produce dispersion. This is why lossy transmission lines always exhibit pulse spreading—the absorption that causes loss necessarily makes different frequencies travel at different velocities. You cannot have loss without dispersion (though you can have dispersion without loss in lossless but dispersive media).

Minimum Phase Systems: A causal, stable system whose transfer function has neither poles nor zeros in the right half of the complex frequency plane is called minimum phase. For such systems the phase response is uniquely determined by the magnitude response through a Hilbert transform. A uniform transmission line departs from minimum phase only by its pure delay term, so once that delay is factored out, phase can often be reconstructed from measured magnitude alone—a useful check on the consistency of S-parameter data.

Physical Interpretation

Why does causality create these mathematical constraints? Consider a pulse propagating through a transmission line. The leading edge of the pulse must be affected by the line's properties at all frequencies, because even a sharp edge has a broad spectrum. The way the line attenuates different frequency components (loss) necessarily affects how those components add up in phase (dispersion) to maintain the constraint that the pulse cannot arrive before it was sent.

The clearest illustration is the ideal brick-wall filter. Specify a response that passes every frequency below a cutoff perfectly and blocks everything above it, and the resulting impulse response is a sinc function extending infinitely in both directions—including before the impulse arrives. The magnitude was chosen freely, the phase was left unconstrained, and causality was the casualty. Every physically realizable channel, including every transmission line, has a gradual band edge whose shape and phase are locked together by the Kramers-Kronig relations.

Verification in Practice

When working with measured or modeled transmission line data:

• Verify that tabulated frequency-dependent parameters satisfy KK relations (many simulation tools include KK checking)
• Be suspicious of material data showing loss without corresponding dispersion, or vice versa
• When curve-fitting models to measurements, impose KK constraints to ensure physical realizability
• For broadband digital signal integrity, confirm that the transmission line model is causal across the entire frequency range of interest, from DC to several times the Nyquist frequency
• In time-domain simulation, causality violations appear as pre-shoot—a response that begins before the stimulus arrives—which is a reliable sign of a non-physical model or of an S-parameter set that was extrapolated carelessly toward DC

SPICE modeling for signal integrity covers how these causal, frequency-dependent line models are constructed for circuit simulation.

High-Speed Digital Design

• Impedance control: Maintaining characteristic impedance along signal paths minimizes reflections and ensures signal integrity; see impedance control for stackup and tolerance practice
• Length matching: Using propagation delay calculations to match trace lengths in differential pairs and parallel buses
• Loss budgeting: Accounting for both conductor and dielectric losses when predicting channel performance at high data rates
• Dispersion management: Understanding how frequency-dependent parameters affect rise time and eye diagram closure

RF and Microwave Systems

• Impedance matching networks: Using transmission line transformations and quarter-wave sections
• Filter design: Exploiting distributed element behavior to create filters, couplers, and power dividers
• Resonant structures: Using half-wave and quarter-wave resonances
• Calibration and de-embedding: Removing transmission line effects from measurements

Power Integrity

• Power distribution network design: Managing impedance of power delivery traces at high frequencies
• Decoupling strategy: Accounting for inductance and propagation delay in capacitor placement
• Return path management: Understanding how current return paths affect effective inductance

Measurement and Characterization

Every quantity in this article is measurable, and two instruments dominate. Time-domain reflectometry launches a fast edge and reads the reflected waveform, mapping impedance against position along the line: a step up in the trace indicates a local impedance increase, and the round-trip time locates it. Vector network analysis measures S-parameters across frequency, from which propagation constant and characteristic impedance are extracted, usually after de-embedding the fixture with calibration structures.

The two views are Fourier transforms of one another, and disagreement between them usually signals a calibration or de-embedding error rather than a property of the line. Time-domain measurements covers the technique and its pitfalls, notably that the spatial resolution of a reflectometer is set by the rise time of its own step, not by the sample rate of its receiver.