When a wire stops being a wire
The previous guide planted one unsettling idea: at high frequency a circuit's behaviour depends on how big it is compared to the wavelength of the signal travelling through it. Recall that wavelength is lambda = v / f, where v is how fast the wave moves — about two-thirds the speed of light on a typical cable, so roughly 2 x 10^8 m/s. Because that speed is finite, the far end of a connection always lags the near end: the signal is a thing in flight, not a value that appears everywhere at once. Once a connection is long enough for that lag to matter, the comfortable old assumption — that a wire is a perfect node with the same voltage at both ends — quietly breaks.
Engineers draw a fuzzy boundary with a rule of thumb: while a wire is shorter than about one-tenth of a wavelength you may keep treating it as an ideal lumped node, but longer than that you must treat it as a transmission line — a medium that guides a wave with delay. Put real numbers on it. At 1 kHz audio the wavelength on a cable is about 200 km, so a 1 m lead is 1/200,000 of a wavelength — utterly lumped, exactly as your DC and audio intuition expects. At 1 GHz the wavelength shrinks to about 20 cm, so even a 2 cm trace is already a tenth of a wavelength and behaves as a transmission line. The one-tenth figure is a convention, not a hard law; fussy designers use one-twentieth, and for digital it is the signal's rise time, not its clock frequency, that decides.
Here is the mental shift that makes everything else click. At DC and audio a wire is just a wire — a perfect handshake, voltage identical at both ends. Cling to that picture at RF and it will baffle you. The honest reframing is the one from the last guide: a transmission line is not a node, it is a journey. A pulse launched at one end is a real travelling disturbance that takes time to arrive and, as we are about to see, can come bouncing back. Take the journey seriously and the rest of this guide simply follows.
Characteristic impedance: the resistance a wave feels
To see where the magic number comes from, zoom into a cable and chop it into tiny slices. Each slice has a little series inductance — the wire's flywheel that resists any sudden change in current — and a little capacitance across to the return conductor — a tiny bucket that resists any sudden change in voltage. String thousands of these together and you have an LC ladder: series L, shunt C, series L, shunt C, on and on. Pour a voltage step into the front and it charges the first bucket through the first flywheel, which charges the next, and the next, so the disturbance rolls down the line at that finite speed we met above.
Now the surprise. As the wave rolls forward it must keep charging each fresh bucket, which means it draws a steady current from the source — even though the line wastes no energy and you have not yet said anything about what is at the far end. The ratio of the wave's voltage to that current is fixed by the ladder itself: Z0 = sqrt(L / C), the inductance per unit length over the capacitance per unit length. This is the characteristic impedance. It is measured in ohms and acts like a resistance to the advancing wave, yet no heat is made — the energy is stored and handed bucket to bucket. A worked example: a coax with L = 250 nH per metre and C = 100 pF per metre has Z0 = sqrt(250 x 10^-9 / 100 x 10^-12) = sqrt(2500) = 50 ohm.
So why 50 ohms in particular? It is a famous engineering compromise, not a law of nature. For an air-filled coax, lowest loss happens at around 77 ohm while the highest power handling peaks near 30 ohm, and 50 ohm sits in the workable middle — close enough to both that the industry simply standardized on it so every source, cable, connector, instrument and load could share one number and plug together. Video and TV instead use 75 ohm, the lowest-loss choice for a given cable diameter. Treat 50 ohm as a convention to honour, not a target handed down by physics, and the whole RF ecosystem clicks into place.
Reflections: the echo at the end of the line
The entire point of Z0 is what happens when the wave reaches the far end. If the load resistance equals Z0, the arriving wave finds exactly the voltage-to-current ratio it has been expecting all along — the load looks identical to more line stretching off to infinity — so the wave pours its energy in and nothing comes back. The line is matched. But if the load differs, the wave cannot fit: it carries a voltage and a current locked at the ratio Z0, the load demands a different ratio, and the only way to satisfy both is for some of the wave to bounce back toward the source. This is signal reflection, and it is the central headache of high-frequency wiring.
How much bounces is captured by one tidy formula. The reflection coefficient is Gamma = (ZL - Z0) / (ZL + Z0), where ZL is the load impedance and Gamma is the fraction of the voltage wave sent back. Three landmark cases anchor it. A matched load, ZL = Z0, gives Gamma = 0 — no echo, everything delivered. An open circuit, ZL = infinity, gives Gamma = +1 — the whole wave bounces back with the same polarity. A dead short, ZL = 0, gives Gamma = -1 — the whole wave bounces back inverted. Everything real lands in between: a 50 ohm line driving a 75 ohm antenna input has Gamma = (75 - 50) / (75 + 50) = 25 / 125 = 0.2, so 20 percent of the voltage echoes straight back toward the source. Notice how this overturns the matched-only world — the load now matters enormously.
This is not only a radio problem — you can see it on a logic board. A fast digital edge is rich in high frequencies, so an unterminated PCB trace IS a transmission line, and the echo shows up as ringing: overshoot and stair-steps on the waveform as the pulse rattles between the driver and the open-ended receiver. Drive a clean step into a long mismatched cable and watch the reflection arrive exactly one round trip later. That is precisely how a time-domain reflectometer, the cable-fault finder, works — it sends a pulse and times the echo to locate the break. The DC-intuition trap one more time: at DC an open end does nothing at all, but at speed an open end is a perfect mirror.
Standing waves and VSWR
When the forward wave and its reflection share the same cable, they interfere. At points along the line where they happen to add, you get a voltage peak; where they cancel, a null. Because the two waves travel in opposite directions, this interference pattern does not move — it stands still in space as a standing wave, with fixed bellies and nodes pinned along the cable. A perfectly matched line carries only the forward wave, so it is flat: no peaks, no nulls. The worse the mismatch, the deeper the ripple between peak and null becomes.
The standard yardstick for that ripple is the Voltage Standing Wave Ratio — the ratio of the largest voltage peak to the smallest null along the line: VSWR = (1 + |Gamma|) / (1 - |Gamma|). A perfect match, Gamma = 0, gives VSWR = 1:1, a dead-flat line. Our 50-into-75 example, Gamma = 0.2, gives VSWR = 1.2 / 0.8 = 1.5:1. A total reflection, |Gamma| = 1 from an open or a short, gives VSWR = infinity — all peaks and nulls, nothing delivered. VSWR is handy precisely because an antenna analyzer or a directional meter reads it directly, without you ever knowing the exact complex load.
reflection at the load: Gamma = (ZL - Z0) / (ZL + Z0) |Gamma| VSWR return loss power reflected ------- --------- ------------- --------------- 0.00 1.00 : 1 infinite 0 % <- perfect match 0.10 1.22 : 1 20 dB 1 % 0.20 1.50 : 1 14 dB 4 % <- 50 ohm line into 75 ohm 0.33 2.00 : 1 9.5 dB 11 % <- a "2:1" antenna 0.50 3.00 : 1 6 dB 25 % 1.00 infinite 0 dB 100 % <- open or dead short return loss(dB) = -20 x log10|Gamma| power reflected = Gamma^2
Return loss says the same thing in decibels: return loss (dB) = -20 x log10|Gamma|, and bigger is better — a 20 dB return loss means only 1 percent of the power bounces. Three practical bites follow. First, reflected power never reaches the load, so a humdrum 2:1 VSWR quietly throws away about 11 percent of your signal. Second, the standing-wave voltage peaks can rise above a cable's or a component's rating and arc over. Third, and most dangerous, the source no longer sees a friendly constant 50 ohm — it sees an impedance that swings with frequency, which can detune an oscillator or overheat and destroy a transmitter's output stage. That is why a ham operator frets over an antenna's VSWR before keying up.
What to do about it: terminate, match, and a word of honesty
The cures run from blunt to elegant. The bluntest is termination: put a resistor equal to Z0 at the end of the line (or in series at the source), and the wave is swallowed with no echo, exactly as if the cable ran on forever. RF benches live on 50 ohm terminations, and fast digital boards add series or parallel termination on their most critical traces. But when the load is fixed and simply is not 50 ohm — an antenna, a transistor's input — you cannot just bolt a resistor across it without wasting the signal, so you build a small impedance-matching network instead: a couple of reactive parts, an inductor and a capacitor, that transform the stubborn load up or down until it looks like Z0. The hose analogy fits perfectly — matching is choosing a hose that neither chokes the flow nor wastes it.
Designing that network from the algebra is fiddly, because the load is usually complex and changes with frequency, so RF engineers reach for the matching network's best friend: the Smith chart, a brilliant circular map that turns impedance matching into geometry. On it you can read off, by eye, which series or shunt part to add and how big. That whole craft is the next guide; here just hold the goal firmly in mind — push |Gamma| toward zero so the line runs flat and every available watt actually arrives at the load.