When wires give up
Every circuit you have built so far rests on a quiet assumption: a signal travels on two conductors — a wire and its return, a transmission line with a signal path and a ground. A coax cable, a pair of traces on a board, the power line on a pole: all the same idea, a wave guided by the gap between two pieces of metal. It works beautifully from DC up through the gigahertz, and most of electrical engineering lives happily inside it. But push the frequency high enough — into the tens of gigahertz, the realm of radar, satellite uplinks, and 5G millimetre-wave — and that trusty two-conductor cable starts to choke. The reasons stack up, and together they are why a radar engineer will reach not for a cable but for a hollow metal pipe.
Three things go wrong at once. First, loss explodes. As you'll see in a moment, high-frequency current crowds into a paper-thin layer of the conductor, so the effective resistance climbs with frequency and the signal bleeds away as heat — a long coax run at 30 GHz can lose most of its power in a few metres. Second, power handling collapses. Pack a kilowatt of radar power into a thin coax and the centre conductor, already starved of cross-section by the skin effect, simply melts or arcs across the dielectric. Third, the dielectric itself leaks — the plastic insulation that holds a coax together has its own losses that worsen sharply with frequency. A hollow pipe with nothing inside but air sidesteps all three.
The hollow pipe and its locked door: cutoff frequency
Picture a rectangular brass tube, open at both ends, like a length of square downpipe. Launch a microwave into one end and it doesn't travel straight down the bore — it zig-zags, bouncing off the side walls at an angle, the way light skitters down a hall of mirrors. Each bounce is a reflection off the metal, and the wave works its way along the pipe as the sum of all those zig-zag reflections. This is the heart of a waveguide: a wave doesn't flow *through* it the way current flows through a wire; it bounces down it, guided by walls it can never penetrate.
And here is the magic, the single fact that defines waveguides: a waveguide is a high-pass filter for the air inside it. Because the wave has to fit *across* the pipe — its electric field must vanish at the conducting walls, so at least half a wavelength must span the width — there is a lowest frequency that can squeeze through. Below this cutoff frequency, the wave simply will not propagate. Send a signal below cutoff into the pipe and it doesn't travel; it dies away exponentially within a few centimetres, reflected straight back out. The pipe is a door that only opens for waves short enough to fit. For a standard WR-90 waveguide (about 23 mm wide, used across the X-band) the cutoff sits near 6.5 GHz — feed it 4 GHz and nothing comes out the far end at all.
RECTANGULAR WAVEGUIDE — a wave zig-zags down the bore
metal wall (top)
==========================================
\ /\ /\ /\ /\
\ / \ / \ / \ / \
\ / \ / \ / \ / \ -> propagation
\/ \/ \/ \/ \/
==========================================
metal wall (bottom)
|<------ width 'a' ------>|
CUTOFF: fc = c / (2a) (dominant TE10 mode)
WR-90: a = 22.9 mm -> fc ~ 6.56 GHz
f < fc -> wave DIES (evanescent, no travel)
f > fc -> wave PROPAGATES down the pipe
Useful band sits ABOVE cutoff (X-band: 8.2-12.4 GHz)Modes: the shapes a trapped wave is allowed to take
On a two-conductor cable a wave can have essentially one field pattern — electric field straight across from signal to ground, magnetic field looping around, both perfectly transverse (the TEM mode). A hollow waveguide is different and richer: with only one conductor, a pure TEM wave is impossible, so the trapped wave is forced to take on one of a discrete family of standing-field patterns called modes. Like the resonances of a flute, only certain shapes 'fit' across the pipe — and each one has its own cutoff frequency. The lowest, simplest pattern is the dominant mode (TE₁₀ in a rectangular guide), with a single, gentle hump of electric field across the width.
Engineers work hard to operate a waveguide in just its dominant mode, in the band above the dominant cutoff but below the next mode's cutoff. Why so fussy? Because different modes travel at different speeds, so if two modes carry the same signal they arrive smeared apart in time — multimode dispersion, a kind of self-inflicted echo. Keeping to one mode keeps the signal crisp. (You've already met this exact problem in a different guise: an optical fibre is a waveguide for light, and 'single-mode' versus 'multimode' fibre is the same dominant-mode discipline, just at 200 THz instead of 20 GHz.)
The skin effect: current that flees to the surface
Now to the second half of the frontier — and the reason cables suffer in the first place. At DC, current spreads itself evenly across the whole cross-section of a wire, using every square millimetre of copper. But raise the frequency and something counter-intuitive happens: the current abandons the middle of the wire and crowds into a thin shell at the outer surface. This is the [[ee-skin-effect|skin effect]], and the deeper you go below the surface, the less current flows. At high enough frequency the core of a thick copper bar carries almost no current at all — it might as well be hollow.
Why does current flee outward? The cause is Faraday's law, the same induction you met earlier in the track. The current flowing in the core sets up a magnetic field, and when that current is *changing* (as AC always is), the changing field induces little swirling eddy currents inside the conductor. By Lenz's law these eddies oppose the change — and the geometry works out so that they cancel the current in the centre and reinforce it at the surface. The faster the current changes, the more violently the eddies fight it, and the thinner the conducting skin becomes. Self-induction, in short, evicts the current from the interior.
We measure the thinness of that conducting shell with the skin depth δ — the depth at which the current has fallen to about 37% of its surface value. It shrinks with the square root of frequency, and in copper it is startlingly small. At 50/60 Hz mains it's about 9 mm (so a fat power conductor is fine). But climb the spectrum and it collapses: about 2 mm at 1 kHz audio, 66 µm at 1 MHz, just 2 µm at 1 GHz — thinner than the copper foil on a circuit board, thinner than a strand of spider silk. At microwave frequencies, the current is riding a skin you'd struggle to see under a microscope.
SKIN DEPTH IN COPPER — current crowds into a thin shell
DC: High frequency:
.===================. |####| |####|
|###################| |# current #|
|### full cross- ###| |# only in #|
|### section used###| |# the SKIN #|
|###################| |####| |####|
'===================' <-- dead core (no current) -->
skin depth d = sqrt( rho / (pi * f * mu) )
Copper: 60 Hz -> d ~ 8.5 mm
1 kHz -> d ~ 2.1 mm
1 MHz -> d ~ 66 um
1 GHz -> d ~ 2.1 um (!!)
d falls as 1/sqrt(f): 100x the frequency -> 10x thinner.
AC resistance Rac ~ rises as sqrt(f) once d << radius.What skin effect does to everything you build
Here's why this matters far beyond exotic microwave gear. A conductor's resistance is its resistivity times length, divided by the cross-sectional area the current actually uses. The skin effect *throttles that area* — at 1 GHz, only the outer 2 µm of a wire carries current, so a thick cable behaves as if it were a thin-walled tube. The dead copper core is doing nothing. The result: AC resistance rises with the square root of frequency, and once the skin is much thinner than the wire's radius, doubling frequency multiplies loss roughly by 1.4. This is the deep reason a transmission line gets lossier the higher you go, and why a coax that's perfect for 1 GHz is hopeless at 30 GHz.
The consequences ripple into work you'll do every day. On a PCB, a multi-gigabit serial link's trace loss is dominated by skin effect — which is why high-speed boards plate their traces smooth (surface roughness adds even more loss when the current rides only the surface) and route critical links on outer layers. In power and audio inductors and transformers, fat single wires are useless at high frequency, so engineers use [[ee-skin-effect|litz wire]]: a rope of many thin, individually-insulated strands, woven so each strand spends equal time near the surface. Because every strand is thinner than a skin depth, the current spreads across all of them and the effective resistance plummets. It's a beautiful trick — beat the skin effect by making every conductor *all skin.*
- Long cable runs lose more the higher the frequency — the √f loss climb is why a 30 GHz signal travels metres on coax but kilometres on optical fibre, whose photons don't care about copper skin at all.
- PCB high-speed links budget for skin-effect loss and surface roughness; designers smooth the copper and shorten the trace, because every dB of loss closes the eye diagram a little more.
- Magnetics below ~1 MHz use litz wire — many thin strands so the current uses the full copper instead of a useless thin shell, cutting AC resistance dramatically.
- Waveguides turn skin effect to advantage — the wave only ever touches a few microns of the inner wall, so a guide is often silver- or gold-plated *just on the surface*, since that's the only metal the current sees.
The whole track, in one wave
Step back and look at the ladder you've climbed. You started with two invisible fields, electric and magnetic, anchored to charge and its motion. You watched them learn to regenerate each other and sprint off as a self-propagating [[ee-electromagnetic-wave|electromagnetic wave]]. You learned to trap that wave on two conductors as a [[ee-transmission-line|transmission line]], with its own [[ee-characteristic-impedance|characteristic impedance]] and its [[ee-reflection-coefficient|reflection coefficient]] measuring what bounces back at a mismatch. And now, at the frontier, you've seen the wave break free of the second conductor entirely — bouncing down a hollow waveguide above its cutoff — and you've seen why, on any real metal, the current itself hides in a microscopic skin. Static fields, Maxwell's equations, guided propagation: it is all one continuous story about where the energy lives and how to herd it.
And this rung is a launch pad, not a finish line. Everything ahead grows straight out of it. RF and microwave engineering is the art of matching impedances and shuttling waves through cables and waveguides without losing or reflecting them. Antennas are the deliberate opposite of a waveguide — structures built to *let* the wave escape into free space rather than trap it. High-speed digital links — every USB, PCIe, and Ethernet lane in your laptop — are transmission lines fighting skin-effect loss and reflections to push more gigabits down a copper trace. The same two fields, the same cutoff and skin physics, dressed for a hundred different jobs.