Phase Velocity vs Group Velocity on a Transmission Line

On a lossless TEM line in a uniform dielectric, vp = vg = c/sqrt(εr), which is 0.690c for PTFE. In a hollow waveguide β is not proportional to ω, so vp > c, vg < c and vp·vg = c².

Engineer’s note: Phase velocity matters, but with wideband modulated signals like OFDM what really counts is group velocity, or how flat the group delay is. If different frequencies in the band arrive at different times, you get intersymbol interference and EVM falls apart. Always check phase linearity across a filter's passband.

For practitioners

A transmission line has one propagation constant β(ω). Two different velocities come out of it. One is the ratio ω/β. The other is the slope dω/dβ. On a coaxial cable or stripline the two are the same number, so engineers often treat "the velocity" as a single property of the line. Hollow waveguides, and any structure where β is not proportional to ω, are different. There the two velocities split apart, and confusing them leads to wrong delay numbers and broken de-embedding.

Two velocities from one function β(ω)

Take a single-frequency wave on a line, v(z, t) = cos(ωt − βz). A point of constant phase satisfies ωt − βz = constant, so it moves at the phase velocity:

vp=ωβv_p = \frac{\omega}{\beta}

A real signal is never a single frequency. A narrowband signal is a carrier inside an envelope, and its spectrum covers a small range of ω around the carrier. Expand β(ω) to first order around the carrier frequency. The envelope then moves at the group velocity:

vg=dωdβ=(dβdω)−1v_g = \frac{d\omega}{d\beta} = \left( \frac{d\beta}{d\omega} \right)^{-1}

Phase velocity depends on the value of β at one frequency. Group velocity depends on how β changes with frequency. The two are equal only when β is proportional to ω, so that ω/β and dω/dβ are the same constant.

The TEM line: one number for both

For a lossless TEM line with inductance L and capacitance C per unit length:

β=ωLC\beta = \omega \sqrt{LC}

If the dielectric is uniform and non-dispersive, L and C do not depend on frequency. β is then a straight line through the origin, and the ratio equals the slope:

vp=vg=1LC=cμrεrv_p = v_g = \frac{1}{\sqrt{LC}} = \frac{c}{\sqrt{\mu_r \varepsilon_r}}

For a non-magnetic dielectric (μr = 1) this becomes c/sqrt(εr).

Worked example, PTFE dielectric (εr = 2.1, μr = 1):

  • sqrt(2.1) = 1.449
  • vp = vg = c/1.449 = 0.690c ≈ 2.07e8 m/s
  • Delay of a 1 m matched line: τg = l/vg = 1/2.07e8 ≈ 4.83 ns
  • At 1 GHz: β = 2π·1e9 / 2.07e8 ≈ 30.4 rad/m. The wavelength is 0.207 m, so 1 m holds about 4.83 wavelengths and the transmission phase is about −1739°.
  • Phase slope: 4.83 ns × 360°/ns per GHz ≈ 1.74° per MHz, falling with frequency.

Phase delay (−φ/ω = l/vp) and group delay (−dφ/dω = l/vg) give the same 4.83 ns here. That match is what "non-dispersive" means in practice.

Reading the dispersion diagram

The clearest way to see both velocities is to plot ω against β. Phase velocity is the slope of the line from the origin to the operating point. Group velocity is the slope of the tangent to the curve at that point.

0kc2kc3kcωc2ωc3ωcβ (rad/m)ω (rad/s)ω = cβspeed cTEM line, PTFE εr = 2.1:vp = vg = c/sqrt(2.1)= 0.690c ≈ 2.07e8 m/swaveguide (air-filled):β = sqrt(k² − kc²)vp·vg = c²line from origin:vp = ω/β > cvg = dω/dβ < cP
Dispersion diagram of ω against β for a PTFE TEM line, the speed-c line ω = cβ and an air-filled waveguide mode. Computed from the formula.

The PTFE TEM line is a straight line through the origin. At every point the chord from the origin and the tangent are the same line, so vp = vg = 0.690c. The dashed line ω = cβ marks speed c. The PTFE line is steeper in β, which means it is slower than c. The waveguide curve behaves differently, as the next section shows.

The hollow waveguide: vp above c, vg below c

In an air-filled hollow waveguide, each TE or TM mode has a cutoff wavenumber kc. Above cutoff:

β=k2−kc2,k=ωc\beta = \sqrt{k^2 - k_c^2}, \quad k = \frac{\omega}{c}

β starts at zero at the cutoff frequency ωc = c·kc and is not proportional to ω. Differentiate β² = ω²/c² − kc²:

2βdβdω=2ωc2  ⇒  vg=c2βω=c2vp2\beta \frac{d\beta}{d\omega} = \frac{2\omega}{c^2} \;\Rightarrow\; v_g = \frac{c^2 \beta}{\omega} = \frac{c^2}{v_p}

So vp·vg = c² for every mode. Written in terms of frequency:

vp=c1−(fc/f)2,vg=c1−(fc/f)2v_p = \frac{c}{\sqrt{1 - (f_c/f)^2}}, \qquad v_g = c\sqrt{1 - (f_c/f)^2}

At f = 2fc, sqrt(1 − 0.25) = 0.866. That gives vp = 1.155c and vg = 0.866c, and their product is c². Close to cutoff, vg goes toward zero and vp grows without limit. Far above cutoff, both approach c. On the diagram, the curve starts at ωc on the ω axis and bends toward the dashed ω = cβ line. The chord from the origin to any point on the curve is steeper than c, and the tangent at that point is shallower than c.

A phase velocity above c breaks no physical law. A crest carries no information. Energy and modulation travel with the envelope at vg, and vg stays below c.

Group delay: what a VNA actually measures

A VNA measures the transmission phase φ = ∠S21 and computes group delay from its slope:

τg=−dϕdω\tau_g = -\frac{d\phi}{d\omega}

For a matched line of length l, φ = −βl, so:

τg=ldβdω=lvg\tau_g = l \frac{d\beta}{d\omega} = \frac{l}{v_g}

Group delay therefore gives you group velocity. A slope measurement also avoids the 2π ambiguity in the absolute phase. Phase delay needs the full unwrapped phase, −φ/ω, and gives l/vp. On a TEM line in a uniform dielectric, both numbers are the same. On a waveguide section they are not.

A single "electrical length" or velocity factor describes a line completely only when vp = vg. If a fixture includes a waveguide section or another dispersive structure, a delay value fitted to the phase at one frequency will not match the phase slope, and the reverse is also true. Correct the fixture across frequency with a full de-embedding model, not with one delay number.

When the difference matters

  • Dispersive guides. In hollow waveguide, vp and vg can differ by a large factor near cutoff. Delay, length and phase calculations must state which velocity they use.
  • Frequency-dependent materials. If εr changes with frequency, β is no longer proportional to ω. The TEM result c/sqrt(εr) then holds only point by point, and vg differs from vp.
  • Wideband signals. When vg changes across the signal band, different parts of the spectrum arrive at different times and the envelope spreads out. A narrowband signal sees only one vg and keeps its shape.
  • De-embedding and port extension. A port extension based on a single delay assumes a non-dispersive line. See Port Extension and De-Embedding for Fixtures and PCB Traces.
  • Cable phase after calibration. Phase across a cable scales with β·l at each frequency. This is why a small change in effective length, for example from flexing, shows up first as a phase change. See Test Cable Flexure and Phase Stability After Calibration.

In plain words · for students and beginners

Phase velocity and group velocity in plain words

An everyday picture

Picture a long line of people doing a stadium wave. One person stands up, then the next, and so on. The point where people are standing moves along the line. That moving point is like a wave crest, and its speed is like phase velocity. Now suppose a group of fans walks along the aisle carrying a banner. The banner is like the envelope, the overall shape that carries the message. Its speed is like group velocity.

Where the picture breaks down: in a stadium the two speeds have nothing to do with each other. On a transmission line both speeds come from the same quantity, β (the "phase constant", which tells you how many radians the wave's phase changes per metre). Phase velocity is ω divided by β, where ω is the angular frequency in radians per second. Group velocity is how fast ω changes when β changes.

amplitude (normalized)time t1+10−10position z (m)envelope peakcrestcrest (phase velocity vp)envelope (group velocity vg,carries the signal)amplitude (normalized)time t2 (later)+10−10position z (m)crestOn a TEM line in a uniformdielectric both move together:vp = vg
A wave crest and its envelope moving along a dispersive waveguide at two times, showing phase velocity vp and group velocity vg. Concept drawing.

The drawing shows a waveguide, where the two speeds are different. The crest dot moves further than the envelope peak in the same time. On an ordinary coaxial cable, both move together.

The same example, step by step

Take a cable filled with PTFE (a common plastic insulator). Its εr (relative permittivity, which tells you how much the material slows the wave compared with vacuum) is 2.1.

  1. Take the square root: sqrt(2.1) = 1.449.
  2. Divide the speed of light by it: c/1.449 = 0.690c.
  3. In metres per second that is about 2.07e8 m/s.
  4. A 1 m cable then delays the signal by 1 / 2.07e8 ≈ 4.83 ns.

On this cable, phase velocity and group velocity are both 0.690c. You do not need to choose between them.

For a hollow metal waveguide working at twice its cutoff frequency (cutoff is the lowest frequency the guide can carry):

  1. Compute sqrt(1 − (1/2)²) = sqrt(0.75) = 0.866.
  2. Group velocity = 0.866c, which is slower than light.
  3. Phase velocity = c/0.866 = 1.155c, which is faster than light.
  4. Check: 1.155 × 0.866 = 1.000. The product is always c².

Mistakes beginners make

  • Thinking "faster than light" is impossible, so the formula must be wrong. The crest carries no message. The message rides on the envelope, which moves at vg, below c.
  • Using one velocity factor for everything. A single number works on a cable with a uniform dielectric. It does not work in a waveguide or when the material changes with frequency.
  • Reading group delay as phase delay. A network analyzer (a VNA, an instrument that measures how a device reflects and passes signals) reports group delay from the slope of the phase. That gives l/vg, not l/vp.

Remember

  • Phase velocity vp = ω/β is the speed of a crest. Group velocity vg = dω/dβ is the speed of the envelope that carries the signal.
  • On a lossless TEM line in a uniform dielectric they are equal: c/sqrt(εr), or 0.690c for PTFE.
  • In a hollow waveguide vp·vg = c², so vp > c and vg < c.

Frequently asked questions

What is the difference between phase velocity and group velocity?

Phase velocity vp = ω/β is the speed of a single wave crest. Group velocity vg = dω/dβ is the speed of the envelope of a narrowband signal, and the envelope is what carries information.

Are phase velocity and group velocity equal on a coaxial cable?

Yes, if the line is lossless and the dielectric is uniform and non-dispersive. Then β = ω·sqrt(LC), so both velocities equal 1/sqrt(LC) = c/sqrt(εr), which is 0.690c for PTFE.

How can phase velocity in a waveguide be faster than light?

In a hollow waveguide vp·vg = c², so vp is above c while vg is below c. A crest carries no information. The signal travels with the envelope at vg, which stays below c.

Does a VNA group delay measurement give phase velocity or group velocity?

Group delay is τg = −dφ/dω, and for a matched line of length l it equals l/vg. On a TEM line in a uniform dielectric this is also l/vp, but on a dispersive structure the two are different.

References and further reading